Atomic spectrometry update: review of advances in the analysis of metals, chemicals and materials

Eduardo Bolea-Fernandez a, Robert Clough b, Andy Fisher *b, Bridget Gibson c and Ben Russell d
aDepartment of Analytical Chemistry, Aragon Institute of Engineering Research (I3A), University of Zaragoza, Zaragoza, 50009, Spain
bSchool of Geography, Earth and Environmental Science, University of Plymouth, Plymouth, PL4 8AA, UK. E-mail: afisher@plymouth.ac.uk
cIntertek Sunbury Technology Centre, Shears Way, Sunbury, Middlesex, UK
dNational Physical Laboratory, Nuclear Metrology Group, Teddington, Middlesex, UK

Received 25th September 2024

First published on 18th October 2024


Abstract

This update covers the literature published between approximately June 2023 and April 2024 and is the latest part of a series of annual reviews. It is designed to provide the reader with an overview of the current state of the art with respect to the atomic spectrometric analysis of various metals, chemicals and materials. Data processing appears to be the hottest topic in many of the areas. This is especially true for LIBS and (TOF)-SIMS, where huge amounts of data can be acquired. Methods have been used to decrease the dimensions of the data whilst still retaining the most important information. This can then be input into a machine-learning algorithm so that the provenance of a sample, the sample type, or, in the case of TOF-SIMS data, a clear characterisation of the surface of the sample can be obtained while using less computing power and less processing time. Although these algorithms have been used for some years, their use is expanding into new areas. Another development is the combination of complementary techniques on the same instrument platform. This enables data from the two techniques to be obtained simultaneously and from the same spot on the sample. With regard to the different analytical techniques used, LIBS is continuing to increase in popularity, bolstering its reputation as being the rising superstar of the analytical world.



Atomic Spectrometry Update (ASU) – Call for new writers

The Atomic Spectrometry Update (ASU) editorial board is looking to increase the size of the production team (see our recent Editorial (https://doi.org/10.1039/D4JA90044H)). We are looking for writers working in the field of atomic spectroscopy to contribute to one of the six reviews: advances in environmental analysis; advances in the analysis of clinical and biological materials, food and beverages; advances in XRF; advances in elemental speciation; advances in the analysis of metals, chemicals and materials; and advances in atomic spectroscopy and related techniques. You will work as part of a team to help produce the review on an annual cycle. Full support will be provided. Writers are allocated specific topics within the review and all abstracts for the review are provided. This is a great way to keep abreast of the current developments in the field, potentially improve your scientific writing skills and to publish annually. If you are interested, have relevant experience and a good command of written English, please contact the ASU General Editor, Prof Steve Hill for more information (steve.hill@plymouth.ac.uk).

1. Introduction

This is the latest review covering the topic of advances in the analysis of metals, chemicals and materials. It follows on from last year's review1 and is part of the Atomic Spectrometry Updates series.2–6

Several reviews covering different aspects of analytical chemistry that discuss more than one sample type have been published. Rather than finding the same review being discussed in more than one place in this update, those reviews are commented upon here. Other papers, e.g., ones that report a new software package for a certain technique that could be applied to any sample type, will also be discussed in this section.

An example of the development of software was presented by Iro et al.7 These workers produced software called voxTrace, which is a voxel-based Monte Carlo ray-tracing code for the simulation of X-ray spectra. The authors explained that confocal μ-XRF using capillary optics is a very powerful technique that can undertake 3D analysis of numerous sample types including those of cultural heritage, materials science and even of biological origin. There is a problem with the quantitative interpretation because of the large computing capacity that may be needed to take into account the energy-dependent transmission properties of polycapillary half-lens optics, and the absorption and scattering effects inside the sample. The software simulates the measured spectra enabling consideration of effects such as secondary excitation and both elastic and inelastic scattering. Another important feature of the voxTrace software is an energy independent voxel size of the sample. This allows the modelling of surfaces and scans with a step size smaller than the confocal volume; hence enabling an investigation of samples with higher accuracy. Simulating spectra, instead of using peak deconvolution algorithms of single spectra and then processing absorption effects independently, can be advantageous. This is especially true for the investigation of samples with materials with overlapping line energies. The software enables many of the problems associated with confocal μ-XRF to be overcome in a relatively short time span and without the need for large computer resources. Instead, the software requires only a standard graphics processing unit and Compute Unified Device Architecture.

There is no doubting that LIBS is the rising superstar of the atomic spectrometry family. It has found use in virtually all sample types, is capable of on-line or operando analysis and may undertake stand-off analysis of samples that are hazardous. In addition, with the advent of portable instruments, it may be taken into the field where it can supply data within minutes rather than hours if samples need to be returned to the laboratory. Although the signal is very dependent on sample type, leading to the requirement of matrix-matched samples for calibration, much research has been put into developing calibration models to try an improve accuracy. It is unsurprising that several reviews of different aspects of LIBS have been presented during this period.

One such review was presented by Guo et al.8 (89 references, in Chinese). This focussed on work undertaken over the last five years in the areas of coal analysis, metallurgical analysis, biomedical samples and the analysis of water. The review discussed the analysis of coal in terms of quality control and environmental impact. Also discussed were the economic savings of on-line analysis in the metallurgical industry. Although the bulk of the text is in Chinese, the abstract, figures and table of biomedical applications are in English.

Another example was presented by Krolicka et al.9 who reviewed, with the aid of 109 references, the application of LIBS to the depth-profiling of multi-layer and graded materials. As well as tables of applications for “inorganic materials” (including coated steels, nuclear components, photovoltaic materials, battery components and archaeological pottery) and polymers, the authors also discussed, at length, the importance of crater morphology that can affect the LIBS signal significantly during depth-profiling studies. The conclusions and future work section highlighted the abilities of the technique and the advantages it has over some other techniques. It also envisaged that in the future, the problems of calibration will be overcome by employing a calibration-free approach. In addition, remote analysis (as demonstrated by the Mars Rover), at-line industrial analysis, identification of materials of unknown composition (e.g., scrap metal classification, etc.) and multiple instrumentation types integrated together (e.g., Raman and LIBS, which provide complementary information from one platform) are all predicted to become commonplace.

A very lengthy review by Jin et al.10 discussed the experimental and applied research of double-pulse LIBS. Double-pulse LIBS is known to have numerous advantages over single-pulse LIBS, the foremost of which is enhanced sensitivity. The review cited 186 references and had some of the applications described in tables. The review was split into numerous sections. After an introduction, the various configurations of DP-LIBS were described, i.e., confocal, orthogonal re-heating, orthogonal pre-ablation and cross-type. The advantages and disadvantages of each were provided. A large section discussing the influence of each of the experimental parameters on accuracy and sensitivity was then provided. This section included sub-sections discussing the influence of pulse widths, laser wavelengths and energies, the laser profile, delay times and use of a vacuum or an inert gas atmosphere. Sections on underwater LIBS, remote LIBS and surface mapping were also given. Methods of enhancing the LIBS signal including nanoparticle-enhanced LIBS and magnetic confinement were also discussed. The applications sections covered topics such as geological exploration and environmental monitoring, metallurgical industry, energy industry (including nuclear, solar and coal) and biological samples.

A paper entitled “Spectroscopic methods for isotope analysis of heavy metal atoms: a review” was presented by Zhang et al.11 The review contained 111 references and discussed the progress of atomic spectroscopy for heavy metal isotopic analysis. The obvious research area would be the nuclear industry. However, other areas of research, e.g., environmental monitoring, biomedicine, archaeological artefacts, were included. The review split the techniques into two main classes, which it termed “linear spectroscopic” and “non-linear spectroscopic” techniques. The “linear” ones included techniques such as atomic emission (including ICP-OES and LIBS), atomic absorption and laser-excited atomic fluorescence spectrometry. The non-linear techniques were saturated absorption spectrometry, four-wave mixing spectrometry and Doppler-free two-photon spectrometry. Interestingly, the use of assorted mass spectrometric techniques was left out of the review. The principles, advantages and disadvantages, instrumentation required as well as the applications of each of the techniques were presented. Many of the applications were tabulated for easy reference.

A paper by Gonzalez et al.12 was not really a review, but more of a beginner's guide to the capabilities of XRF. This included a historical perspective, theory of the origins of different X-ray lines, the sensitivity and working range compared with other analytical techniques and Rayleigh and Compton scatter. Instrumental hardware such as detectors and X-ray tubes were also discussed. The applications section is somewhat brief and is by no means a comprehensive review. Instead, it gave examples of the areas in which it can be used and gives an indication of how many applications in each area have been published.

Atom counting with accelerator mass spectrometry” was the title of a review containing 577 references presented by Kutschera et al.13 This very long review covered the history of the technique from its development through to the present, the theory, the types of accelerator and listed the facilities throughout the world. The largest section detailed the different application areas that uses the technique. These include archaeology, where it may be used for radiocarbon dating, dating using 41Ca and isotope ratio measurements. Other areas of research that were reviewed included biological research, atmospheric science, oceanographic science, lithospheric science and astrophysics.

2. Metals

2.1. Ferrous metals

As always, this has proved to be a popular area of research with the majority of interesting papers involving LIBS. Although many papers continue to report the development of methods to improve calibration accuracy, there have been others describing different methods of improving the sensitivity of LIBS analyses, methods for decreasing problems arising from self-absorption, etc. The classification of different steels using data obtained from LIBS analyses has also continued to be of interest. These studies normally require a data reduction process to eliminate redundant areas of the spectra or to remove interfering species, followed by the training of a model using several well-characterised samples with more samples being used as a test of the model and finally the analysis of some “real” samples. Other analytical techniques have also been utilised for the analysis of ferrous metals and alloys. This is especially true for those studies that have reported insights into corrosion mechanisms. Here, techniques such as XPS dominate.

In recent years numerous methods have been reported that have attempted to improve the accuracy of the calibrations for LIBS analyses. An international comparison entitled “Quantification of alloying elements in steel targets: the LIBS 2022 regression contest” was undertaken.14 The study comprised the determination of two major (Cr and Ni) and two minor (Mn and Mo) components in 15 steel samples. A total of 21 laboratories were invited to participate with each being sent spectra from 42 steel targets so that they could construct their own regression models. The methods used by the three best-performing laboratories were discussed in the paper. These were a single linear partial least-squares model and two artificial neural network (ANN) regressions. Various strategies for selecting the emission lines used and spectral wavelength range were reported. Similarly, various spectral normalisation and data augmentation strategies were also presented.

The potential of using LIBS near or at the production line is one of its main advantages over other techniques. However, the problems of poor calibration are still not resolved completely, and hence inaccurate data can be obtained if analysis is undertaken at different pressures, temperatures, a slightly different sample is used, etc. There have therefore been several studies addressing this. An example by Gu et al.15 determined Mn and Si in 13 certified steels using calibration-free LIBS. After careful optimisation of the operating conditions (delay times, gate width, etc.) the intensities underwent pre-processing to identify and extract the features required and then input to the calibration algorithms. When the data that had undergone no pre-processing were input into a traditional calibration method, it produced calibration curves with very poor regressions, e.g., 0.68 for Si and only 0.33 for Mn. However, after data pre-processing, when input to a multivariate linear regression method, these improved slightly to 0.797 and 0.702 for Si and Mn, respectively. These improved further to 0.996 for both Mn and Si when the data were input to deep neural network (DNN) and to 0.990 (for Si) and 0.998 (for Mn) when input to the text convoluted neural network (Text CNN) algorithm. This led to relative errors of between 0 and 14.38% for the Si and between 0.915 and 23.03% for Mn, with most values for both being <10% for the DNN algorithm. These were improved further to 1.053 and 9.302% for Si and 0 and 9.4% for the Mn using Text CNN.

Another example of a study using chemometric methods to improve the accuracy of the calibration was presented by Lin et al.16 who used PLSR, Extreme Random Tree and Random Forest to treat the data obtained from the LIBS analysis of steel. A univariate calibration of the analyte (Cr) at 425.43 nm was also undertaken. A total of 30 spectra were obtained for each sample and each spectrum was the average of three individual laser shots. Analysis of eight steels was undertaken with the results of six being used to train the models and the other two being used as test samples. Before the data was input to the various algorithms, it underwent pre-processing using wavelet transform to remove noise and to re-set the baseline. The quality of the calibrations was assessed using statistical tools such as the fitting coefficient (R2), root mean square error of calibration (RMSEC), root mean square error of prediction (RMSEP) and average relative error values. For the Extreme Random Tree model (the focus of the study), R2 was 0.9984, RMSEC was 0.0274 wt%, RMSEP was 0.0299 wt% and the average relative error was only 2.43%. This represented significant improvements on the other models, especially over the univariate model and enabled the authors to conclude that their new model offered the possibility of accurate determination of Cr in steels.

Another paper to report the use of wavelet transform to process the data was presented by Zhao et al.17 These authors used LIBS and laser-induced fluorescence (LIF) to determine Cr, Mo and Ni in low-alloy steels and then processed the data using discrete wavelet transform (using Daubechies wavelet with wavelet function db7) to diminish background signals and noise. A brief explanation of how the algorithm works was provided to help the reader. The processing improved R2 values from 0.976, 0.981 and 0.965 for Cr, Mo and Ni, respectively without any processing to 0.995, 0.997 and 0.993. The RMSEC values also decreased significantly indicating that the process offered substantial improvements in both model accuracy and analytical accuracy.

Belkov et al. have published two papers in the area of multivariate calibration for LIBS analysis of low-alloy steel reference samples.18,19 In both cases, the analytes of interest were C, Cr, Cu, Mn, Ni and Si and were determined using a low-resolution spectrometer (resolution of 0.4 nm with a spectral step of 0.1 nm). In the latter case, data pre-processing was undertaken by normalisation of the signals to the Fe signal at 252.0609 nm and by baseline correction. These processed data were then input to a partial least-squares (PLS) method. The models were quantitative for all six analytes, In the former paper,18 three methods of data processing were examined. These were: method of ranking spectral variables by their correlation coefficient with the sought parameter, a successive projection algorithm and an original modification of the method of searching combination moving window. The paper discussed these in some detail. Again, the data were input to a PLS model. The best results were obtained for the searching combination moving window pre-processing. Here, five of the six analytes could be quantified. However, Cu provided less good data and could only be determined qualitatively.

Another method of data pre-processing was described by Ma et al.20 who employed the method of logarithmic transformation to decrease the noise introduced during the plasma plume evolution process and any randomness arising during acquisition. It also managed to mitigate against self-absorption and matrix effects (both common problems during LIBS analyses). The data were then input to a convolutional neural network algorithm that enabled full spectrum-based multi-task quantitative analysis. Fortunately for the reader, the paper discussed the theory and attributes of the process. Results were good, in that for Cr, Cu, Mn and Mo the R2 values were all 0.984 or above. This led the authors to conclude that their approach was promising for the quantitative analysis of steels.

The near-the-line LIBS analysis of steel slag samples was reported by Peterson et al.21 Near- or at-the-line analyses are very useful in industry because they provide rapid results rather than relying on someone taking a sample to a laboratory and then conducting a traditional analysis. This paper reported a comparison between a traditional univariate and a machine-learning multivariate calibration approach. Calibration samples were first analysed using XRF so that reference values could be obtained. All samples were run twice, each run comprising approximately 600 laser shots while moving the sample over several cm2 in a regular pattern using an XY translation stage. The univariate calibration approach used the ratio method in which the emission intensities of the analytes were normalised against the Ca signal at 317.93 nm. A supervised multivariate machine-learning algorithm called Elastic Net was also used. Elastic Net is a type of algorithm that extends linear regression by adding regularisation penalties to the loss of function during training in order to reduce overfitting. A fuller description of it was provided in the paper. Surprisingly, there was very little difference in the results from the two approaches, with the concentrations of most analytes being very similar to each other and to the reference values obtained using XRF. The one difference that was observed was for MgO, which had a much better agreement between the concentrations obtained using XRF with those from the multivariate approach. This was presumed to be because of the many Mg lines used for the analysis. Since MgO is a major component of steel slag, this was deemed a major advantage.

Another application that used LIBS was presented by Bai et al.22 These authors reported its use to determine W in steels present in the EAST and ITER tokamaks. These workers opted for a one-point calibration approach, a variation of calibration-free LIBS, which was undertaken in a vacuum. The analytical lines to be used in the study were chosen by the algorithm “SelectKBest”. This algorithm removes lines that are affected by self-absorption. The one-point calibration LIBS approach uses a standard reasonably well-matched in matrix to the samples under investigation. Any analytes whose concentrations deviate from the known values are then corrected using “correction factors”, which are also applied to the unknown samples. Using this methodology, the errors observed for the major element Fe was lower than 1.64%. The error was somewhat higher for trace elements such as W, where it reached 24.68%. The values for Fe and W still represent an enormous improvement on the accuracy obtained using calibration-free LIBS where errors were as high-as 10.75% and 202.03% for Fe and W, respectively.

A different approach was adopted by Lu et al.23 who examined the long-term reproducibility of the LIBS measurements of lower alloy steel reference samples. The reproducibility of six analytes (Al, Cr, Mn, Ni, Si and Ti) was measured using the same samples, using the same instrument, by the same operator over a period of several days. The calibrations were prepared using 32 samples and obtained using a commercial single-pulse LIBS instrument. The calibrations were then stored internally and used daily over a period of 10 days to analyse six test samples. Initial results were less than encouraging with relative standard deviations (RSDs) of 53%, 63%, 35%, 64%, 49% and 66% for Al, Cr, Mn, Ni, Si and Ti, respectively. A Kalman-filtering system was then employed in an attempt to improve the precision by removing long-term drift. This was at least partially successful in that the RSD values decreased to 27%, 21%, 11%, 25%, 18% and 37%. Although this is still much poorer than one would hope for, they are a significant improvement.

The classification of steels is important to ensure that stainless steels and specialist steels that contain rare components, e.g., W, can be separated from bulk steels during recycling. An example was presented by Bai et al.,24 who used four grades of stainless steel (201, 304, 316 and 430) as experimental samples. The emission intensities from 12 spectral lines of Cr, Fe, Mn, Mo and Ni were used as the input to the Random Forest algorithm for classification purposes. The accuracy of 100 experiments on 300 groups of data was 98.28% with the time required for modelling being only 0.418 s. The average time required for the classification was even more impressive, being 0.019 s. The good accuracy and stability of the model enabled the authors to conclude that it could be used as a rapid on-line method for classification.

Another paper that utilised LIBS data for classification purposes was presented by Jeon et al.25 These authors compared several different methods of data manipulation, e.g., principal component analysis (PCA), Random Forest and a method by which the spectral intensities measured at each wavelength were normalised by dividing them by the mean value of the spectral intensities in each channel. This was an attempt to reduce the influence of instrumental and experimental fluctuations. Once the data had been manipulated, they were input to either Support Vector Machine (SVM) or Fully Connected Neural Network (FCNN) classification algorithms. The paper discussed each of the manipulation and classification protocols. Six representative classes of steel were used and LIBS data were collected for each at seven different laser intensities. As always, a “training set” of spectra were used prior to “real” samples being tested. When the full data were introduced to the SVM and FCNN, the classification accuracy was initially good for the SVM (98.5%) when the same laser energy was used for both training and for the “real” samples. However, this dropped rapidly to less than 18% success rate when different energies were used. The corresponding FCNN figures were 86.64% to less than 24% success. When the data manipulation tools were used prior to classification, some improvement was achieved. For example, when training used a laser energy of 95 mJ and the analysis of samples occurred using 21 mJ, the PCA-SVM combination saw a classification success rate drop to 17.8%, whereas that of PCA-FCNN decreased to 31.6%. These figures improved significantly when Random Forest was first used for data manipulation; especially for the FCNN classification. However, the best performance was achieved when the normalisation method was used in conjunction with FCNN classification rather than PCA or Random Forest. This was especially true for when the widest range of laser energies was utilised.

Further examples of steel classification were presented by Wang et al.26 and by Shang et al.27 In the former paper, Wang et al. used LIBS to analyse a series of materials (valve stem, welding material and base metal) in an attempt to both classify the materials and to predict the time it has been aged.26 The data obtained were first treated using either PCA or short-term Fourier transform (STFT) prior to input to a probabilistic neural network (PNN). The manipulation using STFT provided better data than the PCA. The overall results were compared with those obtained using a variety of other classification methods, including artificial neural network (ANN), k-nearest neighbour (KNN), partial least-squares discriminant analysis (PLS-DA) and soft independent modelling by class analogy (SIMCA). The accuracy of the classification was 100% for the PNN, with the others ranging from 62.5% for SIMCA up to 95% for the ANN. It was also found that the hardness of the material was directly related to the aging time and that this could be monitored using LIBS because the relative emission energy, line widths, etc. were all affected by the hardness. It was concluded that LIBS offered a rapid method of classifying the different types of metal as well as identifying those components that were aged and possibly prone to failure. The paper by Shang et al. described the use of a hand-held LIBS spectrometer for the analysis of heat-resistant steel.27 Once obtained, the data were pre-processed using a novel approach that combined Lorentzian fitting correction and Kalman filtering prior to being input to SVM. Accuracy of the classification of the aging of steel type T91 reached 94%, which was an improvement over that obtained using standard normal variation and multi-scatter calibration pre-processing methods.

The subject of surface hardness measurement using LIBS was also addressed by Yang et al. who input the analytical data to the machine-learning algorithms of partial least-squares regression (PLSR), convolutional neural network (CNN), deep residual neural network (ResNet) and deep residual shrinkage neural network (DRSN).28 Both linear and non-linear models could fit the relationship between hardness and the LIBS spectra, but the ResNet algorithm fitted the data the best, yielding an R2 value of 0.9967. Interestingly, the authors attributed the good correlations achieved to the inclusion of the numerous sources of noise obtained during acquisition of the LIBS spectra.

Several other LIBS applications have been published. This includes a study by Gupta et al.29 who compared the performance of calibration free LIBS with that of a version with calibration utilizing PLSR, the samples studied were three copper-based alloys and one iron-based alloy with the analytes determined being Al, Cr, Cu, Fe, Mn, Ni, Si, Sn and Zn. Optimisation of the temporal delay was undertaken for the calibration-free version in an attempt to ensure that the requirements of an optically thin plasma that was in local thermal equilibrium were fulfilled. The abstract failed to give any indication of the success of either method. However, the overall premise that the technique of LIBS can be used for rapid analysis with no sample pre-treatment being required, leading to significantly reduced costs, was emphasised.

Another application was the determination of Nd in magnetic alloys.30 Unusually for LIBS analyses, these authors undertook a sample preparation prior to the analysis, namely a fusion. The instrumental parameters of delay time (0.5 μs), pulse energy (20 mJ) number of laser pulses (401) and the laser spot size (65 μm) were all optimised. Three different calibration protocols were compared. These were: external standard calibration (in which different amounts of Nd2O3 was spiked onto alumina), multi-energy calibration and slope ratio calibration. The success of the different methods varied significantly. The external calibration method yielded data that were reasonably accurate, with relative errors of <20% being achieved. The multi-energy calibration was far less successful, with errors exceeding 57% being achieved. This was ascribed to the lack of emission lines with suitable sensitivity being found. The most successful approach was the slope ratio calibration, which had relative errors of less than 3% compared with reference values.

The final LIBS application of relevance to this section was presented by Quackatz et al.,31 who compared the results of the determination of H in arc weld metal and steels that had been obtained using LIBS with those obtained using the standard method ISO 3690. The standard method requires the sample to be heated so that the H diffuses out of it and is then measured ex situ using either thermal desorption spectroscopy or a technique called carrier gas hot extraction. In this study the authors used carrier gas hot extraction as the standard method. A successful analysis using LIBS would allow a rapid, virtually non-destructive method that was capable of spatial resolution to be used. As well as determining H in the bulk materials, the steel pieces were cut using water jets and the H determined in the edges. Certified reference materials were used for quality purposes for the standard method but were also used to construct a calibration curve for the LIBS analyses. A clear correlation between the H concentration determined using carrier gas hot extraction and the LIBS intensity was observed. The LOD was estimated to be approximately 2 mg kg−1. A depth profile analysis was also undertaken after the samples had been charged either electrochemically or by high-pressure hydrogen gas. It was concluded that LIBS held a great deal of promise for time and spatially resolved analysis and that it would be a useful complementary tool to the standard method.

Techniques other than LIBS still reach the scientific literature. This is especially true of studies into corrosion and corrosion mechanisms. A paper by Dorofeeva et al.32 studied the corrosion resistance of AISI 321 stainless steel and the effects that ion implantation had. The coupons of metal (10 × 10 × 1 mm) were first polished to remove roughness and then placed in an ion implanter. The O was first implanted followed by both Al and B being implanted simultaneously. The materials were then placed in a corrosion chamber and exposed to 3.5% sodium chloride solution in acetic acid (pH 3.5) for a period of a month. The corrosion rate nearly halved (from 1.26 μA cm−2 for the non-implanted sample to 0.708 μA cm−2 for the implanted). The samples were then analysed using a number of techniques including SIMS, XPS and TEM to ascertain why this should be. A bilayer was noted in which the O penetrated no more than 25 nm whereas the Al and B penetrated more than 250 nm. In the O rich layer, it was also noted that chromium oxide and nickel chromate had formed. In addition the O had also combined with the Al implants to form a layer rich in alumina. The XPS results showed some clear trends. At the surface, the main Fe species was FeII (73.964% of total Fe) with metallic Fe (10.97%) and FeIII (15.06%) being the other components. Deeper into the sample the FeIII component remained similar, but the metallic Fe had increased significantly to approximately 49%. The Cr data also showed a trend, with metallic Cr increasing with depth. The CrIII at the surface was 91.25% of total Cr with the CrVI being a minor component (6%). These values decreased to 37.82 and 0%, respectively, at a depth of 250 nm.

Another paper to investigate corrosion was presented by Huang et al.,33 who studied the effects of 0.5 M sulfuric acid on the high-entropy alloy AlCoCrFeNi. Analysis of the materials was undertaken using both XPS (for the surface and near-surface region) and hard X-ray photoelectron spectroscopy (HAXPES) for the bulk analysis (i.e., at depth). The amount of metal leached into the sulfuric acid solution was also determined using ICP-MS. Initially, the Co, Fe and Ni dissolved in the acid solution. The Al also showed some dissolution for the first day, but then stopped. The Al and Cr were re-deposited on the surface of the material. The authors speculated that the dissolution of the Co helped decrease the dissolution rate of the Cr. The enriched surface of Al and especially the Cr formed a passivation layer that helped prevent further dissolution of the material. Despite the Cr-based passivation layer, the species with the highest concentration on the surface were still Al compounds. The authors concluded that HAXPES could be a powerful technique for investigating high-entropy alloys.

It has been known for many years that the presence of transition metal ions in solution can be problematic during the determination of elements that form vapours when using tetrahydroborate prior to atomic spectrometric detection. A paper by Buoso et al.34 investigated the acid decomposition products of tetrahydroborate as potential materials to form vapours with Bi (BiH3) and Sb (SbH3). It is known that tetrahydroborate decomposes in acid solution in the following manner:

BH4 → BH3(H2O) → BH2(H2O)OH → BH(H2O)(OH)2 → B(OH)3

It was hypothesised that one or more of these products would suffer fewer interferences in the reduction of Bi and Sb than the original tetrahydroborate. To study this, the authors used HCl at a concentration of between 0.1 and 1 mol L−1 and the test ions CoII, CuII, FeII, FeIII and NiII. The experimental results indicated that the presence of Fe caused minimal problems. However, Co, Cu and Ni led to significant errors when they were present at concentrations as low as 100 mg L−1 when tetrahydroborate itself was used. When the intermediates were used, the concentration that could be present without causing interference increased significantly. The authors proceeded to analyse three certified materials; two iron-based ones (SRM 662 and SRM 663) and one copper-based material (SRM 399). As expected, the results for when tetrahydroborate was used had poor agreement with certified values, whereas the intermediates showed good agreement. The authors stated that further work was required. However, it is encouraging that these interferences can perhaps be circumvented without recourse to the assorted complexing agents that have been used previously.

2.2. Non-ferrous metals and alloys

This has been a very busy area of research during this review period. Numerous sample types have been analysed and many different techniques have been employed for the analysis. As with the analysis of ferrous metals, the most common technique used has been LIBS. However, the focus has been slightly different. There have been numerous papers using statistical methods to pre-treat the data or to improve accuracy of the calibration, but many others have focussed on modifications to the technique to try and improve sensitivity. Other techniques used to give interesting applications have included XRF, SIMS, GD-MS and LA-ICP-MS.
2.2.1. Reviews and overviews. A review by Shen et al.35 containing 197 references was entitled “Secondary ion mass spectral imaging of metals and alloys”. After an introduction, the review discussed, with the aid of diagrams, the main differences between TOF-SIMS and magnetic SIMS instrumentation. Following that there was a historical perspective, a section on the principles of SIMS, a description of the different types of spectrometer and their associated attributes (resolution, mass range, sensitivity and whether or not they are sequential or simultaneous) and a summary of the capabilities of commercial instruments that are currently available. A brief discussion of operando or in situ SIMS and a section describing the three modalities of SIMS, namely, depth-profiling, imaging and spectral analysis, were then given. The main bulk of the review concentrates on applications. This large section is split into numerous sub-sections that concentrate on corrosion behaviour, thin films and oxide layers, metals and alloys for biomedical applications, semiconductor materials, geological formations and minerals, characterisation of particles, nano-SIMS, complementary techniques and nanomaterials. Since huge amounts of data can be obtained using SIMS, there is a need for efficient data handling. A section describing multivariate analysis and machine learning was therefore very welcome. The review concluded with an outlook for the future. The review is a very useful read for anybody new to the area.
2.2.2. Copper and copper-based alloys. As discussed briefly before, LIBS has been the most popular technique employed for the analysis of non-ferrous metals and this is true for the copper-based materials. Their analysis has concentrated mainly on modifications to LIBS rather than the use of statistical tools to classify the materials. This was exemplified by a paper presented by Matsumoto et al.36 who reported the use of underwater LIBS. This technique is known to be potentially very useful because of its standoff ability, i.e., there is no need to go to the expense of raising something from the seabed to do an analysis. However, it is also known to suffer several important problems, not least that of poor sensitivity arising from a combination of plasma quenching by the water and from light loss through the fibre-optic signal collection device. In this study, a copper plate was immersed in sodium chloride solution and then the signal fluctuation monitored as the LIBS laser repeatedly irradiated the same spot. It was noted that the intensity of both the Cu and Na emission increased with increasing number of laser shots. It was also noted that there was a strong correlation between the intensity of the emitted light and the bubble collapse time. By normalizing the signal intensity to the collapse time (having removed the offset, i.e., the time taken for the bubble to appear) the precision of the repeated analysis improved from 41.4% RSD to 27.0% RSD for the Cu and from 28.8% to 17.7% for the Na. Other ways of improving the precision were also confirmed. These included the exclusion of the data from the first firing on a new copper plate and normalizing against the signal obtained from the dissolved species. The authors concluded that their findings could facilitate underwater analyses using LIBS.

Metal classification and recycling is an important industry, since there is a limited supply of materials in the world. Significant work has been undertaken in the last few years to enable rapid at-site classification of metals at scrap yards. Another example of this was presented by Bernat et al.37 These authors used three mobile instruments, namely LIBS, XRF and spark-OES, to determine the effects of surface roughness and of any coatings present on scrap metal on the analytical accuracy of the analyses. Since scrap can come in assorted shapes, can exhibit surface roughness and can be coated with varnish or paint, these factors could potentially affect the accuracy of the data obtained to the extent that the material could be mis-classified. Materials with surface roughness of between 0.03 and 6.7 μm and coated with various types of varnish (e.g., alkyd, water-based, oil-phthalic, acrylic and oil-alkyd) were tested. Results were very interesting. A sample with a surface roughness of <2 μm had no adverse effect for the spark-OES and LIBS instruments but caused significant problems for XRF. Samples with a surface roughness of greater than 2 μm had an effect on all three techniques. The effects of the varnish or paint coatings were even starker than those from surface roughness. A single paint coating could change the concentrations of the alloying elements measured by more than 10%. In the case of spark-OES, it could prevent the spark from forming properly and hence no measurement could be made.

There are several methods available that can improve the sensitivity of LIBS analyses. Included in this number is nanoparticle-enhanced LIBS (NELIBS). This has been used for several years and can lead to improvements in sensitivity of up to an order of magnitude because of improved energy coupling with the laser. A study by Soumyashree and Kumar38 examined the effects of pressure and the pulse energy on the plasma plume and signal produced during the NELIBS process. Two sample types were tested – copper and aluminium – and the experiments were conducted using a Nd:YAG laser operating at its primary wavelength (1064 nm). Detection was achieved using an Echelle spectrometer equipped with an intensified CCD. Low-pressure measurements were achieved by placing the sample in an evacuated chamber. The sample was first cleaned by firing a total of 20 times before analytical measurements were made. Silver nanoparticles (1 μL) were then injected onto the clean spot, dried using air and the chamber re-evacuated prior to the NELIBS procedure. Data handling was achieved using a Matlab algorithm. Interestingly, although the NELIBS yielded improved sensitivity compared with LIBS, examination of the plumes produced indicated that the diameter, aspect ratio and expansion velocity were similar. This was true for both metal samples analysed. However, signal enhancements in the case of NELIBS showed a pressure dependence with larger enhancements seen in atmospheric pressures at longer delays.

A new method of LIBS sensitivity improvement was proposed by Zhang et al.39 These authors coated the surface of the sample (aluminium, chromium, copper, iron or tungsten) with a layer of epoxy resin of thickness 0.2 mm and then let it cure. The laser was then fired once to make a small circular hole in the resin layer. Subsequent measurements were made through this small hole. The operating conditions in terms of resin to hardener ratio, thickness of the layer and the laser energy of the pre-ablation process, were all optimised. Increases in sensitivity ranged from 66.7% (for Al at 394.46 nm) to 699.6% (for Cu at 510.6 nm). Most enhancements were in the range 96–130%. As well as the improved sensitivity, improved precision was also obtained, with the range decreasing from 6.5–25.5% RSD to 0.2–12.9% RSD. Although these improvements are useful and the process is inexpensive, the authors admitted that the process is overly long. Future work would look at improving sensitivity further using double-pulse LIBS, microwave enhancement etc. as well as shortening the procedure.

A paper by Chen et al.40 reported target-enhanced double-pulse LIBS analysis of bismuth brass samples. The authors emphasised that the analysis is beset with problems when calibration free LIBS is employed because of many lines not having known transition probabilities and because of self-absorption. In an attempt to overcome these problems, the authors employed a one-point calibration strategy. This is usually where one sample of known composition is used and then any deviances from the expected values are corrected using factors that can also be applied to the unknown samples. On this occasion, the methodology appeared to work with relative errors of less than 4% being achieved.

Depth-profile measurements of tungsten/copper functionally graded materials obtained using LIBS was reported by Ivkovic et al.41 The materials used for the calibration were available commercially and entitled W93Cu7, W90Cu10, W80Cu20, W70Cu30 and pure W and were named after their approximate composition. The exact composition was determined using XRF. Each coupon was 10 mm in diameter and 1 mm thick. The functionally graded material was a disk of 6 mm diameter and 1 mm thickness with the Cu concentration gradient in the thickness dimension. The LIBS measurements were made using a Nd:YAG laser operating at 532 nm and at a power of 100 mJ per pulse. The craters formed by the laser were examined using optical profilometry so that amount of sample ablated could be estimated. The Cu concentrations at each depth was determined using a univariate calibration and then a ratio of the Cu signal at 521.82 nm made against the W signal at 522.47 nm. The procedure was sufficiently successful for the authors to conclude that it had “potential applicability for quantitative analysis of multi-layered materials” with the analysis of the tiles or cooling channels of nuclear reactors in mind.

Other techniques have been used to analyse copper-based materials. Lv et al.42 determined Cu isotope ratios in bronze using fs LA-MC-ICP-MS using both matrix-matched and non-matrix-matched calibration. Copper isotope stock solutions NWU-Cu-B and NWU-Cu-A were used as reference materials and monitor standards and the reference materials GBW02137, GBW02138, GBW02139, and GBW02140, which have a diverse Cu content and isotopic composition, were used as samples. Since they are provided in powder form, they were fused into disks and polished prior to use. Aliquots of the powders were dissolved so that the isotopic composition could also be determined using conventional nebulisation of solutions so that a comparison with the LA data could be made. When GBW02138 was used as a standard, accurate delta 65Cu data were obtained, but this was not the case when a pure copper wire was used; indicating that matrix effects existed. The authors analysed GBW02138 using LA-ICP-MS over a period of eight months and obtained very consistent results leading the authors to conclude that the material is very homogeneous and therefore ideal for acting as a matrix-matched reference material for Cu isotope analysis of bronzes.

2.2.3. Aluminium and aluminium-based alloys. The analytical technique of LIBS, and modifications thereof, are again dominant in the literature for aluminium-based materials. Although many papers are still attempting to address and circumvent problems associated with accuracy, calibration and interferences, others have used it for the classification of scrap, determined the effects of the sample surface on accuracy and have used LIBS to analyse molten material.

Numerous studies have focussed on interference removal or reduction of background in an attempt to make LIBS a more robust technique. Zhang et al.43 reported the development and testing of a method to eliminate the effects that the matrix has on the accuracy of the analysis. The method was plasma parameter correction based on plasma image–spectrum fusion. This meant that the plasma temperature and electron number density of the plasma, which can change depending on what sample is present, can be “corrected” by using features in plasma images and spectra. Details of the exact methodology and the theory were presented at length in the paper. The approach was tested on pressed samples of soil as well as the metals cast iron and the aluminium alloy GSB 04-1661-2004. A comparison was made between the data treated using the proposed method and those obtained using image-assisted LIBS, a protocol devised by the same research group previously. The results for the proposed method were a significant improvement over the image-assisted LIBS, with R2 values all exceeding 0.997 and the average root mean square error (RMSE) being improved by 29.63% and the average relative error decreasing by 38.74%. It was hoped that this further development may help in removing matrix effects enabling LIBS to become more widespread in industry. A similar paper was also presented by Nie et al.44

Self-reversal is another problem associated with LIBS analyses, especially when the analyte is at very high concentration. Hedwig et al.45 developed a method for the determination of Al in titanium–aluminium alloys where the Al can compose between 5 and 50% of the mass. The method involved creating an air spark on the sample simultaneously with the plasma plume. In a similar manner to double-pulse LIBS, the single laser pulse produces the plasma, which is then further excited using the spark. The spark had to be placed carefully though. The tail of the air spark had to be placed between the sample surface and the outer part of the target plasma. The method was reported to be cheap, but successful in removing the self-reversal for the Al 394.40 nm and Al 396.15 nm lines completely. Consequently, calibrations were linear over the concentration range 5–50%.

Background correction is another area of LIBS analyses that still requires research. Noise can occur through fluctuations in plasma intensity, laser sample interaction, laser sample plume interaction and just random noise. Chen et al.46 developed an automatic background correction system based on the idea of “window functions and differentiation combined with a piecewise cubic Hermite interpolating polynomial”. The paper gave a detailed mathematical description of the process. In simulations it out-performed asymmetric least-squares and model-free background correction, yielding the highest signal-to-background ratios. The method was tested on seven different aluminium-based alloys yielding a significant enhancement in the linear correlation coefficient between spectral intensity and the concentration of the analyte Mg. In addition, the regression was also slightly improved and the error was reduced compared with the other models.

The surface of the sample can also have a large effect on the accuracy of a LIBS analysis. If the surface is uneven, e.g., if it is a piece of scrap metal that may be twisted or bent, then serious errors can arise, especially if the source laser and detection system are fixed into position. This could be because the laser becomes de-focussed from the sample surface or the collection optics are further away leading to less efficient collection, or a mixture of both. Chen et al.47 devised a data transfer method to reduce the spectral fluctuations. Experimentally, they focussed a laser on any of 14 aluminium alloys and then moved it away in 0.5 mm increments up to a distance of 2.5 mm away. As usual, a training set of spectra (200 for all 14 alloys giving a total of 2800) was required to set the model (k-nearest neighbour) up. Each distance from the start point was then analysed 50 times for each alloy giving a total of 700 laser firings per distance. The accuracy of the method was extremely poor, being in the range 11.48% to 88.28%, with a large majority being 53% and lower. Feature selection of the raw data undertaken using an algorithm called joint distribution adaptation was then used to select only the most useful features to be input to KNN. The corresponding accuracy data when both the feature selection and KNN were used improved significantly to 73.71% up to 98.42%, with the majority being over 80%.

Lanzinger et al.48 used LIBS to analyse alloy particles as part of a cleanliness study. A total of 10 analytes in 51 CRMs were used as a training set. A further eight CRMs were used for validation. Analytical data were baseline corrected using the Matlab® function msbackadj and normalisation to the signal intensity of the aluminium line at 266.04 nm was performed on each spectrum prior to being input to PLS for calibration and concentration calculation. The LOQ was determined to be approximately 0.05% for nearly all analytes, the exception being Cu where it was 0.1%. The LIBS data were used in conjunction with microstructure analysis and the combination of the two techniques enabled the successful analysis of particles as small as 45 μm to be transferred to bulk material.

The classification of alloys normally requires a dataset for the training of a model, possibly with the addition of some sort of data reduction to remove redundant areas of the spectrum, remove interferences, etc. The model then attempts to predict which type of alloy the sample is. There have been several papers that have undertaken classification studies, two of which were for the classification of scrap metal. One example, by Qu et al.,49 used a 100 μJ, high repetition frequency (10 kHz) laser for the analysis, which enabled the size and the cost of the instrumentation to be minimised. Two modes of operation were tested: fixed mode, where the laser focusses on the same spot, and motion mode, where the sample is moved on a translation plate so that the laser hits different parts of the sample. The operating conditions for each were optimised. An integration time of 80 ms and an integration window of 80–160 ms was optimal for the fixed mode and an integration time of 50 ms and a sample movement rate of 7 mm s−1 was optimal for the motion mode. A total of 14 types of aluminium alloy were tested with seven from one manufacturer being used as the training set and the other seven samples from a different manufacturer being used as the test set. The LIBS data were input to a back propagation artificial neural network (BP-ANN) for classification. The best data obtained enabled an accuracy of classification of 97.71%. The motion mode provided better data because in the fixed mode, the laser continually ablates the same spot; hence, as the crater gets deeper, the focus point of the laser is no longer at the sample surface.

The combination of LIBS and RGB and 3D cameras, followed by statistical analysis of the analytical data, has been used for the classification of aluminium-based alloys by Diaz-Romero et al.50 Two models were tested: one was a single output model and the second was a multi-output model that uses the structure of the single model to enhance learning and avoid over-fitting. Over-fitting is a common error with machine-learning algorithms where too much information is discarded hence making the model more restrictive and less efficient. A description of the two models with useful schematics was presented in the paper. The models were used for two tasks: differentiating between cast and wrought aluminium and also to classify the scrap alloys into three commercially interesting fractions. A total of 773 scrap pieces were used to train the models for the tasks. The single model was better at distinguishing between cast and wrought aluminium, with a precision, recall and F1 score all of 99%. The multi-output model was better at classifying the three types of alloy with corresponding values for precision, recall and F1 score of 86%, 83% and 84%, respectively. It was concluded that the fusion of LIBS and computer vision image methods is a successful way of classifying materials that could be used rapidly at scrap yards.

Another example of using LIBS to obtain analytical data (Mg in 17 aluminium alloys) was presented by Ding et al.51 These authors input the data obtained using LIBS into two different algorithms, PLS and Random Forest, and compared their prediction ability. The PLS had poor performance, yielding a correlation coefficient of 0.6809 and a root mean square error of 1.2042. The Random Forest performed somewhat better, yielding values of 0.8571 and 1.0918 for the same parameters. Further work was then performed that pre-screened the data being input to the Random Forest algorithm. When the data were pre-screened to reduce it to only variables with an importance of greater than 0.11 being input, the Random Forest prediction performance was improved further to a correlation coefficient of 0.9461 and an RMSE of 0.9534. Another big advantage of using a smaller data set as input to the algorithm is that it reduced the computational time required by 91.67%.

Ma et al.52 also devised experimentation to reduce computation time and loading during the statistical analysis of LIBS data, they described a small-scale stacking model for the qualitative analysis of aluminium alloys. The model used three layers (stacks) of treatment of the analytical data. The first stack involved the Random Forest spectral feature selection and specific spectral line spreading to reconstruct the data. The second layer utilised three heterogeneous classifiers to extract features from the reconstructed spectra in different feature spaces, generating second-level reconstructed data. The third and final layer utilised the reconstructed dataset for qualitative prediction. The paper discusses the processes involved in each of these layers in detail enabling the reader to understand the overall methodology. The results were a significant improvement over established methods of classification, out-performing KNN, SVM and Random Forest including those that also utilised PCA as data pre-screening. Using only 15 spectra for training the classification accuracy was 96.47% compared with the improved Random Forest prediction accuracy of 71.76%. Given that most classification experiments rely on many more spectra for training purposes, this was impressive. When more spectra were used for training, the stacking method achieved a 100% success rate. Although the Random Forest prediction rate also increased, the stacking method still out-performed it by >6%. The main advantage of the process was the much-reduced time and effort required in obtaining data to train the model.

There have been several other LIBS-based applications reported that describe modifications to the LIBS process that improves the sensitivity or background to signal ratio. Wubetu et al.53 compared time and polarisation-resolved LIBS with “ordinary” LIBS for the analysis of aluminium alloys. The signal-to-background ratio for the Mn line at 415 nm was significantly increased when the polarisation-resolved LIBS was used compared with just LIBS. The authors investigated the mechanism of this enhancement and also the extent of the polarisation. The latter was achieved by monitoring two matrix lines (Al atom line at 396.15 nm and the double ion line at 569.6 nm) and two unpolarised neighbouring lines. It was concluded that polarisation-resolved LIBS was a useful and simple addition that was capable of improving the performance of LIBS analyses.

Another publication that reported attempts to improve sensitivity of LIBS analysis was presented by Khan et al.54 These authors developed a special sample holder that could be placed in a strong magnetic field. The methodology was entitled magnetic field confined laser-induced plasma. The transverse magnetic field applied a strength of between 1 and 3.5 kG to a laser-induced plume from an aluminium alloy. The power applied to the laser was varied between 80 and 160 mJ. Gate times and other parameters were also optimised. Under optimal conditions, the signal enhancement could be as high as a factor of seven compared with standard LIBS under optimal conditions. However, this was very analyte-dependent with some analytes having only a marginal enhancement. Detection limits were also improved, with that for Fe improving from 135 to 34 mg kg−1, Mg improving from 160 to 124 mg kg−1, Si from 132 to 52 mg kg−1 and Zn from 86 to 26 mg kg−1. Precision was also improved for some analytes, with that for Mg improving from 23% to 12% and that for Si improving from 25% to 14%. The precisions of the measurements of electron temperature and electron number density were also improved.

The contribution of the vapour phase to the LIBS signal of Mg in liquid aluminium was studied by Thorarinsdottir et al.55 Magnesium is an important alloying element for aluminium alloys, but having a lower boiling point may mean that it produces a vapour pressure above the liquid aluminium. By conducting LIBS experiments in the vapour above the molten aluminium, it was possible to demonstrate an exponential dependence of light emission on temperature. The authors then conducted LIBS experiments on a solid sample of the same material, i.e. it had the same Mg concentration. These experiments proved that a significant part of the Mg emission from molten aluminium originates from the vapour phase. Using different samples containing Mg concentrations spanning four orders of magnitude, it was possible to determine that the self-absorption of the Mg signal was significantly higher for the molten samples than the corresponding solid ones. The amount of self-absorption in the molten samples exceeded that expected from theoretical models. The melt temperature, gas flow, measurement geometry, laser pulse energy, and repetition rate, as well as the activity of the given species in solution, make the interpretation of LIBS results highly challenging. The authors concluded that a good understanding of these parameters was required for accurate results to be obtained.

A modification of LIBS is spark-induced breakdown spectrometry (SIBS), where instead of a laser to ablate and excite a sample, a spark is used. Pinjun et al.56 fused the two techniques together so that the LIBS laser first formed a plume of sample and then the spark further excited that plume so that the emitted light could be detected. They then compared the performance of this hybrid technique (they termed it SIBS) with that of LIBS during the analysis of aluminium alloys for their Cu and Zn content. The double excitation process is analogous to double-pulse LIBS where a second pulse, either from the same laser or from a different one, is used to further excite the material. This often leads to fewer interferences and improved sensitivity. The paper described the instrumental setup required for the LIBS and SIBS to be performed. It was noted that the SIBS led to a significant improvement in LOD. For instance, under the same laser power, the LOD for LIBS alone was 96 and 83 mg kg−1 for Cu and Zn. These improved to 16 and 12 mg kg−1 during the SIBS analysis, representing an improvement by a factor of six or seven.

As with the analysis of ferrous metals and alloys, LIBS is not the only analytical technique to have been used. An interesting study into the corrosion of aluminium alloys was undertaken by Mukhametzianova et al.57 who used diffusive gradients in thin (DGT) films followed by LA-ICP-MS. Several gel types including polyacrylamide–Chelex–Metsorb, polyurethane–Chelex–Metsorb and polyurethane–Chelex–Zr(OH)4, were tested during the study. The samples were a commercial grade aluminium of thickness 1.95 mm and a high-strength copper–aluminium alloy of 0.7 mm thickness. Samples were soaked in a sodium chloride solution (1.5%, pH 4.5, T = 21 °C). The polyurethane–Chelex Zr(OH)4 material showed the best performance, with the analytes Al, Cu and Zn all being retained and it possessing a higher Al capacity than the other gels. In addition, it was more robust, was less likely to tear and had less shrinkage. Immersion of the samples in the salt solution followed by allowing the solutes to migrate onto the gel was then followed by analysis of the gel using LA-ICP-MS. It enabled the authors to identify that Al and Zn co-solubilise and that corrosion starts in discrete pockets rather than uniformly across the sample surface, An extremely sensitive technique resulted, with detection limits being 0.72 pg, 8.38 pg and 0.12 pg for Al, Zn and Cu, respectively from an area of gel of 0.01 mm2. The study advanced the assessment of aluminium alloy degradation in aqueous environments and hence supported the design of corrosion-resistant materials.

It can be difficult to obtain an accurate result from the analysis of an Al–Zn–Mg–Cu alloy rod when it has been prepared using a device that changes the gradient of each component throughout the synthesis. The composition is therefore not homogeneous and requires techniques that can analyse extremely small areas for an accurate assessment to be made rather than relying on a bulk analysis technique where clear errors will result. Yang et al.58 used μ-XRF with a spot size of about 20 μm in an attempt to circumvent this problem. After optimisation of the operating conditions (testing voltages, testing currents, dwell time for each pixel, etc.) the authors analysed the material. Results were validated by comparison with data obtained using spark-OES and were in good agreement. Limits of detection for Cu, Mg and Zn were 0.002%, 0.068% and 0.007%, respectively.

2.2.4. Other alloys and metals. There are several other metals that have had some interesting analyses performed on them. These have therefore been brought together into one section. The determination of H in a Ti–6Al–4V titanium-based alloy is problematic because if the surface of the material is damaged using heat, e.g., by a laser or spark during analysis, the H re-distributes itself. Using GD-OES, Weiss et al.59 determined H and a suite of other analytes (C, N and O as well as the matrix materials Al, Ti and V) in an attempt to develop a correction factor. Unfortunately, the presence of H in the sample also affects the accuracy of determination of these analytes and so pure metallic samples of Al, V and Ti were also analysed to help compensate for this. No correction was required for the first 5 μm depth of the Ti–6Al–4V material because of the presence of an oxide layer, in which there was no H. At greater depth, correction was required though. Using a series of experiments in which the surface is ground away and the new surface analysed using GD-OES, it was noted that each new surface was rich in H and a layer below that was H deficient. This indicated that the H diffused from the adjacent region into the area under analysis artificially elevating the real H concentration. The authors attempted a discussion explaining this phenomenon and presented a model to compensate for it.

Two papers have used LIBS to analyse nickel-based alloys. In one, presented by Zhao et al.60 the problems associated with on-line quantitative analysis for laser processing using LIBS was addressed. These problems include fluctuations in the plasma, uncertainties associated with laser ablation and Bremsstrahlung, all of which can prevent accurate results being obtained. The authors therefore developed a system that would overcome these limitations. Included in their setup was a specially designed optical Fourier filtering structure to overcome the high-frequency ablation noise component of the plasma and a spectral screening system based on plasma time waveform monitoring to suppress the influence of plasma fluctuations. Seven certified nickel materials were used as calibrants and the data processing required several stages. These were: background correction; spectrum peak searching; spectrum fitting; certified compositions of all the standard samples being obtained; calibration curve construction; selection of the inner relative standard and calibration wavelengths; and finally, linear fitting. Each of these processes was discussed in the paper. Results for nine analytes showed a significant improvement in terms of accuracy and precision.

The other paper to report the analysis of nickel-based materials was presented by Choi et al.61 who used LIBS analysis followed by data processing to classify nickel-based alloys. These authors constructed a low-cost, low-resolution instrument that was capable of determining the major analytes in the samples. A comparison of data obtained was made between those where just the raw intensities were used and those that had used a ratio method. Unexpectedly, the method using the raw intensities provided the better results in terms of calibration accuracy; but only for the major elements Cr, Ni and Fe. Classification of the alloys was then attempted with KNN and LDA being used. These had classification correctness values of 95.0 and 98.3%, respectively. The LDA model was then modified so that the two closest sample classes that were responsible for the incorrect classification were modelled separately. This two-step LDA model yielded 100% correct classification. It was concluded that by just using the major components for the LIBS analysis, even a low-cost and low-resolution instrument can be used effectively if the correct chemometric model is adopted.

Zirconium metal and/or zirconium oxide was analysed in two papers by Ikeda et al.62,63 who employed microwave-enhanced LIBS for the task. In the first of these papers, the microwave enhancement was hoped to compensate for the signal loss that arises through the use of fibre-optic cables that transmit the light from the source to the detector. Increasing the microwave power had a large effect on the signal-to-noise ratio of both Zr atom and ion lines produced from zirconium metal with an improvement of several hundred-fold being obtained compared with standard LIBS. The microwave enhancement of Zr ion lines was even greater than for atom lines, although the signal-to-noise ratios were similar. The conclusion was that microwave enhancement enhances both the excitation and ionisation of the sample. The second paper63 analysed both zirconium metal and the oxides, which are both used in the nuclear industry. Enhancements were also found in the zirconium oxide material. Although signal enhancements were observed for both ion and atom lines, the microwave enhancement did not affect the level of continuum background emission.

Three papers have addressed the analysis of gold or gold-based alloys. A paper by Li et al.64 reported the target-enhanced orthogonal triple-pulse LIBS analysis of gold alloys. Several methods of signal enhancement have been developed over the years, but this was a new development. A pre-ablation laser collinearly propagated with the reheating laser was introduced in an attempt to further improve the signal enhancement factor obtained using the double-pulse equivalent. The experimental setup of the three Nd:YAG lasers was shown in the paper. The first laser was weakly defocussed on the surface of the18 carat gold sample and acted as the ablation laser. Laser 2 (pre-heating laser) and laser 3 (reheating laser) were collinearly aligned and focused on the target surface. The target used to enhance the signal was a potassium hydrogen carbonate pellet. The inter-pulse delays of the different lasers were optimised to ensure maximal signal. Under the optimised conditions, the signal-to-background ratios improved significantly when compared with single-pulse LIBS for all analytes but was both analyte and wavelength dependent. The enhancement factor for the Au 267.595 nm line was 11.3, Cu 324.754 nm had an enhancement factor of 8.4 but the Cu 327.396 nm yielded only a 6.8-fold enhancement. Two Ag lines also showed enhancement factors of 5.4 and 8.7. Interestingly, the Zn signal was not observed using single-pulse LIBS, but was observed using the setup proposed. Investigation of the craters produced in the sample indicated that the target-enhanced triple-pulse LIBS produced much smaller craters in both diameter and depth.

The sensitivity of the LIBS analysis of gold alloys was also enhanced in a study by Ahmed et al.65 who employed electric field assistance to improve the LOD of Ag, Au and Cu. The electric field was applied through the use of two aluminium electrodes placed 3 mm apart and at right angles to the sample target. A full description of the setup and operating conditions were presented in the paper. A voltage of 360 V was applied across the electrodes leading to signal enhancements and improvements in LOD. For instance, the LOD for Ag improved from 4.3 mg kg−1 to 2.06 mg kg−1 when the electric field was applied. The LOD for Cu was improved even more with a value of 14.3 mg kg−1 being improved to 1.7 mg kg−1. The relatively low-cost modification was thought to be an effective method of improving detection limits when compared with other more expensive techniques.

The third paper to analyse gold was presented by McCoy-West et al.,66 who determined Ag isotopes. Instead of employing several ion-exchange steps to isolate Ag the authors developed a method using a single anion-exchange column followed by conversion of the silver from chloride form to nitrate using ascorbic acid and ammonium hydroxide. The process offered several advantages over other methods such as speed (therefore enabling more rapid sample turnover) and less chance of analyte loss. The method was applied to silver metal and gold pieces. The Ag isotope composition was compared with NIST SRM 978a silver standard. The accuracy of the method was assessed in two ways. The natural gold standard CEZAg was used to test the external reproducibility of the chemical separation and conversion method yielding a result that was within the acceptable uncertainty of previous determinations. In addition, the analysis of gold nuggets provided data that were comparable to those obtained by a different laboratory using a more traditional multi-column methodology.

Matsuda et al.67 developed a statistical method that could convert XRF data into a 2D colour image of the surface of a magnesium alloy. This made it possible to visualise any significant differences in data. The normal interquartile range was used as the basis for segmenting the colour range scale of the image while setting the median concentration as the central value of the scale. The procedure was applied to the determination of Al, Mn, Si and Zn in two magnesium alloys. Using the method, it was possible to identify visually any inhomogeneities present. It was concluded that the methodology could just as easily be applied to the visualisation of SEM-EDS data and to research fields as diverse as semiconductor film deposition and plating uniformity.

The final paper of relevance to this section was a study by Arkhipenko et al.,68 who used WDXRF and a fundamental parameter method to analyse waste samarium–cobalt magnets. A two-stage process was developed that enabled the determination of both the main components (Sm, Co) and impurities (Cr, Cu, Fe, Hf, Mn, Ni, Ti and Zr) present in the samples. As well as the fundamental parameter method the authors also prepared calibration curves, which could be used to quantify the analytes. The results of the fundamental parameters were, at best, mixed. Excellent agreement with certified values were obtained for Co but most other analytes were not close to expected values. For example, in one sample the experimental values for Fe and Sm were 18.8 ± 1.1 and 26.2 ± 1.6%, respectively, whereas the certified values were 12.46 ± 0.18 and 34.21 ± 0.32. Some of the analytes at lower concentration were at least an order of magnitude wrong, e.g., Cr, where the experimental value was 0.04 ± 0.008% whereas the certified value was 0.0014 ± 0.00008%. Therefore, the fundamental parameters method could be regarded as being no better than semi-quantitative. The results obtained from the calibration curves were far superior and tended to be in good agreement with both certified values and the values obtained using ICP-OES.

3. Organic chemicals and materials

The analysis of organic chemicals and materials has continued to be a popular area of research. The most popular part has been the analysis of pharmaceutical materials, although the analysis of cosmetics, paints and explosives has also drawn attention. Interestingly, unlike many of the sections in this update, LIBS has not been the method of choice for pharmaceutical materials. Although there have been a few applications, the majority have used more traditional techniques such as ICP-OES/-MS and XRF. The analysis of paints and explosives, however, has continued to favour LIBS.

3.1. Organic chemicals, paints and explosives

A review containing 215 references entitled “Taking a deeper look into the roles of amines in atomic absorption spectrometry” was presented by Aller and Pereira.69 The review was wide-ranging and covered the obvious topics of metal complexation/extraction prior to analysis using both FAAS and ETAAS. In addition, it also covered topics such as how the presence of amines affects the atomisation processes. Other sections included the actual determination of amine- or N-containing compounds using AAS. Examples included the determination of N in fertilizers and nitrate/nitrite in groundwaters.

An application by Santos et al.70 described the LIBS analysis of organic liquids. The use of LIBS in any liquid is challenging because of quenching effects of the plasma formed and because of potential splashing of the sample. The authors circumvented these problems by using an inorganic clay-like material that was a mixture of 70[thin space (1/6-em)]:[thin space (1/6-em)]30 bentonite[thin space (1/6-em)]:[thin space (1/6-em)]sepiolite as a support. The substrate enabled a molecular LIBS analysis of the solvents (e.g., toluene), fuels such as diesel and gasoline and even aqueous-based samples to be obtained. Minimal spectral interference was observed enabling species such as –CN and C2 to be identified. It was concluded that the substrate material was extremely suitable for obtaining molecular LIBS spectra from organic liquids.

Work into using LIBS to monitor paint removal from aircraft tails or skins has continued. Two papers by Li et al.71,72 discussed the real-time monitoring of paint removal from the aircraft parts. In one example,71 the authors explained how removal of only the top layer of paint is required leaving the yellow primer underneath unaffected and certainly not damaging the anti-static black primer that protects the aircraft from lightning. The experimental details were discussed at length. The output parameters of the fixed laser were a pulse width of 350 ns, scanning speed of 2800 mm s−1, line spacing of 0.04 mm, one laser scan, and a repetition frequency of 90 kHz. The average laser power was adjusted to obtain specimens with different paint removal depths. Although the top coat paint and the yellow primer both contained similar metals, the relative concentrations were very different. For instance, in the top coat, approximately 98% of the metal content was Ti, whereas in the primer the content of Ba, Sr, Ti and Cr accounted for 36.62%, 27.73%, 19.57% and 12.75% of the detected metal elements. The difference between the LIBS spectra of cleaning depths 44.79 μm and 51.50 μm were compared and analysed. The characteristic peak of Sr I at 460.66 nm in the LIBS spectrum could characterise the paint removal depth and delamination boundary. The second paper72 used similar technology but then used the chemometric tools of PCA, PLS-DA and orthogonal PLS-DA to obtain classification and identification models. The PCA was used mainly as a means of decreasing the data for the model-forming algorithms to work on. It removed signals from redundant areas of the spectrum and reduced noise, and in doing so, accelerated computing time for the modelling algorithms. The PLS-DA and orthogonal PLS-DA were used both with and without PCA data removal. The orthogonal PLS-DA model proved to be superior when information from 12 spectral lines were input. The predictive accuracy of 94.4% was achieved enabling the swift classification, recognition, and prediction of LIBS spectra data from the topcoat, primer, and aluminium alloy substrate. These two papers are another example of how LIBS can be used on-line and in real time to facilitate industrial processes.

A third application that discussed the analysis of paints was presented by Turner and Filella73 who used a portable XRF instrument to determine Cr and Pb in European road paints. A total of 236 paint samples from 11 different countries were collected for analysis. Most of the paints were white or varying shades of yellow, but examples of red, blue and green were also obtained. The most common metals present were identified as being Ba, Ca, Fe and Ti. However, the concentrations differed markedly with location. As an example, the median Ba concentration in Spain was 45[thin space (1/6-em)]400 mg kg−1 compared with only 1370 mg kg−1 found in the UK. The main focus of the research was on Cr and Pb. Lead was found in 148 of the samples with the highest recording being 17.2% by weight. Although it was found in all of the different coloured paints, the large majority with a Pb concentration above 1000 mg kg−1 were yellow. Chromium was detected in fewer samples (81), with the highest recorded concentration of 1.94% by weight. It was concluded that lead chromate was the most common but not the only paint pigment used and that Pb was often combined with other species. A depth study was undertaken and found that both Cr and Pb had been used in paints recently, historically and had been overpainted. The overall conclusion was that these two elements are still in common use and therefore still represent an environmental threat.

The analysis of explosives has been a popular area of research with three papers being published, all from the same research group. In one example, Zhang et al.74 used both LIPS and high-speed Schlieren imaging to determine propellant products. It is difficult to determine the detonation products because of the fast reaction times. However, knowledge of them helps to evaluate the performance of the explosive and formulate improvements. The system developed enabled the time-resolved and spatial distribution of the materials produced. It was deduced that N2, O2 and nitrogen oxides preceded the formation of hydrocarbon gases. This was followed by carbon oxides and hydrogen. A second paper by this group used LIPS to discriminate between and classify different explosives.75 Five high explosives and five organic materials containing benzene rings were analysed using LIPS. Each material had six regions of the spectrum analysed and then k-means clustering analysis in two-dimensional principal component space was used to identify parts of the spectrum that would of most use in discriminating between the samples. The data were then input to SVM and/or a supervised quadratic discriminant classifier. Both of the classifying algorithms had great specificity and an accuracy of above 99%. The third paper from this group was similar to the second but focussed on identifying the dominant intrinsic factors of sensitivity of high explosives, i.e., what happens if they receive an impact, receive friction, etc.76 This is achieved by determining species such as –CN and C2, which are known to be intimately linked with sensitivity. Other factors not measured using LIPS were also taken into account for the classification. These included the oxygen balance and the multi-core oil–water partition coefficient. A total of nine explosives and 16 non-explosive organic molecules were analysed and the data obtained input to a machine-learning algorithm (Random Forest). It was possible to distinguish between sensitive and non-sensitive explosives as well as the non-explosive materials.

An interesting application was reported by Chen et al.,77 who reported the use of LIBS followed by input of the analytical data to a machine-learning algorithm for the classification of waste paper from incineration smoke. Four types of paper were used in the experiment: tissue paper, corrugated paper, printing paper and newspaper. By measuring species such as Al, C, Ca, Mg, N and O and inputting the data into LDA and Random Forest, it was possible to classify each paper type with an accuracy of 95.83%. As an academic exercise it was interesting, but how useful it would be at an incineration plant is less clear.

There has been concern that tattoo inks contain toxic metals and are therefore harmful to health. A paper by Evgenakis et al.78 determined eight analytes (As, Cd, Co, Cr, Cu, Hg, Ni and Pb) in 20 different tattoo inks in the Greek market that had been imported from America and China. Samples (5 g) were freeze-dried and then 50 mg was digested using nitric acid with microwave assistance. The digests were then diluted to 10 mL prior to analysis using AAS. A slightly modified procedure was required for the Hg, where acid digestion occurred using a mixture of nitric and hydrochloric acids. The Hg was determined using vapour generation-AAS. All tattoo inks contained Cr, Cu, Ni, and Pb, while As, Cd, and Co contents were below the detection limit in 5, 3, and 8 out of the 20 inks, respectively. The Hg was only detected in three inks. In general, the concentrations of the analytes were similar in both American and Chinese inks. However, the most contaminated ink, by far, was a brown one imported from America. Method validation was achieved through spike/recovery experiments. The authors believed that their method of freeze-drying followed by acid digestion offered significant advantages (mainly of speed and lack of analyte loss) over other preparation methods.

3.2. Pharmaceutical samples

This section has seen the largest number of papers published. Trends are hard to see, but it is evident that the use of LIBS is far less prevalent than it is in many of the other sections of this update. Some papers do report the analysis of solid materials directly but, with one or two exceptions, these tend to use XRF analysis.

A review containing 153 references and entitled “Analytical approach of elemental impurities in pharmaceutical products: a worldwide review” was presented by Aleluia et al.79 After an introduction, there was a lengthy section on elemental impurities, their importance and occurrence. Then followed a section entitled “Analytical strategies for the determination of inorganic species”. This section was split into numerous sub-sections covering spectrophotometric methods, AAS, XRF, ICP-based techniques, speciation techniques and “other” techniques. The “other” sub-section included techniques such as LIBS, microwave plasma-OES, SEM-EDS and TOF-MS. Many of the references are discussed in the text but there are also useful tables. Conclusions and a future perspective were also provided. The review gave advantages and disadvantages of each technique and is a good read.

Quite a few of the papers discuss or provide sample preparation methods. It is true to say that sample preparation is often the most time-consuming and labour-intensive part of an analysis. Therefore, any method of improving the efficiency of sample preparation is to be welcomed. To these ends, Danilov et al.80 described a rapid method of preparing vitamin mineral complexes prior to TXRF analysis. Samples from three manufacturers were simply suspended in ethylene glycol. Results were compared with those obtained following an acid digestion followed by ICP-OES analysis. Unfortunately, agreement of results between the two techniques was not good. In addition, the results obtained using ICP-OES tended to be in better agreement with expected values quoted by the drug manufacturers. The TXRF did not detect Cr or Ni but the ICP-OES did. This was ascribed to potential contamination during the digestion procedure because they were not declared by the drug manufacturer, i.e., they were not supposed to be there. The advantage, according to the authors, was the speed of preparation and the lack of contamination. A second similar paper by the same research group again compared TXRF data with those obtained using ICP-OES.81 This time, as well as the sample suspension in ethylene glycol method of preparation, the same acid digestion protocol used for ICP-OES was tested for TXRF. This time, better agreement between sample preparation methods and analysis techniques was achieved. The TXRF data were notably less precise though. Recoveries of spikes ranged from 80 to 120%.

Another paper to report a sample preparation method was presented by Hao et al.82 who described a method for digesting Rosuvastatin calcium tablets. Samples were crushed and powdered and then underwent a microwave-assisted digestion procedure (0.2 g of sample plus 5 mL nitric acid and 50 μL of hydrochloric acid). Once digested, the samples were evaporated to almost dryness before being diluted to 100 mL using 2% nitric acid. The acid mixture used for digestion was optimised to ensure complete dissolution of the titanium dioxide excipient. The resulting digests were then analysed using ICP-MS with As, Cd, Co, Hg, Ni, Pb and V being the analytes and Bi, In and Y being internal standards. The analysis was undertaken using kinetic energy discrimination and with a collision cell. Validation was achieved through spiking/recovery experiments, with the spike being added prior to the microwave-assisted digestion. Recoveries ranged from 91.8–103.6% and precision was 1.8% RSD.

Other papers have also used microwave-assisted digestion of pharmaceuticals. One example was presented by Shah et al.83 who used a mixture of nitric and hydrofluoric acids to dissolve doxycycline hyclate tablets prior to ICP-MS detection. Analytes from class 1 (As, Cd, Hg and Pb) and from class 2 (Co, Ni and V) contaminants were determined simultaneously. These authors also validated their protocol using spike/recovery experiments. Specificity, linearity of calibration, LOD and precision were all evaluated and the conclusion was that the overall methodology was a suitable approach. Another example was presented by Senger et al.84 Most digestions use acid or a mixture of acids to ensure complete dissolution. This paper reported the “greening” of the process, i.e., it attempted to not use any acid. Instead, it undertook the digestions using only hydrogen peroxide and obtained the ICP-OES data for Ag, As, Ba, C, Cd, Cr, Cu, Li, Mo, Ni, Pb, Sb, Se and V. The temperature program was optimised, with 250 °C being optimal. Using the optimal program, up to 500 mg of material could be digested without leaving undissolved residue. When the procedure was applied to different anti-hypertensive pharmaceuticals, recoveries from spiking experiments were 91–110% for all analytes and precision was better than 6.4% when six replicates were prepared on the same day. Intermediate precision was also measured. This is where six replicates were prepared over three days and analysed. Here, values better than 10% RSD were obtained. As a further quality control measure, the reference material NIST 1640a was also analysed. Although this is clearly not matrix-matched to the samples, it would at least demonstrate that the calibration standards were prepared correctly.

Another paper to report a “green” sample preparation method was provided by Abellan-Martin et al.85 These authors developed a dispersive liquid–liquid micro-extraction procedure using a natural deep eutectic solvent (menthol and decanoic acid) to extract As, Cd, Hg and Pb from two drug types. These were: Omeprazole (for peptic ulcers) and levothyroxine sodium for thyroid problems. Following a microwave-assisted digestion of the solid form, the digests were diluted to 25 mL with 8-hydroxyquinoline and thiourea at final concentrations of 1.0% m/v for both reagents in a pH 4.4 acetate buffer. The extraction procedure was employed on the diluted digests. The digest (8 mL) was extracted with 80 μL of extractant, shaken and then centrifuged. An aliquot of 60 μL of the extractant was then taken and analysed directly using ICP-OES and cold vapour generation ICP-OES The procedure for the actual measurement was quite novel. A multi-channel nebuliser was employed in which sample passed through one channel and the reductant through the other. They mixed at the tip of the nebuliser and the vapour swept to the plasma for detection. This introduction method led to a sensitivity enhancement of approximately 25 and LOQ improvement of 50 compared with standard nebulisation. Again, the accuracy of the method was assessed using spike/recovery experiments. The “greenness” of the methodology was calculated using the AGREEprep metrics and was found to be excellent, largely because of the reduced volume of a non-hazardous solvent being required.

One final paper that developed a new extraction protocol was presented by Noronha et al.86 A univariate optimisation was performed to evaluate three different acid mixtures: 2.87 mol L−1 nitric acid; inverse aqua regia (3[thin space (1/6-em)]:[thin space (1/6-em)]1 nitric[thin space (1/6-em)]:[thin space (1/6-em)]hydrochloric acids); and an acid mixture composed of nitric[thin space (1/6-em)]:[thin space (1/6-em)]hydrofluoric acid in a 9[thin space (1/6-em)]:[thin space (1/6-em)]1 ratio. A sample mass of 30 mg and a typical acid volume of 200 μL were used, meaning that the protocol also had “green” credentials. A full description of the method was provided in the paper but rather than relying on microwave assistance, it used ultra-sonication. Statistical evaluation of the three acid mixtures as well as a reference method (microwave-assisted extraction) indicated that most of the results were not significantly different for 22 of the 24 analytes. The two errant analytes were Ag and Pt. Recoveries of spikes were generally in the region of 85–120% and precision was typically better than 10% RSD. The method developed was described as simple and fast. Since it used a reduced amount of sample and the minimal volume of reagents, a significant saving of energy resulted, along with a reduction in the amount of residue generated. Additionally, since it is performed at room temperature and pressure, the proposed procedure is relatively safe compared with other procedures such as microwave-assisted sample digestion.

Two papers have coupled liquid chromatography with ICP-MS for the analysis of pharmaceutical materials during this review period. The first, by Horstmann et al.,87 described the speciation analysis of technetium-based radiopharmaceuticals using both HPLC-ICP-MS and HPLC-electrospray ionisation (ESI)-high resolution mass spectrometry (HRMS). This combination of atomic and molecular spectroscopies enabled direct identification, characterisation and quantification of two novel and four established radiopharmaceuticals. The high-resolution Orbitrap-based HRMS identified the compounds conclusively, while the HPLC-ICP-MS enabled the separation and sensitive detection of Tc species that may be parent molecules, metabolites or contaminants. The reversed-phase chromatography system was described in detail. Since standards of Tc are not readily available, a novel on-line isobaric dilution analysis was described. This involved a Ru post column spike continuously being added after the chromatographic column generating a constant ratio of m/z 99 and 101. The disruption of this ratio arising from the elution of 99Tc was used for its quantification. The second paper of this type was presented by De Maria et al.,88 who studied the metalation of thioredoxin by gold therapeutic compounds including the drug auranofin and two heterocyclic carbene–Au complexes. The materials were analysed using direct infusion electrospray ionisation mass spectrometry enabling the structure, stoichiometry and kinetics of formation to be studied. Reversed-phase HPLC experienced problems in eluting the Au compounds and also led to the breakup of the metal-protein adducts. No such problems were experienced when capillary electrophoresis was used. The ICP-MS/MS as a detection system enabled the simultaneous determination of S and Au isotopes which, in turn, enabled the extent of the metalation of the protein to be determined.

A forensic study of psychotropic drugs was reported by Degardin et al.,89 who used the techniques of XRF NMR, GC-MS, HPLC and near-IR to discriminate between genuine and counterfeit samples. Several differences were observed with many of the techniques. With regard to the XRF analysis, the counterfeit drugs were found to have significantly less Mg present (between 0 and 0.04%) compared with the genuine samples (0.08%); the counterfeit drugs do not have to be prepared to the same rigorous standards as the genuine article and therefore their components tend to be more variable. Despite this, with the assistance of PCA, it was possible to identify some trends that would be of use to law enforcement officers when attempting to prosecute.

The LIBS analysis and discrimination between six very similar cold medications, with data reduction using t-distributed stochastic neighbour embedding followed by a back propagation neural network for classification, was reported by Yao et al.90 All medications had paracetamol and amantadine hydrochloride as active ingredients. Once obtained, the LIBS data were treated using t-distributed stochastic neighbour embedding to reduce the dimensionality. This algorithm performed better than others that were also tried, e.g., PCA, local semantic analysis, independent component analysis, isomap and local linear embedding. The high self-learning and self-adaptive ability of the back propagation neural network enabled an accuracy of classification of 96.7%. The ability to classify different similar medicines was thought to be beneficial when detecting counterfeit products and so this relatively quick and accurate method is fit for purpose.

3.3. Cosmetic samples

An overview of heavy metal contamination in cosmetics was presented by Gyamfi et al.91 It contained only 47 references, so was far from a comprehensive review. However, it did concentrate on 16 papers published between 2012 and 2020 that had determined metals such as Cd, Cr, Fe, Hg, Ni and Pb in a variety of cosmetic samples. Analysis had been undertaken using AAS, ICP-OES and ICP-MS. The presence of such elements is a potential health risk and different countries have different regulations. This makes it challenging for cosmetic companies. The authors stressed the need for standardisation of the acceptable limits.

Two papers have employed the technique of single-particle (sp)-ICP-MS to analyse cosmetic samples. In one by Garcia-Mesa et al.92 both sp-ICP-MS and high-resolution continuum source GFAAS (HR-CS-GFAAS) were used for size characterisation and speciation of ionic and nanoparticulate Zn in cosmetic samples. It was stressed that the size of the particulates could affect the toxicity and therefore, to obtain the best estimate of toxicity, the samples had to be characterised accurately. The sample preparation methods were slightly different for the two techniques, but both involved suspending the homogenised sample in Triton X-100 and then sonication to disaggregate any agglomerates. The two techniques required different concentration of sample and of Triton X-100. The sp-ICP-MS samples then required vortex mixing immediately prior to a further dilution. The optimal conditions (weights of sample, concentration of Triton X-100 and sonication times) were determined using central composite design. For HR-CS-GFAAS, two calibration curves were required: one for ionic Zn and the other for particulate. The latter was achieved using ZnO nanoparticle UV shielding powder standards with size between 18 and 500 nm. How the particle size distribution was calculated from the GFAAS signal was described in detail in the paper. The ionic and particulate Zn concentrations found in eye shadows, and assorted creams and lotions were compared with “total” Zn determined using an acid digestion. Results were in good agreement. The size distribution obtained using sp-ICP-MS was calculated using a standard commercial package. A comparison of HR-CS-GFAAS and sp-ICP-MS data was made as well as with those from TEM analysis using two different diluents (ethanol and 1% Triton X-100). In general, reasonable agreement between the techniques was obtained. The HR-CS-GFAAS method is cheaper and the data treatments are simpler and faster than those for sp-ICP-MS. It also enables speciation of ionic and particulate, which sp-ICP-MS cannot do. In the second paper, by Hebert et al.,93 assorted facial powders and eye shadows of different colours and prices were analysed for several analytes (Ag, Al, Bi, Cr, Mg, Mn, Pb, Sn and Zn), These authors also adopted suspension in Triton X-100 (0.1 g of sample in 50 mL of 1% Triton) as a means of sample preparation. All seven of the samples analysed contained eight of the analytes with five containing all analytes. Only Bi was missing from two samples. Nanoparticles of diameter <100 nm were found in all samples. However, there appeared to be no correlation between cost and nanoparticle content.

Two relatively simple applications have also been published. In one by Liu et al.,94 a green hydrophobic deep eutectic solvent comprising menthol and hexanoic acid was used to dissolve cosmetic samples. The analyte Cd (plus Cu, Mg and Mn) was then extracted using disodium EDTA solution and the analytes measured using FAAS. The LOD for Cd was 0.037 mg kg−1 and precision was 3.5% RSD. Recovery values for spikes were in the range 85.5–118.3%, which was deemed acceptable. An advantage of the process was that the eutectic solvent was readily recycled so waste was very low. The other simple application was presented by Guerranti et al.,95 who determined Cd, Co, Cr, Ni and Pb in 90 cosmetic products (eye shadow, eye liner, blush, mascara and lipstick) available on the Italian market. The samples were prepared using a microwave-assisted nitric acid digestion, filtered and then analysed using ICP-MS. Interestingly, no reaction/collision cell gas was used for Cd, Co and Pb, ammonia was used for Cr and methane was used for the determination of Ni. Method validation was again through spike/recovery experiments, with the spike being added prior to the digestion. The results appeared somewhat variable with some, e.g., Ni in eye shadow, having a recovery close to 100%, whereas others, e.g., Cd, had a recovery of only 70%. Of the analytes, only Cd and Pb were constantly below the limit proposed by the Italian National Institute for Health.

3.4. Fuels and lubricants

The number of papers submitted on this subject was roughly the same as last year; however, many were of low quality, describing research that is neither new nor novel. It was also surprising that there was a lack of work undertaken on fuels containing ethanol and synthetic aviation fuels. Both are ‘hot topics’ in the industry at the moment and both have analytical challenges that need addressing but they are hardly mentioned in the literature. Crude oil, coal and petrochemicals in general can be seen somewhat as a pariah subject currently, which may account for the lack of interest academically, but there is a huge amount of work that needs to be done on bio and synthetic fuels and around processes for recycling plastics. This ‘green’ subject does not seem to be getting the focus it requires, as there are still industries, principally aviation and petrochemicals, for which there are no current alternatives to using hydrocarbons.

Research in coal this year seems almost exclusively to be being undertaken in China, and most papers on this subject are based around different algorithms for the improvement of predicted ash contents in coals by analysing the elemental composition by LIBS. A popular topic also seems to be alternative sample preparation methods for analysis of fuel and lubricating oils using various emulsion and emulsion breaking techniques to transfer elements of interest into the aqueous phase, which can then be analysed against aqueous standards. The aim of reducing solvent use in this analysis is laudable; however, most of these papers are based around a restricted number of elements, none contain all the elements in the industry standards ASTM D5185 and ASTM D7111, and the sample preparation methods proposed are complicated and time-consuming. This is obviously not what is required in a busy laboratory.

3.4.1. Petroleum products – gasoline, diesel, gasohol and exhaust particulates. A lot of the submissions in this area were neither new nor novel, with papers reinventing the wheel of microwave digestions and sample preconcentration by evaporation. Only one paper in this section was worthy of note; it was by da Silva et al.96 and described a method of reverse-phase dispersive liquid–liquid microextraction for the determination of Ca, Cd, Cu, Fe, Mg, Mn, Ni, Pb and Zn in petroleum-derived samples using ICP-OES. This method used 5.0 g of sample and 1600 μL of 3.75% HNO3[thin space (1/6-em)]:[thin space (1/6-em)]isopropyl alcohol 3[thin space (1/6-em)]:[thin space (1/6-em)]1 v/v extraction solution, which was then placed for 10 min in a thermostatic bath at 80 °C. This was followed by 60 s of vortex mixing, 3 min of sonication at 55 °C and 15 min of centrifugation at 3500 rpm. The resulting aqueous-phase volume was adjusted up to 15 mL with water and then ICP-OES analysis carried out. Relative standard deviations were better than 3.5%, linear correlation coefficients in excess of 0.9985, and LODs in the range of 0.57 ng g−1 for Mn to 107.6 ng g−1 for K were obtained. Accuracy was evaluated by analysis of the V21 + K metallo-organic standard, and recoveries ranged from 89.4–98.3% with RSDs of less than 6.4%. A recovery test was carried out and the results were in the range of 87.1–95.7%. This methodology was applied to gasoline, diesel, jet-A1, avgas, lubricant oil and Vaseline samples.
3.4.2. Coal, peat and other solid fuels. Two papers and two reviews on this subject were worthy of note this year. The first paper by Zhu et al.97 described a combined dual-cycle variable selection mechanism with competitive adaptive reweighted sampling to optimise partial least-squares regression. In this work, a new quantitative analysis method was proposed that was used to correct for matrix effects and improve the performance of regression models using LIBS. The matrix correction of elemental carbon and fixed carbon is realised by qualitative classification and by using generalised spectra, respectively. The coefficient of determination of the training set and test set of elemental C was improved from 0.7178 and 0.7095 to 0.9534 and 0.9515. The root mean-square error of cross-validation (RMSECV) and the root-mean-square error of prediction (RMSEP) decreased from 1.1723 wt% and 1.2488 wt% to 0.4504 wt% and 0.4871 wt% and the R2 of the training set and test set of fixed carbon was also improved from 0.8252 and 0.8236 to 0.9814 and 0.9749. The RMSECV and RMSEP decreased from 1.2274 wt% and 1.3548 wt% to 0.4421 wt% and 0.4711 wt% respectively.

The second paper by Lu et al.98 described a study to develop a rapid compact integrated coal quality detection instrument based on LIBS, which can directly measure C content and thus calorific value in coal particle flow. A partial least-squares model, based on data set selection according to cluster analysis, was applied to establish the relationship between coal quality and plasma spectra. The R2 of the calorific value with this method was 0.93, the root mean square error of prediction (RMSEP) is 0.41 MJ kg−1 and the mean absolute error (MAE) was 0.33 MJ kg−1. The R2 of C content was 0.94, the RMSEP was 0.97%, and the MAE was 0.91%. These results indicated that the developed instrument could conduct precise coal quality analysis in real time.

Two reviews were of interest in this area this year. The first by Dong et al.,99 containing 225 references, looked at LIBS and spontaneous emission techniques for monitoring the thermal conversion of fuels. This review discussed the application of atomic emission spectroscopy for monitoring the fuel thermal conversion process, including the characterisation of raw energy materials (coal, biomass) before combustion, during the fuel combustion process (flame) and after combustion (fly ash). The first section focussed on LIBS for the analysis of raw energy materials in different forms like rock, pellets and particle flow. The second section focussed on the characterisation of the fuel combustion process with the use of LIBS and spontaneous emission. The third section was on unburned-C detection in fly ash using LIBS.

The second review by Damdinsuren et al.,100 containing 27 references, was a brief overview of XRF applications in Mongolian brown coal. This study placed specific emphasis on parameters including the ash content and calorific value of lignite, which is the main energy source in Mongolia. In the research, it was observed that the ash content of coal can be estimated by analysis of the concentrations of chemical elements, including Ca, Fe, and Sr and an inverse linear correlation was established between the calorific value of coal and its ash content.

3.4.3. Oils – crude oil and lubricants. Only two papers were worth noting in this section this year and there were virtually no papers on the subject of crude oil. The first paper by Abellan-Martin et al.101 described the application of a multinebuliser for analysis of wear metals in used lubricating oils using MIP-OES. Conventional ICP-OES is the standard method for this analysis (ASTM D5185); however, there are situations where Ar is not readily available and alternative analysis options must be used. This paper tackled the problem of matrix problems and carbon build up in the analysis of oils when using MIP-OES. This was brought about by the addition through the nebuliser of an aqueous stream alongside the organic introduction and these two streams were mixed together at the nebuliser tip. This approach negated the need to use additional air or oxygen to stabilise the plasma and burn off deposits, which can form on the torch as oxygen was provided via the aqueous stream. It also allowed standard and internal standard to be introduced using aqueous solutions. Four methods of analysis were compared: external calibration with organic standards and internal standard addition, standard addition and standard dilution analysis all of which used aqueous standards. The method appeared to work well and the best method for accuracy, precision and ease of use was the standard dilution analysis approach. Six elements were determined, Al, Cr, Cu, Mn, Ni and V, with spike recoveries ranging from 95–106% for the preferred method.

The second paper in this section was by Castillo et al.102 In this paper silicon nanoparticles were assessed for the removal of vanadium-containing asphaltene aggregates. Both GPC-ICP-MS and single-particle ICP-MS were used to assess the aggregate uptake in the nanoparticles with the spectra compared for treated and untreated crude samples. The results indicated that silicon nanoparticles could have practical applications for the separation and removal of V from crude oil and provide information on asphaltene aggregate adsorption mechanisms.

3.4.4. Alternative fuels. This is where analytical methods for fuels containing ethanol, bio-diesel, synthetic aviation fuels of various types and plastic to fuel process samples are reviewed. However, papers on these subjects are almost absent from the literature. Only two papers and one review were worth mentioning. The first paper by Almeida et al.103 described a method for determining Cd, Cr, Cu and Fe in biodiesel samples by EDXRF after extraction induced by emulsion breaking. In this method 7.0 mL of biodiesel and 500 μL of Triton X-100 solution 2%, v/v prepared in nitric acid were used to prepare the emulsion. After 5 min of manual agitation, phase separation was performed by centrifugation for 5 min at 2000 rpm. A 100 μL aliquot of the aqueous phase was collected and placed in the centre of a filter paper. After drying, the analysis was performed directly on the paper using EDXRF. The detection limits achieved were 60, 20, 30, and 60 μg kg−1 for Cd, Cr, Cu, and Fe, respectively. Recovery tests were performed by adding 450 μg kg−1 for the analytes in biodiesel samples and the recovery results varied between 87 and 113%. The procedure was applied to biodiesel samples, and the concentrations obtained for the four analytes were compared with those obtained using ICP-OES after sample digestion.

The second paper in this section, by Endriss et al.,104 addressed the impacts on XRF analysis for rapid determination of the chemical composition of renewable solid biofuels. Rapid determination of the quality of solid biofuels enables transparent fuel trading and optimised plant operation. This paper investigated the impact of interferences caused by the chemical composition, particle size, water content and measuring time on rapid measurement using XRF analysis for solid biofuels. The elements investigated were the minor elements Ca, Cl, K, Mg, Na, P, S and Si and the trace elements Al, As, Cd, Co, Cr, Cu, Fe, Mn, Ni, Pb, Ti and Zn, as described in ISO 17225-1. The results provided new insights into the cause of measurement errors. It was found that grain size <1 mm in sample preparation and water content ≤10% had a clear benefit on the measurements. In the case of samples with high mineral content, interferences between the elements Si and P occurred. Furthermore, the results showed that the time for the actual measurement can be significantly reduced to 60 s compared with the factory recommended measurement time of 750 s. The findings of the study contribute to reducing or preventing measurement errors in future XRF analyses.

The final paper in this section is a review also by Endriss et al.,105 containing 260 references, which looked into the analytical methods for the rapid determination of solid biofuel quality. Fuel properties of solid biofuels have a major impact on the energy-efficiency and emissions of biomass heat and power plants; hence, the fuel quality parameters are often defined and used for pricing in supply contracts. To simplify and accelerate analytical approaches, rapid analysis devices are required to determine fuel properties such as water, ash content, calorific value and chemical composition on-site. This article gives an overview about available technologies and their current use as rapid analysis devices for solid biofuels.

3.5. Polymers

Polymer analysis has been extensively documented in the scientific literature during this review period. However, the two most prominent topics frequently addressed in connection with atomic spectrometry are the management and classification of polymer waste and the analysis of micro- and nanoplastics. These topics, along with other applications deemed significant from the perspectives of atomic spectrometry and polymer analysis, are discussed in detail below.
3.5.1. Reviews. Consistent with the key topics identified during this review period, two relevant reviews – one focused on the identification of plastic waste and the other on the analysis of microplastics – deserve focus in this ASU review. Yang et al. summarised the research progress in the application of spectroscopic techniques combined with machine learning in the rapid identification of plastic waste during the past five years.106 The review contained 232 references and focused on the innovative research of machine-learning methods in plastic identification and on the advantages and disadvantages of various spectroscopic techniques. The review offers theoretical background on machine learning and deep learning, with a focus on traditional statistical analysis methods and machine-learning approaches relevant to the field. Additionally, the authors provided an overview of selected spectrometric techniques – including near-infrared (NIR) spectroscopy, mid-infrared (MIR) spectroscopy, Raman spectroscopy, terahertz (THz) spectroscopy, LIBS and XRF spectroscopy – and discussed the current state of these techniques when combined with machine learning. Tables were provided that highlight studies from recent years where these techniques were combined with machine-learning methods to detect plastics, making it easier for readers to find specific applications and references. A section giving insights on the future trajectory of this field and the proposed recommendations for its advancement was also provided. The authors suggested combining and fusing spectra to overcome the shortcomings of single-spectral techniques for plastic waste identification, although further research in this context, especially when combined with deep learning, is still required. Overall, the review concluded that spectroscopic technology combined with machine-learning methods is expected to become an important means of automated sorting of plastic waste.

The other review of relevance to this section, this time by Wu et al., focused on the progress on characterisation, environmental behaviors and toxicity effects of micro(nano)plastics using atomic spectrometry.107 This review is timely, taking into account the rising concern about the impact of micro/nanoplastics (MNPs) in the environment and human health, and the significant number of research papers on the topic published during this review period. With the aid of 120 references, the review paper introduces the commonly used analytical methods for MNP analysis, highlighting that research on the analysis of MNPs is mainly concentrated over the last three years. It has focused on the identification, characterisation and migration in the environment. It indicates that the study of MNPs by atomic spectroscopy has developed rapidly, with LIBS, ICP-MS and TOF-SIMS all being effective analytical methods at present. However, the review did not go into much experimental detail of the techniques. A relevant section focused on the sample preparation and pretreatment of MNPs prior to analysis, with significant information provided on labelling strategies (a table summarised the different labelling methods and their applications) and on the separation, and digestion of the sample before MNP detection. In the conclusion section, several challenges hampering the application of atomic spectrometry in the analysis of MNPs were highlighted, focusing on (1) the need for developing alternative and more affordable labelling strategies with high detection sensitivity, (2) the miniaturisation of atomic spectrometers to facilitate in situ measurements, (3) the development of a high-throughput system that provides information on the chemical characterisation, morphology, and quantitative analysis of MNPs, and (4) the development of a standard methodology for appropriate validation.

3.5.2. Sorting of polymers for waste management. As mentioned above, one of the main areas of polymer research during this review period is that of waste management, particularly the management of electronic waste plastic (e-waste plastic). E-waste plastic often contains harmful additives, such as brominated flame retardants and heavy metals, which pose significant environmental and health risks. As a result, several of the applications published during this review period focused on the analysis of e-waste plastic and on the risks associated to recycling this plastic material.

Holub et al. evaluated the capabilities of LIBS for the direct in situ quantification of the Pb content in polymer post-consumer recycled materials.108 The manuscript highlighted the production of ad hoc calibration standards and the implementation of the methodology for the detection of Pb in polymer (PVC) recycled materials. A LIBS system equipped with a 532 nm laser source, a Czerny–Turner type spectrometer with a high resolution and a CCD camera was used for the measurement of un-processed and processed post-consumer waste polymer samples. A range of wavelengths from 382.41 to 407.61 nm was captured to achieve information containing the CN transition area and a Pb line (388.30 nm and 405.81 nm, respectively). This provided information both about the matrix polymer (nitrile compound) and the target analyte (Pb). The CN violet 0,0 band was used for standardisation to account for signal fluctuations caused by different matrices of the samples in the selected data set. After LIBS measurements, chemometrics were used for analysis of the data obtained. Measurement precision was significantly affected by sample heterogeneity, with samples that were made from more homogeneous polymer waste debris displaying more precise results. The LOD achieved for real-life processed polymer waste material samples (227 ppm) was found to be low enough to control the Pb level relative to the limits set by the EU legislation (1000 ppm). After setting up a calibration sample set, 15 samples of post-consumer polymer waste were analyzed. The results showed that 13 out of 15 samples exceeded the EU legislative limit. The method showed potential for in situ application in recycling plants to control heavy metal content in polymer recycled materials. The use of a simple univariate analysis allowed for a precision of 87.5% and a satisfactory accuracy (74.1%). To further improve the figures-of-merit, multivariate techniques, including principal component regression (PCR) and partial least-squares regression (PLSR), enabled accuracies of the predicted Pb content of more than 90% to be achieved with more than 86% relative precision.

Similarly, Turner et al. evaluated the presence of hazardous elements from the recycling of e-waste plastic in artificial plastic plants.109 In this study, 203 parts or offcuts from 175 plastic plants, acquired from European shops and venues, were analyzed using XRF to determine the elemental content, while a selection was analyzed using FTIR to establish the polymer type. The FTIR analysis revealed that the usually green moulded components were made of polyethylene (PE) or polypropylene, while leaves and colourful petals were generally made of polyethylene terephthalate (PET). After XRF analysis, variable concentrations of potentially hazardous elements were found in the moulded parts, including Br (6.1–108[thin space (1/6-em)]000 mg kg−1), Pb (7.6–17[thin space (1/6-em)]400 mg kg−1) and Sb (58.6–70[thin space (1/6-em)]800 mg kg−1). These concentrations suggest that the moulded material appears to be sourced from post-consumer recyclates that are contaminated, to varying degrees, by e-waste plastics, introducing brominated flame retardants, the flame-retardant synergist, Sb2O3 and Pb. Additionally, the presence of Ba, Fe, Ti, and Zn reflects the addition of common inorganic pigments and fillers, such as barium sulfate, various Fe oxides, titanium oxide and zinc oxide, to reduce costs and improve performance or appearance. High concentrations of Cr (>1000 mg kg−1) were restricted to samples containing high concentrations of Pb or that were painted red, suggesting the presence of lead chromate and red chrome oxide, respectively. Although there are no standards for these chemicals in plastic plants, regulations for electronic plastic were exceeded, limiting their recycling and disposal.

Also in the context of plastic management for waste electrical and electrical equipment recycling, Chaine et al. studied the presence of brominated flame retardants in e-waste plastics to compare the regulatory systems between Europe and Latin America.110 For the identification of brominated flame retardants in plastics, an internationally validated screening methodology using XRF was adopted at different processing operations. The authors found that, using a threshold value of 830 mg kg−1 for Br as a tracer, more than 70% of the plastics would be recyclable, while that fraction dropped to 50% in Uruguay. The results highlighted the impact that regulatory frameworks have on the quality and recyclability of recovered material.

The final waste management application focuses on characterising the polymer properties and identifying additives in commercially available research plastics.111 The study aimed to provide baseline data for polymers that are commonly used as benchmark substrates for research and that are commercially available. The authors characterised 59 polymers from common commercial vendors across 20 different polymer classes, representing >95% of the global plastic production by mass. This characterisation included molecular composition, polymer morphology, molecular mass distribution, thermal properties, elemental compositions, the presence of additives, and the effects of cryo-milling on structural and thermal properties. This work provided a baseline analytical characterisation of plastic and plastic additives for recycling studies and, more importantly, it offers the research community readily accessible, fully characterised substrates. However, it is important to note that the measurements were conducted on a specific set of polymers from particular companies with provided lot numbers. Distributors may change between lot numbers, resulting in different formulations. Additionally, the aging of the plastic can also alter its properties.

3.5.3. Micro- and nanoparticulate polymers. One of the primary areas of polymer research during this review period has been that of the analysis of micro- and nanoplastics. A significant increase in the number of publications on these materials compared with previous years has demonstrated the growing relevance of this research topic within environmental sciences. Clearly, this is an emerging field in which a fast development can be anticipated, with atomic spectrometric techniques playing a key role in the analysis and risk identification of these minute plastic particles.

Among the various analytical techniques available for MNP analysis, ICP-MS operated in single-event (or single-particle; sp) mode has emerged as particularly prominent over the past year, with numerous significant contributions in this area. The technique of sp-ICP-MS can simultaneously provide various types of information of the micro or nanoparticles present, such as the particle number (particles per mL) and mass concentration (mg L−1), the elemental composition, size (spherical equivalent diameter – nm), and size distribution. In this context, three different strategies for analysis can be identified: (1) characterisation of the materials by relying on their carbon content, (2) detection of the particles through metal labelling in exposure experiments, and (3) identification of particles through multi-element fingerprinting. Table 1 summarises the advancements in micro and nanoparticle polymer analysis via sp-ICP-MS during this review period, while some illustrative examples are discussed in detail below.

Table 1 Advancements in micro- and nano-polymer analysis via sp-ICP-MS during this review period
Sample type/matrix Polymer type Analyte Sample introduction system Detection strategy, size range, information obtained Reference
Teabags and face masks PS, PE, PP, and PES 13C Concentric nebuliser + cyclonic spray chamber Direct, 1–6 μm, size and PNC Sakanupongkul et al.112
Seawater Multi-element-doped PS beads 12C, 31P, 133Cs, 140Ce, 153Eu, 165Ho, and 175Lu PFA MicroFlow pneumatic nebuliser + cyclonic spray chamber Direct, 3 μm, size and quantification of REEs Harycki and Gundlach-Graham113
Water Multi-element-doped PS beads 12C, 31P, 133Cs, 140Ce, 153Eu, 165Ho, and 175Lu Microdroplet generation system (MDG) Direct, 3 μm, size Vonderach et al.114
Water/PVDF and glass microfibre filters PS, PMMA, and PVC micro-polymers 13C Laser ablation system (LA) Direct, 2–20 μm, size Van Acker et al.115
Water containing DOC and/or C-based particles PS MPs and Pd-doped PS nano-polymers 12C, 13C, 106Pd, and 108Pd Single-cell sample introduction system Direct and indirect (metal-doped), 0.2 and 4 μm, size (MPs) and PNC (NPs) Hendriks and Mitrano116
Human cell cultures Pd-doped PS nano-polymers 24Mg, 31P, 56Fe, 64Zn, 65Cu, 104Pd, 105Pd, 106Pd, and 108Pd Single-cell sample introduction system Indirect (metal-doped), 200 nm, NP association with biological systems Hendriks et al.117
Biological samples/organism species PS, PS-COOH, and PS-NH2 197Au Concentric nebuliser + conical spray chamber Indirect (labelling), 100 and 500 nm, PNC Li et al.118
Water and biological sample PS, PMMA, PVC, LDPE, and PVP 90Zr, 118Sn, and 181Ta Concentric nebuliser + cyclonic spray chamber Indirect (metal-tagged), 0.3–1 μm, size and PNC Smith et al.119
Consumer products and ocean plastic fragments PP, PET, LDPE and PS 27Al, 28Si, 48Ti, 52Cr, 56Fe, 65Cu, 66Zn, 88Sr, 114Cd, 138Ba, and 208Pb Concentric nebuliser + cyclonic spray chamber Adsorption of metals, 100–800 nm, metal content Baalousha et al.120


Although micro and nanoparticulate polymers are C-based particles, it was a priori considered that ICP-MS was not appropriate for monitoring the C signal. However, extensive research over the last few years has proved that there is enough potential to measure C-based nanomaterials or, at least, micromaterials using ICP-MS. Hendriks et al. directly measured micro-polymers using C detection via sp-ICP-time-of-flight-MS (sp-ICP-TOF-MS). Mono-disperse polystyrene particles with a nominal diameter of 4 μm were used as model micro-polymers.116 In this work, a detailed optimisation of an ICP-TOF-MS instrument in low-mass mode was carried out. The multiplexed sp-ICP-TOF-MS information allowed for the distinction of individual micro-polymer entities (detection based on 12C+ only) and of other C-containing natural particles (i.e., algae cells) that were characterised by relying on the concurrently detection of 12C+, 24Mg+, and 31P+ in each pulse event. A dilute suspension of polystyrene beads was introduced into the ICP using a single-cell sample introduction system specifically designed for the transport of large entities and operated at a sample uptake rate of only 10 μL min−1. An increased dissolved organic carbon, added to the synthetic freshwater in the form of humic acid, did not compromise the sp-ICP-TOF-MS detection, achieving constant particle number concentrations in solutions with concentrations ranging from 0 to 20 mg L−1 C. It should be noted that the limited 12C+ sensitivity and high intrinsic C background hindered the detection of nanosized plastics and that this approach does not allow differentiation between different micro-polymers. However, this method is deemed suitable for those working in laboratory-based systems where micro-polymers are of known characteristics. Beyond measuring micro-polymers in complex aqueous suspensions, the method would also be suitable for assessing the sorption of inorganic trace elements or hetero-aggregation with other natural colloids.

Efficient transport of relatively large entities, such as micro-polymers, is limited when using traditional ICP-MS sample introduction systems. As a result, extensive research has been carried out during this review period aiming at developing alternative ways for introducing these plastic particles. Van Acker et al. characterised laser ablation as a method for sampling and introducing micro-polymers into a tandem quadrupole-based ICP-MS instrument (ICP-MS/MS).115 In this work, micro-polymers of different types (polystyrene, polymethyl methacrylate, and polyvinyl chloride) and sizes (2–20 μm) were introduced intactly, as demonstrated via comparison of the 13C+ peak profiles obtained upon LA-sp-ICP-MS/MS and sp-ICP-MS/MS detection (the duration of the signal peak profiles was of approximately 500 μs). The authors found that the laser energy density did not affect the particle integrity, and thus, the particle sampling, across a wide range (0.25–6.00 J cm−2). Both the 13C+ integrated signal intensities and peak widths remained constant with increasing laser energy density. Single-shot analysis was capable of separating clustered micro-polymers (up to 7 micro-polymer particles per cluster) during the LA and particle transport processes, allowing temporally resolved data to be obtained for the individual particles. To sample a higher number of micro-polymer particles per unit of time and scan larger areas, a line scanning approach characterised by a small laser beam diameter combined with a high repetition rate showed a superior performance compared with a lower repetition rate combined with a larger beam diameter. The use of LA as a means of sample introduction allowed micro-polymer particles with sizes of up to 20 μm to be efficiently introduced into the ICP, although no information concerning transport efficiencies was provided. A good linearity between the 13C+ signal intensity and particle volume was obtained for PS, polymethylmethacrylate (PMMA) and PVC micro-polymer particles of different sizes. It should be noted that the 13C+ signal intensities were proportional to the absolute mass of C in the micro-polymer particles and were not affected by the polymer type. Interestingly, the authors reported narrower signal intensity distributions for the LA approach when compared with the traditional solution-based approach, which was tentatively attributed to the differences in plasma conditions between both methods (dry vs. wet plasma). The developed method proved successful for analysing a mixed suspension containing four sizes of PS micro-particle and for sampling particles deposited onto different filter materials that are regularly used to extract micro-polymer particles from environmental matrices. The latter offers greater compatibility with other non-destructive analytical techniques compared with solution-based analysis, enabling the acquisition of additional relevant information, such as the polymer type, prior to LA-sp-ICP-MS analysis.

Compared with micro-polymer particles, nano-polymers (<1 μm) are a contaminant of even greater concern because of their high environmental concentrations, enhanced environmental mobility and greater bioavailability. However, the presence of nano-polymers has been consistently neglected in quantification studies due to the lack of sufficiently powerful and validated analytical methodology. As indicated above, the limited instrument sensitivity of ICP-MS for C monitoring hinders the direct detection and quantification of nano-polymers by relying on their C content. Alternatively, labelling strategies have been developed for the “indirect” detection of nano-polymers. To date, this approach can be deployed for laboratory experiments aiming at improving our understanding of the nano-polymer behavior. Smith et al. created irregularly shaped metal-tagged nano-polymers by cryo-milling lab-generated plastics containing 1% w/w concentrations of an organometallic additive; these metal-tagged NPs are detectable by sp-ICP-MS analysis.119 Metal tags were incorporated into each of the polymers via solution blending, where mixtures of organometallic additive and polymer were sonicated in an organic co-solvent. Cryo-milled plastic particles were added to deionised water at a concentration of 1 mg mL−1 and sonicated in an ice water bath. Subsequently, the supernatant containing the nanoparticulate fraction was removed prior to sp-ICP-MS analysis. The nano-polymer particles were counted and sized using the metal mass detected in each individual nano-polymer particle assuming a uniform distribution of metals in the plastic particles. The authors highlighted the advantages of their nano-polymer labelling strategy when compared with alternative labelling techniques. Among others, the detection of the metal-tagged nano-polymers synthesised by this method relies on the entire volume of the NP, in contrast to other labelling techniques that rely on adsorption, direct attachment, or the presence of metal atoms on the nano-polymer surface. As a result of this metal distribution, the metal additives are present at lower concentrations per particle than for other labelling techniques, thus minimizing the impact on the nano-polymer particle surface properties. Moreover, a significant experimental advantage of this method is the ability to use unique metal tags for different nano-polymer types, allowing for the concurrent detection and quantification of polyvinylpyrrolidone tagged with Ta, PMMA tagged with Sn and polystyrene tagged with Zr in the same sample solution. However, it should be highlighted that this method focuses on the so-called top-down approach, whereby the nano-polymer particles are generated by cryo-milling of the metal-tagged bulk polymeric material, rather than following a bottom-up approach. The latter might be more suitable for studying real-life nano-polymers occurring naturally in the environment, while this method is intended for laboratory experiments aiming at improving our understanding of the nano-polymers behavior.

In order to tailor the plastic products to a variety of needs and applications, metal and metalloid containing additives and catalysts are essential. These metals and metalloids are embedded in the polymer structure and can thus be used for the detection of micro and nano-polymers through elemental fingerprints. Baalousha et al. investigated the fingerprints in real-life nano-polymers generated from new plastic products and from environmentally aged ocean plastic fragments using sp-ICP-TOF-MS and TEM-EDX.120 The metal and metalloid content was compared with that of natural nanoparticles extracted from three soil samples. Data processing and data reduction consisted of classifying every detected particle signal into single- and multi-metal nano-polymers. The multi-metal nano-polymers were classified into clusters of similar elemental composition using a two-stage agglomerative hierarchical clustering. These model real-life nano-polymers exhibited irregular shapes, were rich in metals and metalloids, such as Al, Ba, Cd, Cr, Cu, Fe, Pb, Sr, Ti and Zn, and were depleted in rare earth elements. The relative abundance of metals and metalloids was higher in the environmentally aged particles than in the new plastic products. This was most likely attributed to sorption/attachment from the surrounding environment. The study illustrated the importance of nano-polymers as a source of metals and metal-containing nanoparticles in the environment and highlighted the need to develop analytical methods for tracking real-life nano-polymers through elemental fingerprints as an alternative to metal labelling.

The technique of ICP-MS has seen a remarkable increase in research activity for micro and nano-polymers analysis, warranting special attention in this review. However, other techniques and methodologies also deserve significant attention. As examples, LIBS and Raman spectroscopy have shown interesting features for research on micro-polymers. Vaisakh et al. evaluated the feasibility of an integrated LIBS-Raman spectroscopy system in monitoring water resources for micro-polymer-heavy metal detection with a unified sampling and pre-treatment protocol.121 Raman spectroscopy was proposed to characterise micro-polymers based on the polymer type, while detection of trace elements adsorbed on the micro-polymer surface was carried out using LIBS. Please note that LIBS can also provide structural information allowing for the identification of the polymer type. The surface adsorption process was evaluated under laboratory-controlled conditions to assess the extent of adsorption of trace elements on plastics over time. Micro-polymers were collected from various estuary river sites and were subsequently preprocessed to select the appropriate size fraction and remove organic matter. Raman and LIBS databases were created from pure plastic and metal samples aiming to identify the polymer type and heavy metal concentrations, respectively. All 21 micro-polymer samples collected from different sampling spots were characterised into the basic polymer type (PE, PET, and PP) based on the Raman spectral signatures. The capabilities of the LIBS system to determine trace concentration levels was evaluated using the LIBS intensity ratio of the Pb emission line at 405.83 nm to the substrate (Teflon polymer) specific emission from CN molecular band at 388.29 nm, after baselining all spectra. A comparison of the emission line intensities from Pb adsorbed on plastic samples (PE and PP) after finite time intervals demonstrated adsorption onto the sample surface. It should be noted that this experiment was carried out on samples of “macro-dimension”, but an even greater adsorption is to be expected on micro-polymers due to their higher surface area to volume ratio. Based on the LIBS analysis and a systematic study of surface contaminants, micro-polymers were contaminated with heavy metals like Al, Cr, Cu, Mn, Ni, and Zn, while other trace elements, such as Ca, Mg Li and Na, were also adsorbed to the micro-polymer surface. The authors compared the results for polymer type characterisation and surface heavy metal detection for pretreated and non-treated micro-polymers. Based on their results, the method can be applied on the sample directly after the collection process, eventually making the strategy applicable in situ.

3.5.4. Other applications. Elemental analysis of polymer samples using atomic spectrometric techniques is not only valuable for waste treatment and recycling, and for assessing the toxicological effects of micro- and nano-polymers, but is also crucial for other important applications. For instance, it is used to assess the migration of hazardous heavy metals in food-contact plastic packaging, and for fingerprinting to identify sources, particularly in forensic analysis.

As an example of the first, Carneado et al. investigated the factors affecting the migration of Sb from PET to beverages, considering the type of plastic, kinetic factors (time and temperature) and drink matrix (water and juices).122 Total Sb and Sb speciation – SbIII and SbV – were accomplished using bulk and liquid chromatography (LC) ICP-MS analysis, respectively. The HPLC system, equipped with an anion-exchange column (125 × 2.1 mm, 10 μm particle size), was operated at room temperature using 10 mM EDTA at pH 4.0 with 0.5% MeOH as the mobile phase. The maximum level of Sb established by the EU in drinking water (5.0 μg L−1) was not exceeded in any sample, irrespective of the juice flavor. After development and optimisation of the LC-ICP-MS method for the analysis of juice samples, the majority of the Sb was found to be in the form of SbIII, although this content was significantly lower than the total Sb content. This was attributed to the small amounts of Sb species found in the measurement extracts, which were close to the LOQ. As to migration experiments, both the temperature and time had an effect on the concentrations of Sb and Sb species, with the temperature being the main factor that accelerates Sb migration into fruit juice. This was attributed to the increased degradation of the PET with increasing temperature. Moreover, Sb migration also appeared to be dependent on the matrix, as it was found to be more pronounced in pineapple compared with other juice flavors. To study the extent to which Sb migration was governed by matrix and PET characteristics, a cross-migration experiment was carried out. For this purpose, five different PET containers were filled with all studied matrices (drinking waters, peach, and pineapple juices). The bottles were stored at 60 °C for 30 days. Total Sb and Sb speciation were determined in triplicate after 7, 15 and 30 days of storage. The results indicated that Sb migration strongly depended on both the type of PET container and the sample matrix. The strategy used in this work to assess Sb migration proved to be a useful approach for assessing the potential role of the matrix and the PET type simultaneously. The key role of the matrix composition should be addressed in future updates to the present European legislation describing migration assays using food simulants. It should be noted that when complexing agents are present in the food matrices, simulants may underestimate the migrated amounts.

As to the second application type, plastic materials are often found on crime scenes and are of importance in forensic investigations. Determining the trace elemental composition may yield discrimination between samples from different sources. De Bruin-Hoegée et al. introduced a new standard for quantitative elemental profiling of polymers by LA-ICP-TOF-MS.123 It should be noted that quantitative LA-ICP-MS analysis is well-established for forensic glass analysis, but the lack of appropriate, sufficiently reproducible and homogeneous reference standards hampers accurate quantification of trace elements in polymers. A novel approach was developed for producing PE, PS and PVC standards with known elemental concentrations (23 elements at three concentration levels). All chemicals were dissolved in various organic solvents (THF, ACN, MeOH or DCM) and dried after mixing. The preparation method was optimised by evaluating the effect of mixing time, drying temperature, and polymer type on the homogeneity of the standard. The solid standards were analyzed by LA-ICP-TOF-MS, FTIR, and XRF. Fragments of the newly created standards were measured using a 193 nm excimer-based LA system coupled with an ICP-TOF-MS instrument. Collision cell technology mode was applied, using 4% H2 in helium as collision gas, to improve signal-to-noise ratio and remove polyatomic interferences. All samples were scanned for the presence of 52 elements. Additionally, the concentrations of 54 elements were determined using XRF, while FTIR was used for characterising the polymer type of forensically relevant objects. To evaluate the homogeneity of the new standards, line ablation and multiple spot ablation measurements were performed. The average RSD values were found to be 10% but decreased to 6% after signal correction by selecting one element as the internal standard. The novel standard showed an enhanced homogeneity compared to existing polymer reference standards, with RSD values ranging from 15% to 167%. It should be noted that the NIST 612 glass standard still had better RSD values (<5%), although the non-matching matrix affects the elemental profile. The reproducibility was also assessed by measuring standards produced from separate stock solution, yielding an average RSD of 17%. To characterise the standards using at least two independent techniques, XRF analyses were conducted. The average RSDs were only slightly higher than those obtained using LA-ICP-MS and still below 25%. The potential of the new standards was evaluated by analyzing a set of polymer objects with forensic relevance. According to the authors, future research should focus on verification of the standards by multiple laboratories with various complementary techniques. Additionally, a larger dataset of forensics-relevant plastic should be constructed to enable the application of more sophisticated chemometric and machine-learning methods for discrimination between different classes, thus providing intelligence information on the type of material encountered. However, it should be noted that, although the new standards showed excellent performance, the poor elemental homogeneity of polymer objects encountered in forensic casework could hamper practical application.

Continuing in the context of forensic discrimination, Komatsu et al. characterised the metal components in white-based polyester single fibres using X-ray fluorescence spectrometry and X-ray absorption fine structure analysis using synchrotron radiation.124 Polyesters contain metal compounds derived from polymerisation catalysts, transesterification catalysts, and matting agents. Semi-microbeam XRF spot analysis refers to spot analysis under a 20 μm-diameter semi-microbeam. Monochromatic 14.5 keV X-rays were focused on 20 μm × 20 μm (V × H) with Kirkpatrick–Baez (KB) mirrors. A silicon drift detector was placed at 90° to the incident X-ray beam. Nanobeam XRF imaging of the slides and cross-sections of the polyester fibres was also accomplished. The 20 keV monochromatic X-rays were focused to a 100 nm × 100 nm (V × H) with KB mirrors. The fibre sides were imaged by on-the-fly measurement under 5 μm stage movement conditions in 0.1 s. The XAFS measurements were performed using the same measurement system as semi-microbeam XRF spot analysis. The energy of the XAFS measurements ranged from −30 eV to +75 eV of the absorption edge of the target element. Semi-microbeam XRF spot analysis of the white polyester single fibres detected Br, Ca, Cl, Co, Fe, Ge, Mn, S, Sb, Ti and Zn. Background noise levels were extremely low in synchrotron radiation experiments because the incident X-rays were monochromatised with a double-crystal monochromator. Pattern classification was performed based on the presence or absence of detected elements without considering differences in X-ray intensities. Some of the elemental patterns were presented by multiple samples. To identify the samples exhibiting those patterns, the X-ray intensity ratios were subjected to pairwise comparison. This analysis distinguished 429 pairs out of 435 combinations, indicating a discrimination capability of 98.6%. Nanobeam XRF imaging clarified the distribution states of the manufactured components within the single fibres. Analysis using XAFS was performed on the Co, Ge, Mn, and Ti detected by semi-microbeam XRF spot analysis. The XAFS could estimate the types of metal catalysts and matting agents in different single fibres. In summary, this study presented a successful single-fibre identification using semi-microbeam XRF spot analysis, nanobeam XRF imaging and XAFS. However, the use of synchrotron radiation for forensic purposes has the disadvantages of high costs and limited beamtime, thus limiting the applicability of this methodology for identifying trace evidence of extremely important crimes.

The last paper in this section focused on the development of a simultaneous LIBS/LA-ICP-MS method for the investigation of polymer degradation.125 Spatially resolved analysis of polymer thin films – polyimide, polystyrene and polyvinylpyrrolidone – can provide insights into aging and degradation in harsh environments, such as after exposure to a UV treatment and weathering in a corrosive atmosphere containing hydrogen sulfide. For this purpose, testing equipment to expose the materials to controlled environmental conditions and suitable analytical techniques for chemical characterisation needed to be available. Polymer films with a controlled thickness were prepared. Furthermore, an identical film preparation procedure was followed to prepare spiked polymers containing S in the range of 20–5000 μg g−1. These matrix-matched standards were used for calibration purposes in the context of quantitative LA-ICP-MS analysis. The simultaneous LIBS and LA-ICP-MS data were collected using a 193 nm ArF excimer laser equipped with a “TwoVol3” ablation chamber and two fibre mounts capable of collecting the light emitted from the laser-induced plasma. The system was coupled to a 5-channel Czerny–Turner spectrometer with CMOS detectors for broadband LIBS spectra acquisition (188–1097 nm) with a spectral resolution of 0.1 nm and a spectrometer equipped with a “PI-MAX-4” ICCD camera. The LIBS analysis followed by chemometric analysis of the data was used for identifying the different polymers through evaluation of the main elements C, H and O, as well as diatomic species C2 and CN. The LA-ICP-MS enabled the determination of the S uptake (quantitative depth profiles) diffusing into the polymers as H2S from the atmosphere. Significant differences in the oxygen signals were measured for the untreated versus UV-treated polymer samples. For all three polymers, the oxygen signal of the UV-treated films was shifted to higher signal intensities compared with non-UV-treated films, indicating oxidation. In contrast, the ICP-MS data for S indicated that the UV-treated polymers showed a significantly lower uptake compared with the non-UV-treated films. The developed method combined with the testing equipment for sample exposure to controlled environmental conditions can be a useful tool to investigate the properties and degradation of polymer materials and coatings.

4. Inorganic materials

This section of the update covers numerous topics including catalysts, building materials, electronic components, glasses, ceramics, nuclear materials and thin films/layered materials. Each will be dealt with in their own sub-sections. As usual, some topic areas are more research-active than others. However, in some cases even though the area is buoyant, there may be little atomic spectrometry. This is true certainly for the catalysts, thin films and electronic components. Often, the work is focussed on the actual catalyst or component with atomic spectrometry used almost as an afterthought to identify what material has been made or in trying to elucidate a reaction mechanism.

A primary iridium standard solution traceable back to the International System of Units was developed and reported by Gozzi et al.126 The starting material was ammonium hexachloroiridatehydrate. Gravimetric reduction of the salt to the metal under hydrogen yielded a material that could be related to the kg (an SI unit). A similar procedure was run on iridium metal as a comparison. The salt and the metal obtained after reduction was then acid dissolved and impurities measured using ICP-OES and ICP-MS. Inert gas diffusion enabled H, N and O to be determined. The trace metal and inert gas fusion enabled the “purity” to be obtained, leading to another part of traceability to be fulfilled. Solution standards were gravimetrically prepared from the candidate material salt. Using a high-precision ICP-OES method, these were compared with those prepared from the dissolved, unreduced high-purity iridium powder. Agreement between the sets of data (within 0.07%) with uncertainties calculated using error budget analysis confirmed the accuracy of the results.

Zhang et al.127 described the preparation and certification of four certified reference materials of elements in gold solution. Based on ISO 17034 and ISO guide 35, the materials comprised 23 analytes at concentrations of 0, 1, 5, 10 and 20 ng mL−1, each present in 2 mg mL−1 gold solution that had been prepared from the dissolution of the high-purity gold material GBW02793. Spike/recovery values were between 96 and 108%, indicating a good level of accuracy. A homogeneity study was also undertaken in which 11 replicate sub-samples were each analysed twice. Results indicated good homogeneity. A long-term stability study in which sub-samples were kept at room temperature for 12 months and a short-term study where samples were stored at either 60 °C or −20 °C for seven days were also undertaken. No discernible trends/losses were observed in either. The solutions were distributed to eight expert laboratories who analysed them with each obtaining results very close to the expected values. The materials were thought to be ideal for use as reference materials during the analysis of gold or gold jewellery.

4.1. Catalysts

As discussed previously, this is an immensely popular area of research, with over 600 papers being published within this review period. However, the large majority follow the same format of preparing and testing a new catalyst and analysing it to see if it is what they had anticipated. A large majority of papers simply use XRF or ICP-OES to find the composition of the material and perhaps use other techniques, e.g., BET, particle size measurements and XRD, to further characterise it. Some may use XPS or XANES to try and elucidate reaction mechanisms. Given that most of the papers are so similar in structure, finding papers that have any novelty regarding the atomic spectroscopy can be troublesome. However, the selection of papers below does show some novelty and are therefore worthy of some discussion.

Many of the more novel applications have used operando or on-line analysis. Such applications clearly have the advantage of not having to take samples half way through the process and then analysing off-line. If the process is found not to be working then off-line measurements means that time will be lost before it can be rectified. An on-line method is also ideal for determining the reaction or dissolution kinetics. Analysis of oxygen reduction catalysts was investigated by Schroder et al.,128 who used operando X-ray absorption near-edge spectroscopy (XANES) to study the dynamics of a Ag-MnOx oxygen reduction catalyst. The Mn valence changes in an ultrathin, porous MnOx layer on a silver thin film was tracked using the operando XANES. The Mn-K-edge measurements as a function of electrochemical environment and oxygen reduction reaction (ORR) conditions reveal that, when driving the ORR at 0.8 V-RHE, the Mn is more reduced and the MnOx redox is non-reversible in contrast to measurements in N2 saturated electrolyte. Ex situ analysis using ICP-OES and XPS indicated that these phenomena did not correlate to metal dissolution. Instead, it may be associated with morphological surface reconstruction related to silver valency. It was concluded that both operando and ex situ work was required to describe the dynamic property changes of electrocatalysts. This was because the Mn oxidation state and final morphology were affected by catalysis, the applied potential and the voltage history.

At present, iridium oxide catalysts provide the best compromise between activity and stability when used for the oxygen evolution reaction. However, iridium is an expensive material, so work has been undertaken to find alternatives or additional substances that mean that less of it is required. A study by Lahn et al.129 reduced the iridium content and increased the more inert titanium (to between 1 and 6%). To counter the potential drop in activity, they also introduced the reactive ruthenium. These new materials were made into thin films and placed under oxygen evolution reaction conditions. The stability of the materials was assessed using both on-line and off-line ICP-MS. For the on-line ICP-MS analyses, a scanning flow cell (SFC) was connected to the instrument to investigate catalytic activity and stability of different alloy compositions simultaneously. The electrolyte was saturated with Ar and pumped at a rate of about 205 μL min−1 from the cell. It was mixed on-line with an internal standard prior to reaching the ICP-MS instrument. For longer time-based experiments, 1 mL aliquots of the electrolyte were taken at intervals, diluted by a factor of five and then analysed. The studies showed that an addition of Ti at a concentration of 5% increased the stability of the catalyst by a factor of three and had the additional bonus of slightly enhancing activity when compared with the Ti-free equivalent Ru–Ir catalyst.

Another paper to report the use of a scanning flow cell coupled with an ICP-MS instrument was presented by Linge et al.130 These authors stated that silver-based electrocatalysts are promising candidates to catalyse the sluggish oxygen reduction reaction in anion-exchange membrane fuel cells and oxygen evolution reaction in unitised regenerative fuel cells. Silver nanoparticles were supported on Vulcan carbon or on mesoporous carbon, giving two sample types to be analysed. They used SFC-ICP-MS to estimate the kinetic dissolution stability window of Ag. The SFC and its operation were described in the paper. Again, sample was mixed on-line with an internal standard (Rh) once it had left the cell and was on its way to the ICP-MS instrument. The stability of carbon-supported Ag catalysts depended strongly on the morphology of the Ag nanoparticles. This could be tuned depending on the chosen carbon support and synthesis method. It was noted that there was very significant Ag dissolution once the potential was above 1.2 V versus the reversible hydrogen electrode. This made these catalysts useless for the oxygen evolution reaction. However, it was also noted that the Ag supported on mesoporous carbon was up to three times more stable than that on Vulcan carbon. They could be useful materials for cathode materials for fuel cells.

A third example of the use of a scanning flow cell was presented by Kormanyos et al.131 These authors used such a setup to study the dynamics and stability of a catalyst, in this case bimetallic Pt–Ru (PtxRuy) thin-film electrocatalysts where the Ru content ranged between 0 and 100%. Also studied were three commercially available carbon-supported counterparts (50–67% Ru content). In common with the other papers of this type, the authors added an internal standard on-line between the cell and the ICP-MS instrument. As well as the dynamics being investigated using SFC-ICP-MS, the authors also used XPS and other techniques (SEM, etc.) to investigate the morphology of the films. Results demonstrated that significantly more Pt and Ru was leached from the carbon-supported materials than from the thin film. The activity change was different too. As the Ru leached from the carbon-supported material (a very rapid and large amount), huge drops in catalytic activity were observed. However, for the films, as the Ru slowly leached, the activity also slowly declined. Another paper by the research group was entitled “High-throughput exploration of activity and stability for identifying photoelectrochemical water splitting materials”. The materials studied were thin films of Fe–Ti–W–O. The results revealed a relationship between composition, photocurrent density and element-specific dissolution. These structure–activity–stability correlations were visualised using principal component analysis.

An alternative catalyst system for the oxygen evolution reaction is that which uses cobalt-based materials. Li et al.132 prepared three different materials (cobalt metal, acid-etched cobalt and Co3O4) by treating the same starting material in a different way. The materials were analysed using near ambient pressure (NAP)-XPS for the quasi-on-line work, XPS and TOF-SIMS. The experimental setup for each was explained in the paper. In particular, the on-line NAP-XPS instrument, which was attached to a glove box without using glue, was explained. Depth profile work was also performed before and after the catalytic process occurred. The techniques showed the quantitative compositions and structures of the three-dimensional active oxide/oxyhydroxide layer reconstructed near the surface. The results suggested that the hydroxyl species and oxygen vacancies coordinated to reversible Co2+ sites within the superficial layer are responsible for the catalytic activity. It was concluded that the study provided valuable insights into the near-surface reactions of the model catalysts.

The use of nanoparticles as catalysts is nothing new. Such materials tend to be very successful because their size gives them a very high surface area and hence, activity. A very popular area of research with nanoparticles recently has been single-particle analysis-ICP-MS (sp-ICP-MS). This has the potential to give a number concentration of the particles present, but also a size distribution. Despite its common usage for nanoparticles, it is far less common to see it applied to the analysis of nano-catalysts. This was the subject of a paper presented by Martinez-Mora et al.133 These authors used the technique to study platinum nanoclusters made by gas-diffusion electrocrystallisation as electrocatalysts for methanol oxidation. Different quantities of polyvinyl pyrrolidone were used to stabilise the nanoparticles prepared and it was this that played a crucial role in preventing diffusion limited aggregation of the Pt nanoclusters. The PVP-stabilised Pt nanoclusters demonstrated superior electrocatalytic activity for methanol oxidation compared with aggregated Pt nanoclusters and commercial Pt/C materials. This was attributed to their semi-porous nature. As well as sp-ICP-MS, the materials were examined using SEM and TEM. The methods combined to indicate that the nanoclusters were typically 30–60 nm in diameter and comprised primary nanoparticles that were much smaller in size (2–4 nm).

Another paper by the same research group tackled a completely different area of catalyst analysis. Rua-Ibarz et al.134 reported a comparison of calibration strategies for quantitative LA-ICP-MS analysis of fused catalyst samples. Calibration of LA-ICP-MS has long been problematic. The accepted methodology usually requires calibration standards as closely matched to the samples as possible so that the laser ablates the same amount of material in both samples and standards. This paper adopted the approach of calibration against certified reference materials (CRMs) combined with internal standardisation. This was regarded as being the reference approach and led to errors of <15%. After fusion of the samples into a glass disk, two other calibration strategies were also tested: the “dry calibration” and the “wet calibration”. In the former, one certified sample was used for calibration. Laser spots of varying diameter (10, 12, 15 and 20 μm) and at different repetition rates (20, 30, 40 and 50 Hz) were used, giving a multi-signal calibration strategy. The “wet calibration” required a small adjustment to the hardware (and re-optimisation of the operating conditions) but required no solid certified material. Instead, it introduced a liquid standard simultaneously with the laser ablated plume of material hence employing a standard additions approach. Both methods were tested and validated using assorted glass reference materials. In addition, two used catalysts were fused into glass disks and analysed using the two LA-ICP-MS calibration strategies and the results obtained compared with those obtained using WDXRF. In both cases, errors were between −9 and + 7%, which was an improvement on the 15% obtained using the accepted methodology.

Two papers have reported the analysis of automobile catalysts. In one by Strekopytov et al.,135 a candidate CRM (LGC 3101) was analysed for its Pd, Pt and Rh content. Isotope dilution analysis was employed for the determination of the Pd and Pt whereas a normal external calibration with an internal standard (Ru) was used to determine the monoisotopic Rh. In all cases, detection was achieved using ICP-TOF-MS and sample preparation involved microwave-assisted acid dissolution. The use of a desolvating inlet interface and He as the collision gas enabled spectral interferences arising from oxides on 105Pd, 195Pt and 196Pt to be removed. The method was validated using used auto catalyst, monolith NIST SRM2557 and unused automobile catalyst ERM-EB503a. The exact same system setup was also employed on a sequential quadrupole ICP-MS instrument and the results obtained from the two systems compared. The expanded relative uncertainties for both Pd and Pt (2.7% and 2.3%, respectively) obtained using the quadrupole instrument were higher than that expected for ID analysis. However, the ID-ICP-TOFMS achieved much lower uncertainties, 1.0% and 0.4%, respectively. Having satisfied themselves that their methodology worked, the authors participated in an inter-laboratory comparison exercise with results being in good agreement with those from other laboratories. The other paper, by Zambrzycka-Szelewa and Godlewska-Zylkiewicz,136 employed an open extraction using aqua regia followed by determination of Pd and Rh using HR-CS ETAAS. The extraction conditions were optimised to ensure maximum extraction efficiency in the shortest time period. Different sample-to-acid ratios (1[thin space (1/6-em)]:[thin space (1/6-em)]40), temperatures (108 °C) and extraction times (two hours) were the variables optimised. Alternative, less sensitive lines of Pd (360.955 nm) and Rh (361.251 nm) were used because of the high concentrations found in the samples. Precision was typically 10% for Pd and 7% for Rh. Method validation was achieved through the successful analysis of ERM-EB504. The concentrations found (Rh: 297 ± 12 mg kg−1 and Pd: 270 ± 5 mg kg−1) were in reasonable agreement with the certified values (Rh: 338 ± 4 mg kg−1 and Pd: 279 ± 6 mg kg−1). Strangely, when a microwave extraction was attempted, the extraction efficiency for Rh was less than 40%. The ability of this instrument to detect the analytes simultaneously is clearly an advantage over older-style instruments.

Anatase titanium dioxide with assorted dopants is a well-known photocatalyst. Optical changes caused by bulk doping are not always represented at the surface where the catalysis occurs. Methodologies that are sensitive to bulk and surface modifications are therefore required for a complete understanding of doping mechanism and its effect on photocatalysis. A paper by Wach et al.137 used X-ray absorption spectroscopy (XAS) with different probing depths to elucidate the chemical composition of both the bulk and the surface. Titanium dioxide films were prepared containing both N and Cu dopants. Soft XAS analysis of the surface was carried out using N K-edge and Cu L-2,3-edge. The results showed that Cu2+ was predominant in the bulk but Cu+ was the dominant species in the surface active catalyst. The XAS analysis also confirmed the dominant process in Cu doping was through a substitutional mechanism. The nitrogen doping was interstitial, causing minor changes to the band structure but affecting the photoluminescence. The authors concluded that the study closed a knowledge gap and that atoms' disposition in the bulk sample affects electronic properties and impacts the catalytic cycle on the surface.

4.2. Building materials

Quality control of data is one of the most important aspects of a reliable analysis. The development of certified reference materials and conducting inter-laboratory comparisons ensures that the results from one laboratory conform to the accepted values obtained by other laboratories. An inter-laboratory study that resulted in a certified reference material was reported by Kazlagic et al.138 Six materials, four of which were cements, one slate and one limestone, were sent to 13 laboratories with instructions to measure 87Sr/86Sr ratios. At least two digestions per reference material with a minimum sample amount of 100 mg per digestion were requested. Participants were asked to perform a complete digestion of the reference material. The Sr needed to be separated from the matrix preferably via chromatographic means to remove interfering Rb and matrix elements with participants reporting Sr recovery and procedural Sr blanks. The 87Sr/86Sr isotope ratio measurements then had to be determined using either MC-ICP-MS or MC-TIMS. Additionally, one measurement of the certified isotope reference material NIST SRM 987 (Sr carbonate isotopic standard) was requested per MC-ICP-MS sequence or MC-TIMS turret, with at least three measurements in total. Participants were also allowed to use any other quality control measure they wished. The paper discussed the statistical approach used to collate and compare the data. No significant differences were observed between data sets obtained using TIMS and MC-ICP-MS. The Sr ratios varied slightly between samples, with three of the four cements (VDZ 100a, (Portland), VDZ200a (Portland composite) and VDZ 300a (blast furnace cement)), having a 87Sr/86Sr of 0.708–0.7093. The other cement (IAG OPC-1 another Portland cement) had a ratio of just over 0.726. The study was a success and the materials are now available commercially.

As always with cement-based materials, the majority of analysis methods used has been those that are capable of analysing the sample directly in the solid form, i.e., XRF or LIBS. One of the most frequently researched topics is the determination of Cl within the cement as it is one of the main problems of rebar-pitting and corrosion, hence weakening the cement or concrete structure. Two papers have reported the LIBS determination of Cl in cement but they took very different approaches. Javier Fernandez-Menendez et al.139 discussed how Cl is normally detected as CaCl, which emits light at 593 nm. This means that it runs the risk of interference from atomic sodium or from calcium oxide (which emits over the range 590–620 nm). These authors used a noble gas-enriched atmosphere to remove both types of interference. Both argon and helium were tested as inert atmospheres, with argon proving to be better at sodium interference removal while also increasing the sensitivity of the analysis. As well as removing the interferences, the inert gas method also overcame problems associated with the absence of a blank cement. The methodology was applied to real cement samples that had a Cl content of between 0.23 and 1.5%. The other paper was by Qiu et al.,140 who used collinear dual-pulse LIBS to determine Cl in cement pastes. These authors also used fast photography and shadowgraphs (an optical method that reveals non-uniformities in media like air, water or glass, in a similar manner to Schlieren imaging). When combined, these methods yielded crucial information. Instead of the second pulse simply re-heating the plume produced by the first, it was found that it re-ablated the sample and that a second shockwave turned the plasma from being a spherical shape to one that looked like an umbrella. Using the dual-pulse system with an inter-pulse delay of 200 ns and measuring the Cl signal at 837.59 nm, the LOD was 1930 mg kg−1, which was a significant improvement on the 2550 mg kg−1 measured using single-pulse LIBS.

A LIBS analysis combined with a statistical interrogation of the analytical data was used by Cai et al.141 for the rapid, on-line analysis of cement and some of its components. After acquisition of the spectra, Savitzky–Golay smoothing was used to de-noise them prior to them being input to sample set partitioning based on joint XY distance. The de-lineated data were then inserted to back propagation neural network for modelling and classification. Data from this model were compared with those obtained using the “hold-out” method using just data from CaO, SiO2, Al2O3 and Fe2O3. This is a statistical technique commonly used in data analysis and machine learning to assess the performance and generalizability of predictive models. The correlation coefficients improved by 26%, 10%, 17% and 4% for CaO, SiO2, Al2O3 and Fe2O3, respectively, and the root mean square errors (RMSE) decreased by 47%, 33%, 43% and 21%, respectively. It was concluded that the results showed a significant improvement in the accuracy of the data obtained from quantitative analysis of cement raw meal and that this was significant for real-time detection.

The presence of matrix effects during the LIBS analysis of cements was addressed by Zhang et al.,142 who employed chemometric treatment of the data. The method combined hierarchical clustering (HC), in which the samples with a similar matrix are grouped together followed by data insertion to assorted regression models. The regression algorithms used were: PLSR, support vector regression and kernel-based extreme learning machine. A total of 58 cement samples were analysed for six analytes (Al, Ca, Fe, K, Mg and Si). Results from HC-PLSR, HC support vector regression and HC kernel-based extreme learning machine all offered significant improvements in accuracy in terms of root mean square error of prediction (RMSEP) and mean absolute percentage error of prediction (MAPEP). An improvement in precision (relative standard deviation of prediction) was also noted. It was concluded that the method offered significant improvements in performance because of the reduced impact of matrix effects.

Several papers have also used XRF for the analysis of building materials. The effects of particle size on the accuracy and precision of measurements of building materials was reported by Mijatovic et al.143 Eight samples (four refractory concretes, a thermal insulation refractory concrete, a high-alumina refractory putty and two refractory bricks) were used in the study. Each of the eight samples was ground in a vibrating disk mill for 0, 30, 60, 120, 180, 240, 300 and 360 s, giving a total of 64 samples for analysis. Samples were then made into pressed pellets for analysis. Ten certified reference materials were used for calibration. Both univariate (simple concentration against signal) and multivariate (PLSR) analysis were used for the calibrations. For the PLSR, EDXRF data were employed as a model, while certified reference materials and ICP-OES (following a fusion sample preparation method) results were used as responses in the partial least-squares regression analysis. The grinding times affected the z-scores of Al2O3, CaO, Fe2O3, K2O, MgO, Na2O, P2O5, SiO2 and TiO2. The ideal milling times were sample-dependent but were typically towards the higher end of the scale, i.e., 180–360 s. This would indicate that as small a particle size as possible is required to minimise particle size effects, which is not entirely surprising.

The use of recycled concrete aggregates instead of naturally occurring ones would be environmentally friendly. However, the variable composition of the recycled aggregates (different chemical composition, different residual mortar content, etc.) and its potential to contain contaminants could impact on the quality of the material it is used to make. A rapid, inexpensive and simple method of analysing these recycled aggregates would therefore be very beneficial to the industry. Dey et al.144 used a hand-held XRF instrument to obtain chemical data and to identify contaminants. The results were correlated with those from standard thermal shock methods and from the chemical analysis of whole rock. It was concluded that the portable XRF methodology, when paired with locally calibrated reference samples, would be appropriate for use in either the laboratory or in the field.

An on-line XRF analyser for determining P2O5content in phosphate slurry was described by Ben Amar et al.145 The instrument could be used in two configurations: horizontal (high) flow and vertical (low) flow. The arrangements were described with the aid of pictures and diagrams in the paper. Certified materials were run so that calibration curves could be constructed for both configurations. Concentration ranges of 13.5–18.5% were obtained for horizontal flow. The corresponding figures of merit for vertical flow were 14.0–15.6%. Operating conditions were optimised for the horizontal flow setup, with optimal excitation energy being 20 or 25 kV, an excitation current of 600 A, a distance of 18 mm between the sample and the detector, a measurement time of 60 s and the use of an aluminium filter between the X-ray tube and the measurement window. Despite the low energy of P X-rays and the matrix effects, it was concluded that XRF was still a promising technique for the measurements.

4.3. Other inorganic materials

Edible salts have been analysed using LIBS for some of their major components (Ca, K, Mg and S)146 and for Ca and Mg.147 Both papers employed a hydrophilicity-enhanced substrate. The first paper analysed five salts from around the world (Australia, Bolivia, France and two from South Korea). The samples were first diluted by a factor of 5000 and analysed using ICP-OES. For the LIBS analysis, a silicon wafer was used as a substrate. Onto the surface of the wafer, a laser was used to etch a series of trenches in a lattice shape. The trenches were estimated to be 10 μm deep, 46 μm wide and 250 μm apart. This lattice trench-work evidently increased the hydrophilicity of the substrate. Tape was placed around the square trench-work to prevent sample solution from spreading outside this area. Salt samples were then dissolved (15 g in 85 g of water) and 15 μL of the resulting solution pipetted onto the trench-work, forming a small salt pool. The sample was then dried in an oven at 50 °C for 30 minutes. The dried samples were still not homogeneous, but they were significantly more so than for the samples simply dropped onto untreated surfaces. The authors adopted the alternating laser ablation data sampling (ALADS) approach during the LIBS analysis. This improved the precision of the LIBS measurements significantly by extracting three precise measurements from a total of 9798 single-shot LIBS spectra. The ALADS could extract three measurements with almost the same intensities from a single sample. However, it could not correct the sample-to-sample variation of laser-ablation sampling efficiency. Correlation with ICP-OES data was therefore poor. However, by normalising the analyte signals from LIBS to the Na signal, a significantly improved precision and correlation was achieved. Using this technique, the LODs were reportedly 14, 0.64, 1.7 and 530 mg kg−1 for Ca, K, Mg and S, respectively. The second paper147 was very similar but appeared not to use the ALAD approach, leading to somewhat higher LODs (87 and 45 mg kg−1 for Mg and Ca, respectively).

Two papers have reported the analysis of fertilizers. In one, by Nisar et al.,148 LIBS was used to determine the trace elements in well-known commercial fertilizers such as nitrophos, diammonium phosphate and single super phosphate. The trace elements Al, Ca, Fe, K, Ni and P were identified in the spectra obtained through comparison of wavelengths with the NIST database. Both quantitative and semi-quantitative analysis was undertaken. The lack of sample preparation was highlighted as one of the main advantages of the technique. The other paper to report the analysis of fertilizers was presented by Degryse et al.149 Zinc is often present in phosphate-based fertilizers but can remain stubbornly unavailable biologically. This study determined whether acid treatment post-granulation of the fertilizer could improve its availability. Monoammonium phosphate was fortified with Zn and then some of it was coated with sulfuric acid. For these coated granules, the Zn was either co-granulated or dissolved in the acid coating. Analysis of the granules both before and after a week of incubation in different soil types was then undertaken using XRF and XANES. The soil was also analysed to determine whether the Zn content had increased. Results showed that the Zn present in the uncoated granules was mainly zinc phosphate whereas in the treated granules it was present as mainly zinc sulfate (80%) if acid coated. If co-granulated, 45% of the Zn was present as the sulfate. After incubation in soil, the amount of Zn remaining in the granules was between 10 and 86% and was present mainly as insoluble salts. The Zn in the soil was present as the phosphate, irrespective of the treatment. It was noted though that the Zn from the sulfuric acid coated granules migrated further into the soil. It was concluded that post-granulation acid coating the fertilizer granules is effective in enhancing phyto-availability of Zn.

The rest of the applications applicable to this section involved the analysis of a mixture of sample types. In one, the Si[thin space (1/6-em)]:[thin space (1/6-em)]Al ratio in molecular sieves was determined using a novel “pouring and curing” sample preparation technique prior to analysis using portable XRF.150 A molecular sieve sample (6.825 g) and 3.5 g metal phase glue were placed into an agate mortar, ground for 5 min, poured into the mould and 2.5 g metal phase glue curing agent added with stirring for 30 s. The mould was placed on a heating plate at 55 °C and cured for 15 min. On cooling the sample was removed from the mould ready for XRF analysis. A calibration graph was constructed by making standards that were mixtures of aluminium oxide and silica that had been prepared in the same way. The binary ratio method was used to prepare calibration graphs. This is based on the net intensity of the analytical line of the two elements in the standard establishing a double logarithmic calibration curve of the analytical ratio of the two elements and the corresponding concentration. The same cured sample was analysed 16 times on different spots so that homogeneity could be assessed. The precision values for Al and Si were 3.12% and 3.64%, respectively, which was sufficiently acceptable to claim that homogeneity was good. Results for Al and Si obtained using the proposed procedure were compared with those obtained using ICP-OES. Close agreement between the two techniques was obtained, with the values for ratios agreeing to within 96.732–106.777%. The method was relatively quick, produced reliable data and did not require the expense and time required for the preparation of a fused glass disk.

The analysis of e-cigarette liquids was discussed by Faria et al.151 who used TXRF as the analytical tool. A total of 38 samples of different brands and containing different nicotine levels were analysed along with some raw ingredients such as a vegetable glycerine, a propylene glycol sample and 15 liquid flavours. Samples were mixed with gallium solution to give a final Ga concentration of 200 μg L−1, which was used as an internal standard, and then 10 μL was placed on a quartz disk. This was then dried at 100 °C for 24 hours prior to TXRF analysis. Analytical results were then input to PCA for data reduction and classification. The reference material NIST 1640 was used as a quality control measure. Of the 16 analytes determined, 11 were used for further investigation. The toxicity was assessed in comparison with drinking water limits and of the 38 samples analysed, 10 were in excess for at least one element. Interestingly, the PCA analysis managed to identify those 10 samples that exceeded guideline values.

Amosite is a form of asbestos that is known to be toxic. Instead of simply determining its metallic components which, although useful, tells us nothing of its toxicity, Pacella et al.152 determined the metals and their amounts that leached into Gamble's solution (a simulated lung fluid). This would give a far better assessment of the toxicity. An XPS study of the particles' surface before and after leaching was also undertaken. Samples (20 mg) were suspended in Gamble's solution at 37 °C for a month with gentle agitation. Aliquots were taken periodically, filtered and then analysed using ICP-OES. The concentration of Ca and Mg leached into solution increased with time whereas Fe was not observed. The XPS surface study showed a decrease in Ca and Mg as well as Si and Na on the surface of the particles. A concomitant increase in O on the surface was also detected. A complex Fe chemistry was observed. The Fe on the surface was rapidly oxidised and Fe from the bulk was promoted to the surface. The authors thought that this methodology could help in the understanding of the toxicity of this material.

4.4. Ceramics and refractories

This has been a relatively quiet area of research during this review period. The papers that have been published have nearly all used analytical techniques that are capable of analysing the samples directly, i.e., without any sample preparation. This is because of the refractory and inert nature of ceramics, making them difficult to bring into solution for a standard ICP-OES or ICP-MS analysis. Techniques such as XRF, LA-ICP-MS or LIBS have therefore dominated.

Confocal micro X-ray fluorescence (CM-XRF) analysis has, hitherto, been thought of as a useful tool in archaeological studies, the analysis of paintings or for forensic applications. This is because it can be used to analyse very small items, enables depth-resolved analyses to be undertaken and is non-destructive. A paper presented by Mori et al.153 discussed its use for differentiating two slightly different ceramic samples. The paper described the components of the lab-built instrument and then went on to describe the analysis and results. A different Fe and Mn profile was observed for the blue paintwork on the two samples. In addition, its ability to determine analytes at different depths enabled the ceramic materials to be differentiated because the Co and Zn depth profiles were different between samples. Although not an earth-shattering discovery, it was a nice addition to the applications it is capable of fulfilling.

Gazulla et al.154 developed a method by which the analytes As, Cd Co, Cr, Hg, Ni, Pb, Sb and Se were determined in titanium dioxide at low-concentration using WDXRF. To prevent the excessive dilution that occurs during the preparation of samples as fused disks, the authors used pressed pellets. They were careful to use plastic-ware when handling the sample and binder to reduce the possibility of contamination by Cr from stainless steel spatulas. Several materials were tested as binders, including mannitol, stearic acid, N,N-bis-stearyl ethylene-diamine, a couple of commercial binders and n-butyl methacrylate, with the last of those proving optimal. Unfortunately, no reference materials could be identified and so the authors resorted to making standards from pure oxides of each element mixed with high-purity titanium dioxide. Using this pseudo-spike/recovery methodology, recoveries in the range 94–106% were obtained. Method validation was also achieved through analysis of titanium dioxide samples using the alternative technique of ICP-OES following a microwave-assisted acid digestion using nitric, hydrochloric and hydrofluoric acids. Statistical analysis indicated that there was no significant difference between the ICP-OES and WDXRF data sets. The advantage that the methodology offered was that it was significantly quicker than other methods and required less of a sample dilution.

Calibration for LA-ICP-MS is known to be troublesome and requires standards as closely matched to the samples as possible. Kobayashi et al.155 developed a method for analysing ceramic powders that utilised screen printing methodology. Sample powder was mixed with liquid resin, which was then made into a film using a screen-printing process. Standards were prepared in the same manner but by adding a known concentration to the sample. A sort of standard additions was therefore set up. Internal standards were also added to both samples and spiked samples. The advantage of the methodology was that the samples and standards were exactly matrix-matched. The methodology was applied to the analysis of titanium dioxide powder although the authors envisaged that it could be used for many sample types.

The analysis of the surface of ceria-based ceramics was explored by Zagaynov and Buryak.156 The dopants added and their concentrations can have a serious effect on the quality and functioning ability of the material and so methods that are both rapid and accurate are required for the analysis. Numerous analytical techniques were compared including HRTEM, XPS, ICP-MS, SEM-EDS and laser desorption ionisation (LDI)-TOF-MS. After extensive comparison, which showed that all of the techniques had certain advantages and disadvantages, it was concluded that the LDI-TOF-MS in both positive and negative mode was the ideal technique because no particular sample preparation step was required, hence speeding up the analysis. An added advantage was that the data produced were in good agreement with those obtained using XPS.

Proppants are spherical ceramic beads or sands that keep fractures deliberately formed during fracking open. They are therefore exposed to formation waters and can therefore potentially adsorb assorted contaminants, which may affect their durability and hence recovery of the shale gas from the reservoir. Hao et al.157 submerged samples of proppant in formation water and then let them soak at 80 °C for periods of three and six months. A blank sample was treated the same way but without the formation water. After the exposure period, the waters were analysed using AAS and the proppants analysed using XRF and XPS (as well as FTIR, SEM-EDS, XRD and TEM). The XPS was used to analyse the surface of the proppant, whereas the XRD was used to investigate the phase composition before and after the exposure. Results indicated that alumina is transformed to Al3+, which then reacts with carbonate or bicarbonate in the water, forming aluminium hydroxide, which then precipitated in the formation water. Similarly, the silicon in the proppant reacted with chloride, forming silicon tetrachloride, which immediately reacts with water to form silica gel on the surface of the proppant. The results showed conclusively that exposure to the formation water damaged the structure of the proppant, which would lead to a decrease in its performance. Having elucidated the aging mechanism, the authors proposed a method of stopping it.

Erbium oxide ceramic is a potential material for the self-cooling blanket of tokamak fusion reactors. Liquid lithium is extremely corrosive and so He et al.158 used time-resolved picosecond LIBS to investigate its effects on the ceramic. Ceramic sheets were placed in a stainless steel reaction vessel and then lithium sheets placed on top. The vessel was then sealed and placed in a furnace where the temperature was then raised to 400 °C and held there for 120 hours. The surface of the ceramic was then investigated using SEM-EDS and the depth profile analysis undertaken using LIBS. The crater depth measured using profilometry was directly proportional to the number of laser pulses. As depth increased, the Li signals at both 610.353 nm and 812.645 nm decreased, whereas the opposite was true for Er at 405.547 nm and 405.951 nm. Experiments where the LIBS analyses occurred at different pressures indicated that the depth resolution of the analysis of anti-corrosion erbium oxide ceramic materials can be controlled by the environmental gas pressure.

4.5. Glass

It should be noted that the analysis of archaeological or heritage glasses will be covered in that section of the review. The reader is therefore directed there for those analyses. This section covers the new methods of analysis for industrial glasses. The forensic analysis of glass will also appear in that section. An example was by Tran et al.,159 who determined 29 elements using PIXE or NAA in automobile window glasses. For more detail, the reader is directed to Section 5.1 of this update.

As usual, this is a relatively small section of the review. However, the studies that have been undertaken cover a quite wide range of subjects and use quite a few techniques. There have been examples from recycling and from numerous different types of glass.

There has been one overview of the analysis of glass published in this review period. The paper, by Le Losq et al.,160 was written in French and contained 63 references. The review was split into several sections covering: Introduction, Methods of chemical analysis, Optical and vibrational spectroscopies and X-ray absorption and scattering. Included in the Methods of chemical analysis section were sub-sections on XRF, ICP-MS and LA-ICP-MS, atomic absorption, ICP-OES and electron microscopy. Other sections included NMR, tomography atomic probe, Vickers Hardness, density, thermal properties of glasses including differential scanning calorimetry. As well as giving examples of applications, the review also covered some of the theoretical aspects.

The use of LIBS has continued for glass samples, although there are still numerous other techniques that are also used. One important research aspect using LIBS is that of recycling. A paper by Pontes et al.161 determined the effects of glass colour and particle size on the LIBS analysis of glass waste and differentiation using chemometric methods. In this paper, the rather unusual step of grinding glass bottles from the same manufacturer but of different colour was undertaken. Bottles of five different colours were smashed and ground into fractions that passed through mesh sizes of 48, 100, 250 and325 μm. These different coloured powdered glasses were then sintered to form coloured disks ready for analysis. The paper discussed how the LIBS analysis occurred and then gave a good description of the data handling. The analytical data (10 laser shots at each of 28 spots on a disk and for 10 disks per sample, giving a total of 2800 spectra per sample) were input to PCA to reduce the number of data points that the classification algorithm would use – hence saving the time and computer power required. The machine-learning-supervised algorithm of SVM was used along with the standard normal variate filter (a method to decrease the effects of noise) for the classification of the glasses. A spectral angle mapper algorithm was also used to examine the data. The PCA-SVM method proved to be highly successful in differentiating different colours, with a success rate of 99.04%. Differentiating between particle sizes was also fairly good, with a success rate of 96.75%. When differentiation of both particle size and colour was attempted simultaneously, the success rate was still an impressive 97.64%. The results were improved further when the spectral angle mapper was also used.

Gadoros et al.162 used LIBS for the analysis of soda-lime float glass. It is known that LIBS causes tiny craters to form on the surface of the samples. These craters can have a serious effect on the quality of data obtained using subsequent laser firings. This study determined the effects of irradiation parameters, such as wavelength, laser power, focussing conditions, etc., on the damage caused to the sample. A Nd:YAG laser with frequency doubler and tripler modules (1064 nm, 532 nm and 355 nm), was used with the energy of the pulses being 360 mJ, 180 mJ and 90 mJ respectively with the half width of 3.2 ns. The repetition rate was 20 Hz. The laser pulses were focused by a single quartz lens of 40.6 mm, 41.9 mm, and 43 mm focal length on the 3 wavelengths, respectively. The laser pulses were focused in front of the sample surface (d = +1 mm), directly onto the surface (d = 0 mm) and below the surface into the bulk of the sample (d = −1 mm). The craters and other damage were examined using optical microscopy and profilometry. All of the laser wavelengths and focusing conditions applied in the experiment proved to be suitable for LIBS analysis. However, some operating conditions caused significantly more damage than others (e.g. operating at 1064 nm and focussing below the surface). There were, however, considerable differences in spectral quality and sample degradation. The optimal settings could ensure a favourable distribution of the absorbed energy preventing cracks and fractures of the glass material. With an increasing number of shots, both the diameter and the depth of the ablation crater increased, but no damage apart from that occurred. The actual number of laser pulses determine the size of the crater, but micro-crack free ablation could be achieved by focusing the laser beam in front of the target surface. Overall, it was concluded that parameters could be set that minimise damage to the sample.

Glass has been analysed using XRF for decades. Several research papers continue to be presented though. Many of these have focussed on the XRF analysis of phosphate glasses. Two examples were presented by Cernosek et al.163,164 In the first of these papers, XRF along with ESR and 31P magic-angle spinning (MAS)-NMR were used to elucidate the composition and structure of phosphate glasses. A model was produced that incorporated the results from the three analytical techniques. The NMR results, in particular, led to some surprising discoveries. The structure of the glasses is far more complex than a simple chain of basic structural units. The authors admitted that they do not yet have a firm idea of the structure of the glasses but said that work was ongoing. The second paper, by Cernosek and Holubova,164 created glasses with general formula xCuO–(50 − x)MgO–50P2O5 (x = 0, 0.2, 10, 20, 30, 40, 45 and 50) using copper oxide, magnesium carbonate and phosphoric acid. The samples were again characterised by XRF, XRD, EPR Raman and 31P MAS-NMR. The model developed was verified using other techniques including dilatometry, Raman and tenziometry. The results of each of the techniques were discussed in detail in the paper. Shielding of the charge of the copper core by its d9 valence electrons was also verified. The results from these two papers seem to indicate that the structures of the glasses are not as simple as envisaged.

A third paper to report the analysis of phosphate glasses was presented by Sulowska et al.165 This paper reported the influence of S ions on the glass-forming ability and structure of silicate–phosphate glasses. Both XRF and XAS were employed in the study as well as other techniques such as 29Si and 31P MAS-NMR. The addition of S facilitated the formation of glasses with higher concentrations of P2O5. The XAS results indicated that the S was present in the reduced form of S2−. The NMR data found no evidence of P–S or Si–S linkages.

It is well-documented that LA as a means of sample introduction can be minimally damaging to the sample; if focussed on specific areas, it can analyse extremely small objects/inclusions and has several other benefits. However, it is also known to suffer problems such as difficulty with calibration, etc. A paper by Mervic et al.166 discussed the elemental fractionation effects (non-stoichiometric effects during vaporisation, transport of ablated particles, atomisation and ionisation in the plasma) exhibited within and between matrices. The use of an internal standard can correct for many potential errors, but not for a large difference in the amount of sample ablated. Such differences can be large if samples and standards of different hardness, density, boiling point, etc., are used. These authors adopted an approach that used signal normalisation against the volume of sample ablated. This was measured using profilometry of the crater left in the sample. Consequently, the concentrations were expressed not as μg kg−1, but as μg cm−3, although they could be inter-converted if the sample density is known. The approach taken was shown to work when a decorative glass with murrina was analysed. The method was validated through comparison of data with those obtained using SEM-EDS and through the analysis of NIST 612.

Li et al.167 discussed the use of nano-SIMS to try and analyse different types of glass for their Li isotopic content. This is normally problematic because of instrumental mass fractionation effects, the extent of which can vary from sample to sample. Therefore, empirical methods of correction are normally required. In this work, O at an intensity of 1 nA was used as the primary beam. Secondary ions of 6Li and 7Li as well as 30Si were detected simultaneously. A significant matrix effect was noted on the 7Li signal, with a mass fractionation of 19% being observed. Various correction schemes, some of which were univariate and others multivariate, were employed to try and identify the source and to correct this error. All of the results indicated that it was the silica that was causing the problem. By determining the 30Si signal simultaneously with the Li isotopes, the authors achieved normalisation with the error reducing to less than 3‰. This on-line correction was effective for numerous glass types, e.g., from ultramafic to highly siliceous.

Other applications have also been published. Included in this number was a paper by Shin et al.,168 which discussed the preparation and analysis of silver tungstate–tellurite glasses and their ability to immobilise radioactive iodine. Several glasses were prepared that contained differing concentrations of silver iodide (0, 10, 20, 30 and 40 mol%). The preparation was a melt-quenching process in which the components were heated together at 850 °C for 90 minutes. Once prepared, the glasses were analysed using XRF (which confirmed no loss of elements during preparation), XRD (which confirmed the samples were amorphous) and XPS. The XPS results demonstrated the effects of increasing the silver iodide on the glass matrix. The oxidation number and bonding state of W, Te, and O did not significantly change, even when the amount of AgI added increased from 0 to 40 mol%. However, the binding energy of I was influenced by the surrounding cations, such as W6+ and Te4+. Leaching tests on the materials demonstrated that the normalised release of all elements satisfied the US regulation of 2 g m−2.

A two-stage analysis of the major components of glasses of the Ga–Ge–Te–I system was reported by Evdokimov et al.169 The glasses could contain 5–15% Ga, 10–20% Ge, 69–75% Te and from 1–6% I. The two stages of the preparation were completely independent of each other. The first stage involved an acid dissolution followed by ICP-OES determination of Ga, Ge and Te. The preparation for the I determination required an alkaline extraction. Experimental details of the preparation procedures were presented in the paper. Accuracy was assessed by comparing the results of an analysis of model solutions and model glass samples with the calculated composition. The uncertainty of the results was reportedly 0.05–0.1% at the 95% confidence limit.

4.6. Nuclear materials

Measurement of materials being developed for fusion reactors remains a popular topic, with LIBS a dominant technique. Studies have looked at factors including improving the quantitative ability of LIBS, modifying instruments to work in high-radiation-dose environments and development of portable instruments. The measurement of isotopic ratios for nuclear forensics and safeguards is extensively studied, with uranium and actinides commonly studied using various mass spectrometric techniques. Within this area, streamlined sample introduction to reduce or remove the need for sample preparation to improve the speed of analysis and reduce secondary waste has been a focus of some studies. Safe and effective fission-reactor operation and decommissioning of nuclear sites continues to be investigated. This includes characterising materials used during reactor operation or for waste characterisation purposes post operation. There is also an ongoing drive to increase the number of radionuclides measurable as part of the decommissioning process, with mass spectrometry the dominant technique.
4.6.1. Fusion. The development of materials for nuclear fusion reactors remains a common area of research. Table 2 summarises some of the measurements of materials relevant to nuclear fusion. Of the analytical techniques used, LIBS is dominant. Trtica et al.188 prepared a review on LIBS hydrogen isotope detection for both nuclear and fusion technologies. The efficiency, reliability and speed of LIBS was highlighted, as was the importance of H isotope measurement, with adverse impacts on material integrity and lifetime mentioned as possible issues.
Table 2 Summary of measurement of materials relevant to nuclear fusion
Analytes and materials Analytical technique Comments Reference
Cr, various materials SIMS Quantifying elements of interest in fusion-relevant materials through fabrication of ion implanted internal standards 170
Li, Sn, liquid metal divertor walls Calibration-free LIBS Design of liquid metal divertor walls compared with solid counterparts, and the need to understand migration and redeposition of liquid metals such as Li, Pb and Sn during a testing campaign. 171
Oxidised Er ceramic LIBS Investigation of candidate material for self-cooling blanket system of fusion reactors. Analysis of static corrosion of Li on ceramic and exposure over different times and pressures. 172
Irradiated W components, W films, deuterium (D)-irradiated graphite Various Development of a surface science facility to enable experiments to study plasma–material interactions, with various modified techniques. 173
D supersaturated surface layers in tungsten High resolution, cross sectional TEM Investigation of properties of D supersaturated surface layers including thickness, internal microstructure and D retention as a function of incident ion energies. 174
Modelling Li plasma dynamics LIBS hybrid model using 3DLINE, FEOS and THERMOS codes Investigating the validity of the commonly used calibration-free Saha–Boltzmann technique for interpreting LIBS results, with implications for characterisation of Li-coated surfaces. 175
Multiple elements, graphite tile LIBS Primarily deposited impurity elements are Al, C, Ca, Cr, Cu, Fe, Mn, Ni and Si from various materials, e.g., a steel first wall. Surface homogeneity and depth distribution of each element quantified. 176
Multiple elements, shutter plate surface LIBS Primary impurities including Cu, Li, Mo, Na and W identified, with surface distribution investigated, as well as thickness of the deposited layer. Results supported by confocal microscopy. 177
Optical emission from Tokamak X-point plasma region Dual route OES (D-OES) Diagnostic tool to obtain wide-spectral-range spectra and high-wavelength-resolution line shapes. 178
Copper alloy (CuCrZr) Nd:YAG laser Controlled ablation for removal of impurities from the surface of plasma-facing components. 179
He, wall materials LIBS Supported by thermal desorption mass spectrometry for absolute concentration of He retention 180
Be-based samples LIBS and calibration-free LIBS Depth profile analysis focusing on impact of different seeding gases (N, Ne, He) on Be-based samples containing D. 181
Mo target with H impurity LIBS Assessing H retention in atmospheric N, He or Ar that can be used during maintenance breaks and broaden the spectral lines and reduce intensity of lines of H isotopes. 182
Al–Li alloy plasma LIBS Coaxial optical structure based on a linear array of optical fibres to improve understanding of spatio-temporal evolution of the plasma to improve accuracy of quantitative LIBS. 183
Laser-induced Mo plasma LIBS Determination of the magnetic field strength and polarity direction near the surface for improved analytical accuracy of LIBS. 184
Frozen-D2O–H2O mixture LIBS Investigation of spatio-temporal evolution of spectra in low-pressure background from laser-induced plasma for improved accuracy of LIBS measurement. 185
Impurities in plasma-facing components Portable LIBS Characterisation and testing of a portable LIBS instrument that can be integrated into a robotic arm for use in harsh environments. 186
WTa coating Resonant laser LIBS Fine tuning of an optical parametric oscillator to the resonant transition of W and analysis of optical emission spectra, for improved quantitative depth analysis. 187


Plasma-facing components were the most common materials investigated, and must be able to tolerate a range of temperature, pressure and radiation conditions. As well as measurement of solid materials, some investigated the impact of gases on the quality of LIBS spectra, whilst others looked at improved understanding of spatio-temporal variations and the magnetic field strength and polarity. Testing was achieved through the use of experimental fusion reactors and the development of bespoke testing facilities, such as that described by Schamis et al.,173 which included a low-energy high-flux broad beam ion source, liquid metal dropper, lithium injection system, RF sputter source and evaporator. Portable LIBS that could be mounted to a robotic arm for improved remote operation capability was also the subject of some studies.189 Techniques other than LIBS were used in some studies, including SIMS, TEM, dual route OES, or in the case of a study by Djigailo et al.,190 a combination of techniques for the composition of Li-containing films.

4.6.2. Nuclear forensics. Identifying the source of nuclear contamination through accurate and precise isotopic ratio measurements is a valuable capability for applications including studying post-fallout behaviour at Fukushima and longer term nuclear weapons fallout, or environmental behaviour of radionuclides following planned releases from nuclear sites. There is an additional nuclear safeguard aspect to prevent and, if necessary, identify the illegal gathering of nuclear materials.

Zhao et al.191 published a review paper on the progress in analysis of radioactive hot particles. Hot particles can contain multiple radionuclides including Am, Pu and U, which can be used to identify the age and origin of nuclear material that is equivalent to the analysis of bulk material. The review considered the techniques used for identifying, screening, locating, and performing isotopic analysis of hot particles, with a focus on the advantages and limitations of different mass spectrometric techniques. A separate study by Holland et al.192 reviewed the importance of nuclear archaeology, which can be used to characterise sites used in the early days of the atomic age that have since been abandoned, such as mines, spent processing and enrichment plants and waste dumps. The paper discusses the methodologies available with use of a case study, explaining how the current state of legacy sites and the quantification of the radioactive inventories can be achieved, supporting local communities and site owners.

The importance of reference materials and comparison exercises was identified in several studies. Lorincik et al.193 reported the characterisation of forensic materials (two powdered U and two metal U samples) that made up the 7th collaborative material exercise of the Nuclear Forensic International Technical Working Group. Several Czech laboratories were able to determine the isotopic composition of U in all samples using several analytical techniques, including single-particle SIMS. The study was able to determine the nature of the material, and that one of the materials had a different source to the other three. The same samples were measured in a separate study by Serban et al.194 using single-quadrupole ICP-MS, where the analytes were measured without chemical separation, and the age determination made through 230Th/234U measurement. LeBlanc et al.195 reported a certification campaign for three Uranium Ore Concentrate CRMs from the National Research Council Canada, obtaining results from 15 laboratories across 10 countries. Sector field and triple-quadrupole ICP-MS instruments were used to determine 64 trace element impurities, with discussion over factors including the importance of sample preparation and uncertainty evaluation. A study by Qiu and Cooks196 had a cautionary message regarding the isotopic abundances of commercial organic compounds and the deviations from assumed natural ratios. The focus was on 10B, used as a neutron absorber in control rods in nuclear reactors, with the deviation from natural abundance in commercial B sources confirmed by ICP-MS. The significance of sample size in particle analysis was the focus of work by Inglis et al.,197 which presented the fundamental statistical methods for interpreting particle data.

As mentioned in the examples above, a range of mass spectrometric techniques has been used in isotopic ratio measurements for nuclear forensics. Multiple techniques were used in the analysis of single micrometre-sized U particles combined with a micro-manipulating technique in a study by Park et al.198 For isotopic analysis of single particles, SIMS was determined to be more accurate and precise than SEM/TIMS and SEM/MC-ICP-MS. Fenclova et al.199 focused on the measurement of Pu isotopes and other actinides using AMS. The paper reviewed the techniques for actinide separation and the preparation of oxide and fluoride starting materials that are key in the suppression of isobaric interferences, material quantity and sensitivity.

Laser-ablation-based mass spectrometric techniques are popular for forensics and multiple other applications, with the advantage of direct measurement of solid materials without sample preparation. Kwapis et al.200 reviewed laser ablation plasmas for nuclear applications. The review included physical breakdown mechanisms and laser interactions, chemical and thermodynamic properties governing LA plasmas, as well as standoff detection of radioactive aerosols and fission gases linked to molten-salt reactor monitoring. Consideration is also given to future outlooks of LA plasma spectroscopy. Varga et al.201 combined LA with MC-ICP-MS for U age dating through the measurement of 230Th/234U ratios. Accurate isotopic measurements were achieved for enriched U samples, however higher mass resolutions were required to remove interferences for natural and low-enriched U samples. Susset et al.202 combined laser ablation with ICP-SF-MS for determining 48 elements in six uranium ore concentrate materials, with detection limits in the ng g−1 range. Conversion of starting materials into glass beads through alkali fusion followed by direct measurement avoided the use of chemical reagents, reduced waste production and the total procedural time. Zirakparvar et al.203 compared LA and solution-based MC-ICP-MS techniques for U isotope ratio measurements from certified reference materials and a nuclear fuel precursor material. The aim was to demonstrate that the latest generation MC-ICP-MS could produce precise and accurate isotopic ratio information, even at low signal intensities, using the all-Faraday cup method. The aim was also to investigate the different amplifier resistor levels and integration time impacts on precision and accuracy of different measurements. There were limitations to the all-Faraday cup approach but it was also shown to be a highly versatile technique.

The use of ICP-MS was common for isotopic ratio measurements, with U isotopes being the most frequently measured. Bradley et al.204 combined a liquid microextraction system with ICP-QMS for spatial U isotope measurement on cellulose-based swipe materials that are used by the IAEA for their environmental monitoring programme. Isotopic values of U particulates agreed with the reference values. The particulates were also successfully extracted after being placed directly within a clay matrix, suggesting the procedure developed could be applied to more complex matrices. Jaegler and Gourgiotis205 also investigated U isotope ratios, focusing on 236U/238U at ratios as low as 10−11 using ICP-MS/MS in reference materials and environmental samples previously characterised by AMS. The developments in ICP-MS/MS instrumentation were credited along with the use of O2 reaction cell gas for the on-line separation of 236U from tailing and polyatomic-based 235U interferences.

Shollenberger et al.206 measured Mo and W isotopic compositions in 16 low-enriched uranium fuel pellets using MC-ICP-MS. The high number of isotopes of each element makes Mo and W candidates for isotopically enriched taggants to materials at the front end of the nuclear fuel cycle. These taggants can be measured in case of the material subsequently being found outside of regulatory controls. However, these materials can undergo fractionation (for example during U enrichment) that may hinder their application. The variable isotopic ratios measured in different samples complicates the use of these elements and demonstrates that the isotopic composition of any potential taggant element must be well characterised. Wang et al.207 developed novel methods for isotopic analysis and dating of 15 uranium ore concentrates using MC-ICP-MS and ICP-MS/MS. As well as reporting the relative uncertainties in the ratios measured for different U and Th ratios, the results also showed the ability to distinguish between uranium ore concentrate samples by simultaneously determining the 235U/238U, 234U/238U and 230Th/234U ratios. Different approaches for age dating were proposed depending on the age of the uranium ore concentrates being studied, with the use of multiple parameters and analytical techniques successfully narrowing the range of potential origins for nuclear forensic purposes.

Techniques other than ICP-MS were also used for nuclear forensic applications. Sanyal and Dhara208 prepared a review of the applications of TXRF for characterisation of nuclear materials. The advantages with regards to low sample requirements and non-requirement of matrix-matched standards were outlined, along with trace-level applications for materials including fuel, coolant and control rods. The technique of TIMS was used for the determination of high-precision isotopic ratios. In one study, the 135Cs/137Cs ratio was assessed in debris samples from the first nuclear test, Trinity.209 The results showed significant fractionation from the predicted fission yields, all with relatively enriched 137Cs. The ratios also varied with relative distance from ground zero and heterogeneities between different lithologies within a single sample. The technique of TIMS was also used by Goswami et al.210 for the measurement of 238Pu/239Pu ratios as part of the recertification of isotopic standards prepared in the 1960s and 1970s. The aim was to lower the uncertainties of Pu isotopic compositions. A novel approach was developed that used the ratio of the isotope ratio of metal oxides to metal ions of U and application of a 233U spike in determining the 238Pu content. An external precision of 0.2% was achieved in the determination of the 238Pu/239Pu ratio for isotopic standard SRM 947. To overcome polyatomic interferences that can affect U isotope ratio measurements in some samples (e.g. from Pb), Goodwin et al.211 investigated liquid sampling atmospheric pressure glow discharge (LS-APGD) coupled with an Orbitrap mass spectrometer. This instrument is capable of high mass resolution (>70[thin space (1/6-em)]000 at m/z = 200) and can therefore replace relatively time-consuming offline chemical separation. In samples spiked with impurity concentrations 1000–5000× those of U, no interference was observed on either the 235U or 238U signal, whilst isotope ratio values were within two standard deviations of the values obtained without interference additions.

4.6.3. Reactor operation and decommissioning. The safe operation of nuclear reactors through measurement of potential fuels and reactor materials was the subject of several studies. Safe and cost-effective decommissioning is a high priority for a number of countries, and methods are continually being developed to expand the number of radionuclides measurable as well as the measurement uncertainties.

Thorium-based fuels are a promising option for nuclear power in some areas, and the properties of Th-based materials have been investigated. Both XRD and Raman spectroscopy were used by Nandi et al.212 for the measurement of phase relations of mixed ThO2/UO2 systems under reducing and oxidising conditions, as such mixed oxide fuels are proposed for Advanced Heavy Water Reactors. The measurements revealed information including the formation of defects upon U substitution into parent ThO2 depending on air or reduced synthesis conditions, and the U evaporation losses established by XRD and XRF following thermal annealing of solid solutions under air at temperatures >1873 K. A separate study by de Souza and Lima213 focused on separation of Th from rare earth elements in monazite for potential use in fuel. The separation method focused on solvent extraction of Th from a leach solution in hydrochloric acid and investigated the ratio of monazite leach and organic solution, initial pH values and concentration of extractants. Cyanex 272 and Cyanex 572 were tested, with aqueous solutions measured using ICP-OES. Under optimal pH conditions, routine separation of Th from rare earth elements was achieved, with up to 90% Th extraction achieved.

A number of papers investigated aspects of molten-salt reactor operation. Suzuki et al.214 performed electrochemical and spectroscopic analyses to determine the corrosion resistance of several nickel alloys in high-temperature chlorides, which could form the basis of structural materials. Corrosion rates were estimated using GD-OES based on elemental analysis of alloy surfaces, with variation depending on the chloride melt composition. This was further investigated using Raman spectroscopy and density functional theory calculation. Understanding of the structure of the melt was key to selecting a compatible system of molten salt and structural material. Wu et al.215 studied the recovery of UO2 from oxide spent fuel dissolved at high temperature by means of chlorination. A procedure based on electrodeposition of U in chloride molten salt was developed. The procedure was optimised to obtain pure UO2, with recoveries of 97%, with the products being characterised using XRD, SEM-EDS, ICP-AES and XPS. As well as optimising parameters including cathode material and diameter as well as the electrolysis voltage, doping was performed with La, Ce and Nd. A maximum decontamination factor of 119 was achieved.

As well as dominating measurements in fusion, LIBS has also been used in measurements of molten-salt reactor conditions. Diaz and Hahn216 demonstrated the measurement of off-gassed Na and Ca from molten NaNO3 and lithium chloride-potassium chloride eutectic, respectively. Samples were melted in custom crucibles and measured using a LIBS instrument designed to work in high-temperature environments. The work showed the ability of LIBS to perform real-time monitoring in high-temperature environments that simulate molten-salt reactor conditions. Andrews and McFarlane217 also used LIBS for on-line monitoring of off-gassed hydrogen isotopes for molten-salt reactors. Three spectrometers of varying resolutions were tested for H isotope shift measurements in LIBS spectra of wetted filter paper. The optimal models for each spectrometer were applied to aerosol samples with varying isotopic ratios, with the calibration strategy developed offering an 82% reduction in volume of calibration samples needed. The results were validated through comparison with an all-aerosol trained model for real-time monitoring of H isotope ratios along with representative K, Li and Na salt species.

Graphite acts as a moderator and structural material in some nuclear reactor designs, with several papers investigating the properties of this material. Yan et al.218 used TOF-SIMS to assess impurity distributions in IG-11/110 nuclear grade graphite, on the grounds that impurities can have adverse effects including diminished fuel efficiency and higher radionuclide production. Results showed that metallic impurities were not uniformly distributed but were within graphite porosities or discrete point inclusions. Accelerated oxidation occurs when impurities are present in the walls of these porosities. The results will help in the development of techniques for improved graphite purification. Shalnev et al.219 focused on the kinetics and change in pore structure during the conversion of nuclear graphite by oxygen and carbon dioxide when recovered from a nuclear reactor. The structure of the graphite remained stable with carbon conversion, with analysis showing that gasification of graphite primarily took place on the outer surface, based on inner surface area and particle size distribution studies. Analysis using ETV-ICP-OES was able to separate the trace elements no longer detectable in partially gasified graphite compared with those accumulated in the solid residue.

Llopart-Babot et al. investigated different methods for assessing the36Cl content in graphite.220 Following combustion of activated graphite materials and trapping to collect volatile radionuclides, different separation techniques were tested, and a comparison of liquid scintillation counting and ICP-MS/MS was made. Bartalini et al.221 demonstrated the capabilities of a laser-based spectroscopic technique for 14C determination in CO2. The method was demonstrated in different materials including graphite following combustion. The performance of the saturated-absorption cavity ring-down (SCAR) system was determined to be equivalent to or better than AMS.

Improved efficiency of immobilisation of radioactive wastes into glass was investigated by Shin et al.168 and Li et al.222 Iodine immobilisation was the focus of the study by Shin et al., specifically long-lived 129I that is easily volatilised and therefore must be captured and treated. Silver tungstate–tellurite glass was investigated for this purpose, focusing on the fraction of Ag2O or WO3, with AgI added to each matrix at 10 mol% intervals from 0–40 mol%. Analysis using XRF indicated no significant elemental loss, with leaching evaluated via the product consistency test A. The normalised release satisfying the US regulation of 2 g m−2. In the paper by Li et al.,222 a compound generated from the reaction of 4-(tris(4-carboxyphenyl)methyl)benzoic acid (H4MTB) with Th in simulated high-level radioactive waste can adsorb Ru. This process can separate Th and Ru by more than 68.1 and 37.3%. The waste liquid before and after separation was vitrified with borosilicate glass and iron phosphate glass and then XRD, SEM and ICP-OES were used to measure the glass morphology and composition, with improved stability and waste loading measured in samples immobilised after separation.

As analytical techniques improve, the number of radionuclides measurable in decommissioning wastes and the detection limits achievable remains a topic of considerable interest. Chukhlantseva et al.223 quantified 14C, 129I and 99Tc using multiple techniques in simulated vitrified high-level nuclear waste from spent fuel reprocessing. Following sample preparation using distillation or extraction techniques, 14C and 129I were measured using liquid scintillation counting, whilst 99Tc was measured using ICP-MS. Satisfactory results were achieved in simulated samples, with the results informing the modelling of engineered safety barriers for a deep disposal facility for radioactive wastes. In the case of 129I, this was the focus of a review by Dano et al. that investigated measurement by AMS.224 The technique of AMS is noted as being the most sensitive for 129I determination, but sample preparation can be further refined to avoid losses at high temperatures during the preparation of targets.

Tandem ICP-MS/MS continues to be utilised for radionuclide measurement. The instrument's reaction cell was the focus of multiple studies. In one paper by Hobbs et al.225 NO cell gas was studied for 50 elements, with a focus on direct 239Pu measurement in the presence of interfering 238U. Building on the limited number of previous studies using NO, 239Pu was directly measured without prior chemical separation in samples with a Pu/U ratio of 5 × 10−8. The concentrations measured were within 6% of the spiked values. Kazama et al.226 investigated Np, Am and Cm behaviour when using CO2 and O2 cell gases. These radionuclides are present in fuel debris in nuclear reactor cores, and a reaction model was developed and tested to simulate gas-phase reactions. There was a similarity in the reaction constants between the model and Fourier transform ion-cyclotron resonance mass spectrometry.

Zirconium-93 is a long-lived radionuclide of interest for radioactive waste characterisation. Papp et al.227 investigated measurement of 93Zr along with 237Np and Th isotopes in wastes and mineral samples. Samples were digested using sodium hydroxide fusion or acid digestion, followed by selective extraction chromatography and measurement using ICP-MS and alpha spectrometry. Chemical recoveries of at least 73% were achieved in evaporation concentrates of a nuclear power plant and other samples including soil and concrete. Measurement of 93Zr in high-level radioactive waste was also the focus of a paper by Morii et al.228 using LA-ICP-MS. The Zr isotopes were directly quantified on DGA resin that was used to selectively adsorb Zr. Stable Zr isotopes were determined using IDMS, with the 90Zr value corresponding to a calculated value from the given Zr concentration in the sample, which was seen as a method to justify quantification of 93Zr in the same samples.

Methods have been developed for some challenging radionuclides that are not routinely measurable. One example is 93Mo, which was measured in irradiated steel using ICP-MS/MS.229 This approach was faster than X-ray-based measurements with less dependence on prior chemical separation. The inventory was estimated as 690 ± 83 Bq g−1, supported by an activation calculation conducted using ORIGEN2. Do et al.230 used ICP-MS/MS for the assessment of 126Sn in concrete rubble, as it is a radionuclide that may have been released following the accident at the Fukushima nuclear power plant. A procedure for dissolution, chemical separation and reaction-cell-based removal of 126Sn from interfering Te was developed, with >95% Sn recovery and decontamination factors of 105. The detection limit from measurement of stable Sn was estimated to be 12 pg g−1, equivalent to 6.1 mBq g−1.

4.7. Electronic materials

As always, this has proved to be a popular area of research as workers hunt for lithium ion batteries that are more efficient, last longer, do not burst into flames routinely, etc. Alternatives to lithium ion batteries is another area of growing research as they could be more sustainable, i.e., their materials are more common than the relatively rare lithium. Similarly, with solar-sensitive materials, higher efficiency is the main driving force behind the research. Despite the huge amount of research going into these materials, many of those that use atomic spectrometry do so in a fairly formulaic way, i.e., to characterise the material they have prepared. Consequently, there is little novelty with regard to the atomic spectrometry. It is noted that the instead of the more common methods used for other sample types, e.g., ICP-MS, ICP-OES, LIBS or AAS, it is mainly X-ray-based techniques and (TOF)-SIMS that predominate in these sample types. Those papers that have offered some novelty are discussed in the section below. The increased use of machine-learning algorithms to help interpret data obtained using SIMS/TOF-SIMS was one of the main features of the novelty.
4.7.1. Wafers, thin films, multilayer materials and surface analysis. The first part of this section will cover certified materials that have been developed, and reviews and overviews of the area. However, the first paper to be cited, by Shard et al.,231 described how the international standard (ISO 18115-1-2023) that provides the terminology used for surface analysis has been updated. This included the clarification, modification and some deletions of up to 70 terms and the addition of a further 50 terms. These alterations were made in response to recent trends and issues raised by workers in the area. Terminology and concepts of some relatively new techniques, e.g., atom probe tomography, near-ambient-pressure XPS and hard XPS, have been included. Also included were 25 new terms that describe resolution and ensure that it is used consistently across all of the surface analysis techniques. In total, the revised document now contains 630 terms that describe the samples, instruments and concepts involved in surface analysis. These terms have been arranged into subject-specific sections for easy reference.

Two papers have reported the development of reference materials that are appropriate for these sample types. Li et al.232 discussed the preparation using magnetron sputtering deposition of three AlCu thin-film materials that can be used for the calibration of energy or wavelength-dispersive XRF spectrometers. Following an acid dissolution, the Cu content of the films was determined using ICP-OES. Uncertainties of the materials were also evaluated. The Cu content in the materials and their associated extended uncertainties were calculated to be: 2.56 ± 0.14%, 14.64 ± 0.42% and 49.46 ± 0.98%. Results of the homogeneity and stability tests also met the standards required for reference materials. The other paper to report the production of reference materials was presented by Boone et al.233 These workers developed an electrophoretic deposition method for the fabrication of glassy micro-analytical materials with excellent platinum-group metal homogeneity that also contain numerous other dopants. A series of materials were prepared in sintered silica using either die-pressing of the silica with nanoparticles of 39 dopants or by the developed electrophoretic deposition method. Spatial homogeneity in all of the samples was tested using LA-ICP-MS. The samples produced using the electrophoretic deposition were far more homogeneous for most of the dopants than those prepared mechanically using the die-pressing. The exceptions were for Ag, Cd and In. The reason for this and for why the others demonstrated much better homogeneity was unclear. It was stated that this required further study. However, it was clear that the methodology developed would be far better for producing reference materials.

Previously in this text, it was stated that LIBS is not used for the surface analysis of electronic materials as much as X-ray-based techniques. Although undoubtedly true, it has found some applications and these have been reviewed by Krolicka et al.234 The review contains 109 references, although it should be stressed that not all of these are related to thin films or other electronic materials. Although some of the review focusses on this type of material, other parts focus on materials such as galvanised steels, archaeological artefacts, etc. One large table summarises many of the applications. Crater morphology and how this affects the LIBS signals when depth profiling were also discussed.

Two papers by the same research group have used a combination of XRR and GIXRF to study materials. In one study,235 Melhem et al. discussed how both techniques require the same instrumentation, use the same mechanical processes and have similar concepts and are therefore ideally suited to combine when characterising thin films and multi-layered samples. This is because a combination of the two can overcome ambiguities when analysing surfaces or depth profiles that the use of only one would present. This would lead to more accurate data for film thickness, density, roughness and elemental composition being achieved. The calculation of a thin-film thickness is still not straightforward though and so this study described the development of a chemometric technique to be used on the analytical data. They deposited a thin film with nominal thickness 50 nm of germanium, antimony and tellurium chalcogenides on a silicon wafer and then put a carbon capping layer on top of that. They then analysed the material using the two techniques. Calculation of the uncertainties associated with the data was achieved using a random weight bootstrap method. This is an approach that relies on re-sampling the dataset to estimate statistics on a population by applying random weights. The results were encouraging, with the C layer having a thickness of 3.5 ± 0.17 nm, a surface roughness of 0.24 ± 0.01 nm and a density of 2.30 ± 0.12 g cm−3. This was in excellent agreement with the nominal density. For the chalcogenide layer of nominal thickness 50 nm, the measured result was 55 ± 1.11 nm. This 10% error was attributed to the uncertainty in the deposition method. The density was calculated to be 5.69 ± 0.09 g cm−3, which was slightly lower than the expected value of 5.88 g cm−3. This was attributed to voids and other imperfections in the film.

The other paper by this research group used a combination of GIXRF and XRR to study the density, roughness and thickness of a Ta/Cr/Pt stack that had been magnetron deposited on silica.236 In this sample type, the heavy element layers (Ta and Pt) act as cladding elements whereas the Cr acts as an X-ray waveguide. Three samples were prepared. In all cases, the Ta and Pt layers were nominally 8 and 14 nm thick, respectively, but the Cr layer was either 5, 10 or 15 nm thick. For The GIXRF curves to be built, i.e., the variation of the intensity of characteristic X-ray lines as a function of the glancing angle of the incident radiation, the X-ray spectra first needed to be processed to determine the net area of each characteristic peak. This was done using the COLEGRAM software package. The authors described how this package worked in the paper. Results were not entirely as expected. Instead of a simple three-layer stack, there were several other inter-layers. For instance, instead of a simple 14 nm thick Pt layer, there was one of 13.8 nm as well as a PtSi layer of between 0.2 and 0.7 nm. There was also evidence of Ta and Ta2O5 layers.

Machine learning is a useful tool for sorting data from surface analysis techniques. Sometimes the data files can be extremely large making it difficult and time-consuming to manually identify some of the features as they can be hidden amongst the other data and can only be found using such tools. Often, what is required is a series of samples with known composition or thickness to “train” the model. Another series of known samples can then be used to ensure the answers being obtained are correct, i.e., a set of validation samples. Only when this preliminary work has been undertaken can the model be used on spectra from unknown or “test” samples. Several examples of this methodology have been presented during this review period. A paper by Aoyagi237 discussed the relative merits of using machine learning for the interpretation of surface spectra. It presented some examples of complex data analysis using conventional multivariate methods such as PCA and multivariate curve resolution, an unsupervised learning method based on artificial neural networks (sparse autoencoder) and a supervised learning method (Random Forest). The author stressed that in the supervised model, as much “training” information as possible should be given to ensure highest accuracy. Depending on which application it is applied to, information such as material types, physical or chemical properties, etc., would help the algorithm classify or predict “unknown” samples more easily.

Another example was a paper presented by Zhao et al.238 who used machine-learning analysis of TOF-SIMS spectra to identify lithium compounds on the surface of lithium metal anodes. In this study, the model was built using 5 different pure lithium-containing materials (12 spectra each), namely, Li2CO3, Li2O, Li3N, LiH and LiOH. The model was then used to see if these compounds existed on the surface of the anodes. In this instance, PCA was first used to reduce the number of dimensions of the data from the TOF-SIMS spectra. The machine-learning algorithm employed was logistic regression used with a five-fold cross validation scheme. The paper took the reader through the process in a relatively easily understandable way. It was noted that, even though the model was trained on pure compounds, it was capable of determining their presence in real samples when they were present as a mixture.

Bamford et al.239 reported the use of TOF-SIMS data from the analysis of polyaniline films followed by multi-dimensional machine learning to identify any difference between different films. Spin-coated polyaniline films on indium tin oxide glass slides were prepared and then heat-treated under atmospheric conditions. The replicate samples appeared equivalent both spatially and in composition. When the data were treated using the self-organising map with relational perspective mapping (an unsupervised machine-learning algorithm), a comparison in both 2D and 3D was possible. Structural flaws (e.g., pinholes) and their molecular characteristics were readily highlighted, revealing transport of contaminants to a buried interface. The developed method was able to analyse nine stitched data sets together and highlighted subtle differences between samples, whereas spectral analysis alone provides results that are difficult to evaluate quantitatively. Using the algorithm, subtle differences between samples were observed, which may manifest themselves as a slight change in peak ratios.

The same research group used a similar approach to characterise multi-layer coatings.240 All of the major constituents of a double silver low-E film were identified including the use of SnO2 as a dielectric, ZnO as a dielectric and as seeding layers, TiOx blocking layers, a Zn base layer and a TiOx topcoat. Repeating layers were identified and were classified as chemically identical using the entire mass spectrum. Chemical changes in layers intended to be identical can be easily identified. It was concluded that the technique could be used in industry to investigate manufacturing issues such as pinhole flaw formation, contamination in one or more layers of a film and poor adhesion.

Another example was presented by Bassett et al.,241 who demonstrated that it need not just be the data from one technique that can be input to machine-learning algorithms. They described the preparation of thin films of nickel and nickel–iron and their analysis using techniques such as XRD, XRF, laser profilometry, optical microscopy, nano-indentation and tribological testing (friction and wear). Once all the data had been collected, they were input to a custom database (Experiment Tracker) to facilitate data transfer between the experimental and machine-learning facilities. From there, they were input to a machine-learning algorithm the authors had developed, called the physics-informed multimodal autoencoder (PIMA). The goal of this study was to quantify efficiency gains with the streamlined/automated workflow over typical human-based measurements. To this end, the authors developed a universal sample holder that could be used in each instrument. To assess workflow improvements, the different tasks associated with data acquisition, sample handling, and data saving for 20 samples were timed for each instrument. Results were impressive, with the streamlined processes being factors of 20, 18, 46 and 13 faster for optical microscopy, profilometry, XRD, and XRF, respectively. The methodology reportedly does not require large scale resources and would be suitable for academic, government or commercial laboratories.

A paper by Muramoto et al.242 described the use of plasma polymerisation to construct ultrathin films that ranged in thickness from 1 to 20 nm and their analysis using TOF-SIMS. The data from TOF-SIMS were then input to PCA to investigate the effects of film thickness on the resulting spectra. It was demonstrated that for these cross-linked plasma polymers, at these thicknesses, the observed trends are different from those obtained from thicker films with lower degrees of cross-linking. The differences include contributions from ambient carbon contamination, which starts to dominate the mass spectrum; cluster-induced nonlinear enhancement in secondary ion yield is no longer observed; the extent of fragmentation is higher because of confinement of the primary ion energy; and the size of the primary ion source also affects fragmentation. The choice of primary ion is therefore important to minimise the differences.

One other example has used artificial intelligence during the study. This was a paper by Filali et al.,243 who used particle swarm optimisation to analyse the SIMS profiles of 75As, 11B and 31P implanted in a silica substrate. Implantation energies were between 100 and 130 keV and annealing was for 30 min at temperatures of between 900 and 1030 °C. The particle swarm optimisation was used to calculate activation energy, diffusion coefficient, junction depth and implant dose. Comparisons between real profiles measured using SIMS and predicted ones obtained from the model were made, showing the validity of the method.

The novel technique of SIMS-OES was described by Miyamoto et al.244 When SIMS analysis is undertaken, secondary ions are emitted and determined using mass spectrometry. However, simultaneous to these ions being emitted, light is also emitted that is representative of the target's constituents. This paper made use of this to determine analytes using two different techniques simultaneously. The combined technique was capable of surface analysis and depth-profiling and could be used to determine main components and impurities. The instrument was described in full. At present, the OES part can only detect analytes that emit between 200 and 800 nm. Future work may attempt to expand this into the vacuum UV region. The results obtained for a HfSiOx film correlated well with those obtained using Rutherford backscattering spectrometry.

Numerous other papers have been published that have not used chemometric analysis of the data. Instead, their novelty lies in the actual collection of the data. A paper presented by Alsaedi et al.245 reported the use of gas clusters as primary ions during TOF-SIMS analysis of metal oxide films. Manganese(II) oxide, manganese(IV) oxide, nickel(II) oxide and cobalt(II) oxide were analysed using carbon dioxide, argon or water primary ions at an energy of 70 keV. The secondary ion mass spectra produced depended on the primary ion chemistry and, to a lesser extent, its velocity. Gas cluster beams containing the carbon dioxide or water both enhanced the yield of metal oxide and metal hydroxide secondary ions when compared to a primary ion beam of argon. For all of the gas cluster ion beams used, the ratio of MxOy+ to Mx+ reached a steady state. For Mn, the MnxOy+/Mnx+ ratio reflected the metal oxidation state whereas the MnxOyHz+/Mnx+ ion ratios did not. It was concluded that the use of gas cluster primary ions offers a novel method of analysing the surface and sub-surface regions of metal oxide films.

The technique of LIBS, although not very common for the analysis of electronic materials, has still found some application. An example was provided by Liu et al.,246 who presented a paper entitled “Experimental and model study of LIBS depth profile for multilayer deposition materials”. Much of the paper discusses the use of LIBS for the analysis of plasma-facing materials in nuclear reactors and the elements that become embedded in them. However, they did develop a model called the laser profile & interface roughness model, a full description of which was given in the paper. One-line calibration-free LIBS was used to quantify the depth distribution of elemental concentrations in nickel–copper multilayers. The thicknesses of Ni layers were calculated using different methods, and the relative errors of the calculations were compared. The reliability of their model was verified using these experimental conditions, with results from the model and from experimental data being in good agreement (a correlation coefficient >0.99). An approach to locate the interface between layers provided experimental data and model data that agreed to within 5.1%. The conclusion was that the model developed was very capable of determining layer thickness.

Another paper to employ LIBS for the analysis of thin films was presented by Xia et al.247 These workers used it to undertake a multi-dimensional study of nickel–zinc ferrite films that had been prepared using magnetron sputtering under different powers and pressures. The LIBS was used to determine film thickness and the results compared with those obtained using SEM. The lateral spatial homogeneity was also determined using LIBS, with 2D maps of Zn/Fe and Zn/Ni plotted. These ratios changed with different sputtering pressures.

Giannakaris et al.248 used both single-pulse and collinear double-pulse LIBS for depth profiling studies of multi-layered structures of thickness 10–250 nm. The double-pulse LIBS assembly used a Michelson interferometer to split the laser beam into two beams of equal energy. The full LIBS setup was complex but was described in the paper. Three different sample types were analysed. One was an indium tin oxide layer on a silicon nitride layer on a substrate comprising two layers of organic coating for planarisation on PET (polyethylene terephthalate). The second had a similar substrate and five other layers, some of which were organic and others inorganic. The third sample was an indium tin oxide layer on a glass substrate. The LIBS measurements could be of atomic, e.g., In or Al, or molecular, e.g., CN, species. Analysis of indium tin oxide layers showed a significant increase in sensitivity when double-pulse LIBS was used. This increase was dependant on the laser fluence but was typically a factor of 3–10 more than for single-pulse LIBS. In addition, the double-pulse approach yielded a better depth resolution (40 nm per pulse compared with 60 nm per pulse). It was also noted that the double-pulse setup enabled spectra to be obtained at much lower laser fluence (150 mJ cm−2 compared with 250 mJ cm−2 required for the single-pulse version). The depth profiles of the samples obtained using both setups were very similar.

Glow discharge (GD) with either –OES249 or –MS250 detection is another surface analysis technique that may also be used for depth-profiling studies. In the paper by Geng et al.,249 a model was developed called MRI-Mixing Roughness-Information-CRAter-Simulation (MRI-CRAS). The mathematics behind the model was presented in the paper. Using GD-OES, depth profiles of Ni/Fe foil were obtained with different working powers and gas pressures. The shape of sputtered crater was either measured by a profilometer or simulated by the model they had developed. The measured sputtering time was converted to the sputtered depth by a built-in interferometer. The sputtering rate was dependant on the power, but was 22 nm s−1 at 37 W and almost 52 nm s−1 at 55 W. These represented a much more rapid method of depth profiling than those of SIMS, XPS, etc. It was concluded that the experimental results for crater depth and shape were in good agreement with those simulated using the model. The GD-MS paper by Huang et al.250 described the formation of zinc oxide layers on steel substrates using three different methods of preparation: magnetron sputtering, a sol–gel method and hydrothermal synthesis. These resulted in layers of very different morphology: thin films, network textures and nanorods, respectively. After optimisation of the DC-GD-MS operating conditions to ensure that a flat-bottomed crater was obtained, the conditions of 5.0 mPa, discharge current of 0.8 mA and sputtering crater diameter of 12.5 mm were used for subsequent experiments. The morphology of the layers had a large effect on the accuracy and performance of the depth-profiling, with the zinc oxide film giving better depth resolution (0.22 μm) and indicating a clear boundary between the material and the steel substrate whereas the other materials did not. The authors employed a fuzzy synthetic evaluation method to investigate the results. This clearly suggested that it was the morphology that had a greater influence on crater depth (and resolution) than the thickness of the layer or the sputtering rate.

Numerous other techniques have been used to study thin-film or multi-layer materials. Imashuku251 discussed the use of optical emission spectrometry during the sputter deposition of zinc oxide thin films. During the radiofrequency magnetron sputtering process, light that is characteristic of the elements present in the sputtered material is emitted. The potential for an on-line “as it is formed” analysis of the films is therefore possible. Films were prepared using an input power of 30–60 W and at a chamber pressure of 2–10 Pa. Films were produced at a constant plasma temperature, corresponding to a chamber pressure of 2 Pa and 5 Pa and an input power of 30–60 W. The author compared the O and Zn line emission ratios for numerous combinations of lines and found that five had an R2 value of greater than 0.95, with the best being a combination of Zn at 481.1 nm and O at 777.3 nm, which produced an R2 value of 0.992. Future work was forecast to include using the same methodology for nickel oxide, indium-tin oxide and titanium dioxide thin films.

Another paper to report the on-line (operando) analysis of thin films was presented by Lüchtefeld et al.252 A lithium cobaltite (LiCoO2) thin-film cathodic material that had been prepared using magnetron sputtering was placed in an electrochemical flow cell, which was coupled directly to an ICP-MS instrument. This methodology enabled both time- and potential-resolved dissolution profiles for an accelerated cycling protocol to be obtained. Uniform dissolution throughout three cycles was observed for the thin films but relative dissolution is higher by a factor of about 50 compared with that of commercial composite samples. For both the commercial and the thin-film lithium cobaltite, the dissolution was greatest in the second of the three cycles. This was attributed to the irreversible phase transitions that occur in the materials during cycling. This work represented the first direct investigation of mechanistic changes in transition metal dissolution from lithium ion battery cathodes.

Grazing incidence XRF (GIXRF) is used fairly commonly for the analysis of thin films but grazing exit XRF (GEXRF) is used less commonly. Nikolaev et al.253 compared the use of GEXRF with GIXRF for the analysis of a planar Ta/Co/Cu/Co/Ta structure. The experimental data for both techniques was input to Monte Carlo Markov chain simulations and the structure of the material “reconstructed”. The results were largely in agreement, with the main discrepancy arising in the concentration data. It was hypothesised that this may be because of uncertainties in the scattering factors. It was noted, however, that the GEXRF data were more precise than those obtained using GIXRF. This was attributed to the multi-probe nature of GEXRF and it using a series of X-ray standing waves rather than relying on just one like GIXRF. Using multiple X-ray standing waves allows the collection of more data in a single scan. In addition, each X-ray standing wave probes a different area in reciprocal space, each corresponding to the wavelength of different characteristic X-rays. It was concluded that, assuming the problems with uncertainty in the scattering factors can be overcome, GEXRF could become a high-precision nano-metrology tool.

High-energy laser desorption ionisation (HELDI) is a technique that uses extreme ultra-violet laser pulses to enhance and homogenise the sensitivity at the nanoscale. This is especially true for elements that are at the lower end of the periodic table, e.g., B, Li, etc. Bleiner254 used HELDI in combination with TOF-MS to investigate nanostructures in 3D, i.e., for studying the lateral surface of the material as well as the depth profile. Several film types were analysed, including copper zinc tin sulfide. The 10 × 10 × 10 data point blocks (each point is a full MS spectrum) were processed using a script written in-house that enabled the extraction and visualisation of the underlying information within a few seconds. Based on a few reference mass peaks set by the author (6, 12, 23, 63, 80 and 120) the script performed a TOF-MS calibration of the raw signals. The average mass spectra were calculated layer by layer. The spectra acquired across a 3D space were then assembled into a 3D matrix. These plots enabled a 3D elemental map to be created where flaws in the film, e.g., crevices, as well as impurities and inter-diffusion between layers, are easily identified. It was also thought that it would help discern between instrumental noise and heterogeneities in the sample.

Yuan et al.255 developed a coated inner wall, mono-capillary ellipsoidal lens that can be used with a μXRF instrument for the determination of thin-film thickness. The coated elliptical lens has a larger acceptance angle and higher utilisation efficiency for light sources compared with traditional uncoated equivalents. When used with the μXRF instrument, compared with common XRF devices, it has a smaller X-ray spot size and higher intensity gain. This resulted in higher spatial resolution and higher detection efficiency during XRF film thickness measurement. A standard curve in the range from 0.85 μm to 56.5 μm for measuring the thickness of Cu coating on silicon was prepared. In the thickness region from 0.85 μm to about 30.00 μm a relative deviation of about 3% was obtained. This deviation rose to above 5% in the thickness range 30–56.5 μm. The device reportedly had the advantages of simplicity and speed.

Kang and Ko256 discussed the use of laser-induced X-ray fluorescence to measure multi-layer materials, albeit not ones of electronic use. They analysed three types of stainless steel, copper covered stainless steel and three types of Korean coins. The same methodology as for PIXE was employed. A double-amplified femtosecond titanium–sapphire laser system with a pulse width of 36 fs, repetition rate of 1 kHz, wavelength of 800 nm, and power of 10 W was used under vacuum conditions of 10−4 torr, The laser was described as being compact and of low intensity. The laser energy required for PIXE is 1018 W cm−2, whereas for the laser-induced X-ray fluorescence, it was only 1016 W cm−2.

It is not often that a speciation paper makes it into the Industrial ASU. However, Fujihara and Nishimoto257 determined the oxidation speciation of Sb leached from thin films of indium antimonide using 0.1 M sodium acetate. As with many analytes, there is a very different toxicity between SbIII and SbV and so determining total Sb means very little with regard to overall toxicity. Total Sb was determined and then SbIII was determined using hydride generation-microwave plasma-optical emission spectrometry (HG-MP-OES) ensuring that the SbV was not reduced to SbIII. This was achieved using a reduced concentration of sodium tetrahydroborate and by not using potassium iodide reducing agent. The SbV concentration could be determined by subtraction of the SbIII concentration from the total Sb concentration. As well as plain indium antimonide films, bismuth-impregnated indium antimonide films were also analysed. The presence of the Bi did not affect the speciation. However, it did seem to stabilise the material so that less Sb was leached from it. The main species found in the leachates was SbIII.

4.7.2. Electronic components. A review entitled “TOF-SIMS in battery research: advantages, limitations and best practises” and containing 52 references was presented by Lombardo et al.258 The goal of the paper was to encourage new and experienced workers with TOF-SIMS to help workers in the field of battery development and to add it to the techniques they use for characterisation. An introduction gave the working principles of the technique along with its advantages and limitations. The advantages were listed as: high surface and chemical sensitivity; high mass and spatial resolution; low detection limit; the sample bulk can be measured when combining TOF-SIMS with a sputter or focussed ion beam gun; all types of electrodes and powders can be analysed; organic, inorganic, and hybrid materials can be analysed; different isotopes can be distinguished between and both positively and negatively charged secondary ions can be determined. The disadvantages were listed as: a lack of advanced chemical information, e.g., on the oxidation state of the detected species; differentiation of chemical species is possible only if the secondary ions generated through the associated fragmentation process are different; secondary ion shielding caused by sample roughness/topography; lateral resolution and surface sensitivity are limited and semi-quantitative information is obtained. Following that, a list of capabilities of current commercial instruments was presented. Sections on best practices and pitfalls and then a lengthy section on recent advances on data processing followed. A final conclusions section summarised the previous sections nicely.

A second review, by Vahnstiege et al.,259summarised the capabilities of the different techniques that can determine the state of charge of individual active material particles in lithium ion batteries. The state of charge is an important factor for batteries, since it gives the remaining capacity in individual battery cells but also gives information on the redox processes of the cathodic components during electrochemical cycling. These properties can be influenced by factors such as active material properties, inhomogeneities in the electrode, degradation of the electrode, charge/discharge cycling properties, etc. Unless these things are identified and, in some cases prevented, the lifetime of the battery can be compromised. A wide range of techniques have been employed to determine the state of charge of battery components. These include XANES, transmission X-ray microscopy, Raman spectroscopy, TOF-SIMS and single-particle ICP-OES. The areas of application, advantages and drawbacks of each of these techniques were discussed with the aid of 84 references.

One final review is of relevance to this section of the update. This was presented by Sultan et al.,260 who reviewed (with 68 references) the on-line ICP-OES applications of real-time element-resolved electrochemistry. Such on-line techniques have been used to measure element dissolution rates directly, to determine element-specific reaction mechanisms, degradation mechanisms, etc. The paper gives some of the most recent examples where an electrochemical flow cell has been coupled with ICP-OES (and occasionally ICP-MS). In addition, examples where it was also coupled with other analytical techniques, e.g., gravimetric hydrogen measurement, electrochemical quartz microbalance and electrochemical impedance spectroscopy, were also discussed.

Numerous components or electronic parts have been analysed during this review period. Some of the more interesting ones will be discussed below. Tin dioxide has been studied as a potential anode material for lithium ion batteries previously, but was found to have severe deficiencies in that it exhibited large volume expansion and severe structural collapse during cycles. Wang et al.261 re-visited tin oxide but this time blended nanoparticles of it with graphene making a composite. The material was analysed using XRD, SEM, XPS and SAXS. In particular, electrochemical SAXS was used during the first to the tenth discharges. For the tin dioxide particles alone, severe pulverisation was observed after the expansion. However, although the composite also expanded after discharge, the particles did not pulverise – even after the tenth discharge. The SAXS data enabled them to attribute this to the multi-hierarchical scatterers in the anode materials being “roughly divided into gap, interspace, tin dioxide nanoparticles, nanopores and so on”. These results indicated that composite structure can buffer large volume changes and effectively prevent the detachment and pulverisation of tin dioxide particles during the lithiation and de-lithiation processes.

Another paper to analyse anode materials of lithium ion batteries was presented by Langer et al.262 These workers examined the effect that different GD-OES operating parameters had on the sputtering rate, crater size and profile of graphite anode materials. Increasing the voltage applied between 500 and 700 V increased the sputtering rate by 100% per 100 V but did not affect the crater shape. Crater shape and size was affected by the argon gas pressure though. The effects of this were tested over the range 160–300 Pa. At low pressure the bottom of the crater was concave in shape. This became flat at medium pressure before becoming concave again at higher pressure. An increase of the duty cycle in the pulsed glow discharge mode led to a linear increase of the sputtering rate, whereas a pulse duration enhanced the sputtering rate in a nonlinear fashion. Different pulsing conditions can therefore enhance the sputtering rate without affecting the crater shape significantly. It was also noted that a lower electrode density (1.28 g cm−3) led to a larger sputtered volume compared with one with a density of 1.43 g cm−3, as well as producing a crater with a larger concavity.

Two papers have reported the operando analysis of batteries but using different techniques. Sometimes a full understanding of how a battery works cannot be obtained by pre-use and post-mortem analysis. Only when the battery is functioning can some important bits of information be gleaned and this is where operando analysis can be useful. In one example by Kohmoto et al.,263 X-ray computed tomography and an unsupervised machine-learning algorithm (non-negative matrix factorisation) were used with a graphics processing unit to detect and analyse deterioration in sections of a battery during charge–discharge tests in real time. The method developed could detect electrochemical changes in a battery through the conventional voltage–capacity diagram. It could also detect physical changes such as the deterioration of the parts of a battery that cannot be found via human inspection. It did both of these directly from the sliced images of the three-dimensional reconstructed volumes. Charge–discharge tests were completed for 100 cycles on a nickel–zinc battery and computed tomography scans made every 10 cycles. At cycle 50, the method detected that there had been electrochemical changes within the battery. It also detected physical changes such as deteriorating parts that cannot be found directly from sliced images of a 3D reconstructed volume. Both continuous (deformations of active materials caused by ion diffusion and electronic speed in a battery) and discrete internal structural changes, such as compounds precipitating randomly in electrolytes during the charge/discharge cycle, could be detected. Since the developed method could detect electrochemical and physical changes in a battery, it has an advantage over established methods that can detect only electrochemical changes.

The other operando-based method was described by Cressa et al.,264 who developed a sample holder that enabled various electrochemical experiments to be undertaken. The entire workflow was based on a single beam scanning electron microscope that was equipped with a magnetic sector SIMS instrument that had been developed in-house. Both the sample holder and the instrumentation were described in full in the paper. Using this novel instrumentation, these workers were able to pause the battery mid-cycle, perform an analysis and then continue the cycle again. This combination of techniques enabled the SEM component to undertake the micro-structural analysis while the SIMS could do the chemical mapping. In this proof-of-concept paper, a lithium lanthanum zirconium oxide (Li7La3Zr2O12) powder was pressed into a circular pellet, split in half to make two semi-circles. One semi-circle was sandwiched between layers of lithium foil, which was then mounted in the custom-made sample holder. The battery was then put through 15 cycles using four different currents (5, 10, 20 and 40 μA). This gave a total of 60 cycles, with each cycle taking an hour. The system identified several artefacts that affected the battery efficiency. Not least of these was the formation of dendrites. The system managed not only to detect the presence of the lithium-rich dendrites but also to map where they were. The example given in the paper was for a solid-state battery but the authors envisaged that their system could have a large number of uses for different sample types. The main advantage offered is that there is no requirement for sample to be moved between different instrument types. Instead, it may be done automatically and simultaneously using this combined instrument.

Re-cycling or recovery of metals from waste electronic equipment has been another popular area of research because many of the metals are either scarce/valuable or are toxic and hence should not be released to the environment via landfill. Many of the papers in this subject area are pretty routine. However, two offered some novelty. Ferreira et al.265 used a knife mill to shred e-waste boards prior to them being analysed in three different ways. An acid digestion in aqua regia was followed by ICP-OES analysis. This was the “reference” method with which the results from the other methods could be compared. The shredded material was also analysed using EDXRF employing the loose powder technique. The material was also pressed at 60 kN for a minute and the pellet produced was then analysed using LIBS. The LIBS analysis determined Al, Cu and Fe in a total of 130 spectra, which were acquired per sample using the raster mode of analysis. The samples were clearly not homogeneous and so numerous signal fluctuations existed. To minimise these effects, numerous normalisation methods were attempted and calibration was attempted using different multivariate tools. These tools included partial least-squares regression (PLS), principal component regression (PCR), maximum likelihood principal component regression (MLPCR) and error covariance penalised regression (ECPR). A brief description of each was provided in the paper. For the Al determination, the best results were obtained using the MLPCR method although it was noted that problems with acquiring LIBS data were observed. The Cu data were best when employing the PLSR method when peak height was used and employing ECPR when peak area was used, although both provided excellent linearity. The Fe was also best using ECPR. Univariate calibration models were also tested for both the EDXRF and LIBS but were not really successful.

The other paper of note with regard to metal recovery was presented by Ichikawa et al.266 These workers cryo-milled printed circuit boards before WDXRF analysis of the loose powders employing the fundamental parameter method rather than calibration curves. Particle size distribution measurements identified two high-frequency areas: 27–45 μm and 250–500 μm. Analysis of the materials found that Cu was associated with larger particles, Br and Sn tended to be more in the moderately sized particles and Fe was associated most with the fine fraction. Other analytes (Al, Ba, Ca, Ni, Pb and Si) seemed to be equally dispersed between different size fractions. X-ray diffraction was used to identify the chemical form that each analyte was present as. The analytes Cu, Ni and Si were present as the elemental form, Al was present as either the element or as corundum (Al2O3), Ba was present as the titanate, Ca the carbonate and Pb as the zirconate. Using multivariate statistics (PCA and cluster analysis) on the XRF data indicated that the amorphous analytes were Br, Ca, Si and Sn.

Three papers have discussed the analysis of different types of insulators or their precursors. A paper by Lee et al.267 discussed the SF-ICP-MS determination of Cl in hafnium precursors of insulators with a high dielectric constant. The materials were diluted using 1-methyl-2-pyrrolidinone because they precipitate in the presence of moisture/water. The Hf in the diluted solutions was determined to ensure that complete dissolution had occurred. Calibration standards for the Cl were prepared from 2,4,6-trichlorophenol. A 100-fold dilution of the precursor led to a LOD of 0.61 μg kg−1 and a LOQ of 1.81 μg kg−1 being obtained. This translates to a sample quantitation limit of 61 μg kg−1. The Cl concentration in the precursors was <LOQ. A spiking experiment where 10 μg kg−1 was added to the diluted material gave a result of 10.52 μg kg−1, representing a recovery of 95.60% with an acceptable precision of 3.58%. The method could potentially determine Cl in a large number of samples quickly and reliably. Although it did not detect cl in the precursors, it would be capable of detecting it if the samples were contaminated.

There is a new class of insulators called the topological crystalline insulators. These have the general formula of Pb1−xSnxTe. Khosravizadeh et al.268 used SIMS to determine the exact ratio of SnCs+/PbCs+ and Sn±/Pb± in thin layers of the material deposited on gallium arsenide. The positive ions were determined using a Cs primary beam energy of 5.5 keV and the negative ions at 14.5 keV, which collide with the surface normal at an angle of 30°. The sputter rates for positive and negative secondary ions were 3 nm s−1 and 5 nm s−1, respectively. Since there was an isobaric interference between 120Te and 120Sn, the 118Sn isotope had to be used for all measurements. The SIMS signal ratios of Sn±/Pb± and SnCs+/PbCs+versus Sn/Pb mole fraction ratio were plotted to obtain the calibration curves for the ternary precursor compound. Depth profiles were also acquired and matrix effects evaluated.

The third paper presented by Song et al.269 described the use of LIBS for a classification study of composite insulator materials. Silicone rubber compositions from different manufacturers differ and all can become weathered (aged) and start to malfunction. This work used insulator materials from seven different manufacturers and the spectral lines emitted during LIBS analysis were used to classify the samples. Having obtained the analytical data, they were input to a recursive feature elimination to identify redundant areas of the spectrum and then input to a linear discriminant algorithm to remove them from the dataset. A back propagation neural network algorithm was used for the actual classification with a success rate of approximately 95%. It was concluded that the methodology offered a rapid, accurate method for the identification of different formulations. This could potentially help in the maintenance of the insulators and identify those most likely to fail.

Nickel implanted into silicon shows promise in solar cells, IR detectors and spintronic devices. A paper by Alam et al.270 discussed the implantation, treatment and subsequent depth profile analysis of such materials. This would lead to an understanding of the nickel distribution as well as identifying any structural changes. A dose of Ni ions (5 × 1016 ions cm−2) was implanted at an energy of 100 keV and then annealed at 800 °C for 2 h. Analysis using XRR-GIXRF, SIMS, RBS, SEM, XANES and EXAFS was then undertaken. The XRR and GIXRF analysis indicated that the implantation process led to a decrease in the surface density and the annealing process led to an increase in surface roughness (from 2 nm to 3 nm). The RBS and SEM analyses led to the same conclusions being drawn as those obtained using XRR-GIXRF. The SIMS measurements for the as-implanted sample showed that the distribution of Ni ions follows Gaussian like behaviour. The projected range was also found to agree closely with the results obtained from the RBS and GIXRF measurements. However, in the case of the annealed sample, the measured SIMS profile of Ni atoms was found to be considerably shifted towards the substrate surface and did not resemble a Gaussian nature. Evidence that annealing makes the Ni migrate towards the surface and to greater depths was found though. This discrepancy was attributed to the inconsistent sputtering yield of Ni ions up to a few nanometers depth in the Si substrate. The XANES data showed clear evidence of the presence of inter-metallic compounds NiSi and NiSi2 in the annealed material, but evidence of Ni2Si in the as-implanted material. Thorough analyses such as this one where mechanisms can be elucidated enable a better understanding of the processes to be obtained.

Solid-state electrolytes have been studied for many years in an attempt to make lithium ion batteries safer. However, it is known that lithium has a much lower diffusion coefficient in such materials. This makes its transport more difficult to monitor. Isotopic tracing can potentially be useful for this analysis. A paper by Gallot-Duval et al.271 used a technique entitled laser-induced breakdown self-reversal isotopic spectrometry (LIBRIS) to track the movement of Li through the electrolyte. Two lithium carbonate powders were obtained, one of which had natural isotopic abundance and the other that had 95.5 ± 0.1% 6Li. These were mixed with the solid-state electrolyte polyethylene oxide containing lithium bis(tri-fluoromethylsulfonyl)imide. After drying, two thin films of 80 μm thickness were produced, one of which was enriched with the 6Li. Two different LIBS systems were then used: one that offered a lateral resolution of 250 μm and the other a resolution of 7 μm. The operating parameters of the systems were detailed in the paper and differed significantly, although both utilised a frequency-quadrupled Nd:YAG laser (266 nm). Also discussed in the paper was the way in which the LIBS spectra were treated. It should be noted though that the 6Li and 7Li emit at slightly different wavelengths (670.7760–670.7911 nm and 670.7918–670.8068 nm, respectively for 7Li and 6Li) and so a sufficiently high-resolution spectrometer can distinguish the two. Any movement in the 6Li could therefore be tracked. Issues with accuracy associated with single laser shots still existed, but future work focussing on instrumental modifications could be implemented to better control the sample surface state and positioning in the ablation focal plane could lead to improvements. Corrections to fluctuations of the laser pulse energy and increasing the efficiency of the plasma emission collection setup may also be useful.

A relatively simple application was presented by Medvedev et al.272 who determined trace analytes in concentrates of germanium dioxide. The material (0.25 g) was first dissolved in hydrochloric acid (3 mL) and heated at 80 °C for 8 h. The resulting solution was transferred to PTFE bowls and evaporated to dryness at ∼80 °C under an IR lamp. The germanium was volatilised in the form of germanium chloride. After evaporation, the dry residue of the concentrate was transferred into solution by adding 100 μL of high-purity nitric acid. This was then diluted using 1.4 mL of 0.5 M nitric acid solution and analysed using ICP-MS. This procedure was adequate for determining analytes present at higher concentration. For those analytes at lower concentration, the dissolved material present in the 100 μL of acid was placed on a pre-cleaned and etched silicon wafer, dried under an IR light and then the thin film of residue formed was analysed using LA-ICP-MS. Limits of detection for the LA-ICP-MS method were at the pg g−1 level for the majority of the 47 analytes, representing an improvement by factors of typically between 1 and 30 over those using conventional ICP-MS.

Critical metals (Ce, Co, La, Mn and Ni) were determined in sulfuric acid extracts of nickel metal hydride batteries using MIP-OES by Cruz et al.273 Traditional calibration uses different concentrations of analyte all measured using the same wavelength. These authors used multi-energy calibration, i.e., they used only one concentration of analytes but measured the response over several different wavelengths per analyte. Only two calibration solutions are required: solution 1 was composed of 50% v/v sample and 50% v/v of a standard solution containing the analytes and solution 2 has 50% v/v sample and 50% v/v blank. This calibration methodology is supposed to assist in circumventing matrix effects and interferences. To this end, the results obtained were compared with those obtained using conventional calibration and those from standard additions. The multi-energy calibration yielded superior results, with analyte recoveries in the range 90–110%, precision from 1.8% to 5.8%, and limits of detection of 400, 60, 20, 1 and 10 μg kg−1 for Ce, Co, La, Mn and Ni respectively.

Another simple application is presented in a paper by Muss and Koch,274 who used X-ray absorption profiles to characterise bonding wires and to identify counterfeit components. The X-ray absorption profiles could easily identify the material the wire was made from, i.e., aluminium, copper, gold and silver. However, it was possible to correlate the full width at half maximum of the absorption peaks with the wire thickness. The method had the advantage of being quick, reliable and having a very low error rate, but required well-characterised wires so that a “calibration” could be constructed.

Normally, XRF is used to determine the elements present in a sample and XAS (either XANES or EXAFS) is used to determine the oxidation state in which they are present. A paper by Maisuradze et al.275 discussed how XRF could be used to differentiate between different oxidation states of Mn present over the whole surface of a manganese hexacyanoferrate cathode material. In a typical XRF experiment at synchrotron facilities, the energy of the beam is chosen to exceed the K-edge energy of the elements under the investigation. Therefore, an energy of 7200 eV was initially selected to evaluate the distribution of Mn and Fe in the electrodes. However, a second energy of 6553 eV, was also chosen since this should enable differentiation of the Mn +2 and +3 oxidation states. This is because the edge maximum of Mn2+ is situated at 6553 eV, while for Mn3+, this value is at 6558 eV. The results for the oxidation-state study using this methodology were comparable to those obtained using XANES. The advantage of the technique was that it was very rapid (especially useful for operando measurements) and it could be used to analyse the whole surface of the material, something that most other techniques cannot achieve.

The on-line analysis of vacuum circuit breakers was described by Zhang et al.276 The system developed employed LIBS followed by Random Forest treatment of the analytical data to monitor the vacuum level of the breaker. Four analytes were determined, one on the target material (Cu) and three in the ambient gas/vacuum (H, N and O). The LIBS system was described in full, but utilised a Nd:YAG laser operating at 1064 nm, at a pulse energy of 30 mJ and with a repetition rate of 5 Hz. A total of180 raw spectra were acquired at nine different vacuum levels (ranging from 10−3 Pa to 105 Pa). As is often the case, analytical data were pre-processed to remove noise arising from scattering, inhomogeneity of the sample, etc., before the data was input to the Random Forest algorithm. Four types of pre-processing were tested: variable importance random forest, first-order derivative, multiple scattering correction and standard normalisation. For the construction of the Random Forest model, 126 sets of the spectra and corresponding vacuum-level labels were used as the training set, while the remaining 54 sets were employed as the test set. Not only was the LIBS data used to identify the presence of the trace elements, but the broadening of the peaks and peak heights were also vacuum-level dependent. Once constructed, the Random Forest model could classify samples according to the vacuum level from the trace elements present. The paper discussed the pre-processing methods and how the Random Forest algorithm works. The best results were obtained when the variable importance random forest pre-processing was used prior to Random Forest classification. This had an accuracy rate of over 99% for vacuum classification. Since it is such a quick method, it may be used on-line and hence has several advantages over more traditional off-line approaches.

4.8. Nanostructures

Atomic spectrometry, through techniques such as XRD, XPS, XRF, single-particle (sp)-ICP-MS and ICP-OES, has a key role in the characterisation and detection of nanoparticles (NPs) with over 200 papers published in the period covered by this ASU. As this section focuses on the analysis of the NPs themselves, papers covering NP detection in a wide variety of sample matrices are to be found in the other ASUs in this series, as cited earlier in this review. In addition, many published articles only mention the technique(s) used without any further analytical detail and as such are not reported on here.
4.8.1. Topical reviews. A number of reviews covering NP measurements have been published this year. The first of these (227 references) covers the analytical chemistry of engineered nanomaterials (ENMs) with regard to their analysis in complex samples.277 The paper discusses the detection, characterisation, and quantification of ENMs, including coverage of methods for sample preparation in environmental, food, cosmetic and biological samples, as well as those used to monitor the fate of ENMs in the environment and biological systems, and the fitness for purpose of these methods. The authors point out that there is a clear difference between pristine ENMs, which have a clearly defined chemical composition of their core, specific surface coatings, and monodisperse size distributions, and the same nanomaterials in complex, real-world matrices, where they typically undergo different transformations. This can lead to difficulties in the identification, characterisation, and quantification of these materials and more complex analytical approaches are therefore required. Thus, the analysis of ENMs in real samples requires a combination of procedures and instrumentation rather than a single analytical technique, which is often used for pristine ENMs. They also conclude that method validation is required, and this should be based on standardised or reference materials and inter-laboratory comparison studies at the international level. Zhou and Beauchemin have reviewed (74 references) the various processes involved in sp-ICP-MS.278 This work gives a good overview of the theory behind and the steps required to produce accurate sp-ICP-MS measurements. This is followed by coverage of flow injection and mono-segmented flow analysis sample introduction approaches and the data processing differences required compared with continuous sample aspiration approaches. The advantages of IR heating of the sample introduction system are also discussed, such as improvements in sensitivity, along with a brief mention of the use of mixed gas plasmas to improve the LODsize. The use of LA for NP detection is also briefly discussed. The detection of NPs at the single-cell level is becoming more common, in part due to NP-based drug delivery systems increasingly being used and advances in this area have been reviewed (128 references) by Davison et al.279 The techniques covered were sp-ICP-MS, which has the advantage that a large number of cells can be analysed rapidly but does not provide spatial information, LA-ICP-MS, which enables elemental mapping at the single-cell level, although standardisation requires further development, and LIBS, although the sensitivity of commercial instruments currently restricts its use for cellular applications. However, developments in nano-LIBS technology are showing potential for cellular research. The paper contains a wealth of information, including the critical step of pre-treatment of cell suspensions and cell fixation, which the authors state requires further study to ensure validity and that novel validation methods are needed. A section on challenges and future recommendations is also included, one of which is that whilst the detection and quantification of metals and metalloids at the cellular level is possible, methods to determine their chemical speciation to unravel their biological role are required. The use of ICP-MS-based techniques to investigate NP toxicity and transformations during in vitro and in vivo toxicological assays has also been reviewed (121 references)280. The paper covered, in depth, analysis for total elemental amount, sp and single-cell, speciation and LA. The ability of LA as a spatially resolved sample introduction approach was explored and discussed. The authors concluded that the complete characterisation of NPs in complex matrices requires the use of multiple analytical techniques, of which ICP-MS can play a major role, to give a full understanding of NP toxicity and biomolecular or cell interactions. A separate review, with 140 references, on the instrumentation required for the characterisation of nanostructured materials in biomedical applications has also been published.281 The atomic spectrometry aspect relates to the use of XRD, TEM, XPS, EDX and SIMS for this purpose. Each instrumental section gives a description of the fundamentals of the technique and then uses case studies, with relevant papers cited, to describe uses of each technique.

Continuing with topical reviews, Loeschner et al. have reviewed (83 references) the application of sp-ICP-MS for the determination of inorganic NPs in food additives and food.282 Most of the work cited was presented in tabular form, with short sections highlighting aspects of some of these papers covering sample collection and preparation, analytical techniques including sample introduction and a welcome section on method validation. The authors pointed out that more screening studies of foods for NPs are needed, as not all NPs detected in foods are of anthropogenic origin but can be the result of natural detoxification processes of ionic metal species. They also call for the instrumental optimisation procedures, including sample rinse conditions, to be reported to enable the transfer of measurement procedures more consistently across laboratories. The lack of suitable matrix-based NP reference materials is also highlighted as a major issue and current limitation affecting the accuracy of sp-ICP-MS calibration, sizing and quantification results of sp-ICP-MS in general and in the food analysis field in particular. Although commercially available NP suspensions are often used for this purpose, the authors also stated that value assignments provided by the manufacturers have been demonstrated to be very limited and therefore, a more thorough in-house characterisation of commercial NP suspensions is required prior to their reliable use as calibration standards. It was also stated that, based on information from inter-laboratory studies and in-house validation publications, sp-ICP-MS tends to perform reasonably well for NP size distribution assessments. However, important challenges remain in obtaining accurate and consistent particle number concentration measurements. This is related to, in part, inaccurate calibration of transport efficiency, instability of NPs after extraction from food matrices and loss of particles to the surface of the sample introduction system or the side walls of sample containers. Borowska and Jankowski have assessed, citing 116 references, the current status and future prospects of atomic and molecular spectrometric methods for complete NP characterisation in biological and ecological systems.283 The review focused on: NP synthesis yield, the determination of the particle size, stability, solubility, aggregation/agglomeration state and particle number concentration, elemental and isotopic composition and surface characterisation. Amongst the conclusions drawn is that MS-based techniques are the most commonly used for NP analysis and that, once again, validated methods for quantitative NP detection and characterisation in biological and environmental samples are needed. Meng et al.284 reviewed (123 references) the use of sp-ICP-TOF-MS for analysing environmental samples. The review covered the theory of TOF mass analysers, sample introduction systems, the discrimination of background and NP signals, quantification and environmental applications, and also included a short section on single-cell analysis. With the number of ICP-TOF-MS instruments on the market and in laboratories increasing, their use for this type of application is likely to grow rapidly; thus, the review makes a good starting point for those new to this field.

4.8.2. Metrology and inter-laboratory comparisons. Robust uncertainty estimates for measurements of NP characteristics are not often reported so it is good to see this undertaken for particle size by sp-ICP-MS.285 Yamashita et al. estimated UNPsize for two different calibration methods, namely with a particle size standard or an ionic calibrant solution. In this work, the ICP-MS was operated without a collision cell gas and with a dwell time of 100 μs, the transport efficiency was determined using the particle size method and particle events were differentiated from the baseline signal using the blank signal plus five SD of the blank. Uncertainties were estimated according to the principles of the guide to the expression of uncertainty in measurement and full details are given in the paper. For the particle size standard approach, the main uncertainty source was due to the variation in the signal intensities of both the target NPs and the particle size standard, and the size distribution of the particle size standard. The relative uncertainties for 50 nm Ag NPs were 15.0%, 9.9%, and 10.8% when particle size standards of 30 nm, 60 nm, and 100 nm silver NPs were used as calibrants, respectively. For the ion standard solution approach, the sources of uncertainty were the concentration of working standard solution, sample flow rate, transport efficiency, slope of calibration curve and variation in the signal intensity of the ion standard solution and of the Ag NPs. The relative uncertainties for the 50 nm silver NP were 18.5% for 1 ng g−1, 7.6% for 10 ng g−1 and 4.7% for 100 ng g−1 solutions. Thus, the authors recommended that a higher concentration calibrant is used in order to minimise measurement uncertainty, noting that the detector cross calibration between pulse and analogue modes must also be ascertained and set as part of the instrument tuning process.

There have been four inter-laboratory comparisons involving the analysis of NPs reported this year. An accurate assessment of the transport efficiency of aspirated solutions to the plasma is a fundamental requirement for sp-ICP-MS measurements. Geiss et al.286 reported, in a comprehensive paper, a comparison with 7 participants, to assess the variation and accuracy of two commonly used approaches for measuring the transport efficiency, the particle size and particle frequency methods. Each method was assessed on three different days with six well-characterised (details given) Au NP suspensions using different ICP-MS instruments and sp-ICP-MS software. The results obtained showed that the particle frequency method systematically resulted in lower transport efficiencies (0–300% relative difference), which depended largely on the choice and storage conditions of the NP suspensions used for the determination. Thus, the authors recommended that the particle size method is used when the principal measurement objective is particle size of unknown NPs, whilst if the particle number concentration is the main parameter required, the particle frequency approach could be preferable as it might better account for particle losses in the sample introduction system. Hendriks et al.287 described a comparison, with 9 participant laboratories, to evaluate and validate a standard operation procedure, which was included as supplementary information, for the determination of particle mass, particle number concentration and isotopic compositions, with Pt NPs of nominal 50 and 70 nm diameters, using sp-ICP-TOF-MS. The mass equivalent spherical sizes of the two Pt NP suspension studied were measured as 40.4 ± 7 nm and 58.8 ± 8 nm, respectively. The size results had 16% or better RSD among all participants, which was reduced to less than 4% RSD after the exclusion of two outliers. Good agreement was obtained between the different participating laboratories regarding particle mass, but the particle number concentration results were more scattered, with a 53% RSD among all laboratories, which the authors stated is consistent with results from previous comparison studies conducted using sp-ICP-Q-MS instrumentation. For the 194Pt[thin space (1/6-em)]:[thin space (1/6-em)]195Pt isotope ratio measurements, a precision of 1% RSD or better was obtained for an average of 1000 particle events in the scan duration window whilst the accuracy of other isotope ratios, which involved lower abundant isotopes, was limited by counting statistics. The third inter-laboratory comparison covered here, with three participants, reported the use of centrifugal ultrafiltration, with ICP-MS as the analyser, as a screening method for nanomaterial release from drinking water contact materials in the water purification and supply network.288 The accuracy, precision, and reproducibility for the proposed method were assessed using mixtures of ionic and NP Au in a standard, widely utilised model water matrix (NSF International Standard 53/61). The results obtained, using a variety of solution compositions, showed that both forms of Au could be consistently discriminated at the μg L−1 level, within a 10% margin. It was also found using a mass balance approach that Au NPs were retained on membranes within the ultrafilters. It is noted by the authors that further work is needed to assess the suitability of the techniques for more readily dissolvable NPs. The final inter-laboratory comparison reported on assessed the use of TOF-SIMS for the surface analysis of TiO2 NPs.289 A total of 11 participants returned ToF-SIMS data, in positive and (optionally) negative polarity, using sample preparation methods named as “stick-and-go” as well as optionally “drop-dry” and “spin-coat”, with full details of these given in the paper. The results showed that the largest sources of variation within the entire data set were caused by adventitious hydrocarbon contamination or insufficient sample coverage, with the spin-coating protocol applied in this comparison exercise showing a tendency toward insufficient sample coverage. The sample preparation method or the participant had a lesser influence on results. The conclusions drawn from the study were that no sample preparation method stood out as clearly superior to the others, which is a disadvantage for method standardisation but has the advantage that suitable methods are available for both nanoparticle powders and suspensions.

4.8.3. sp-ICP-Q-MS studies. The measurement of the transport efficiency of the sample to the plasma is a key step in obtaining accurate results for sp-ICP-MS. The two most common approaches for determining the transport efficiency are the particle size and particle frequency methods, often using NPs of a different element to the target NP due to the availability, or lack of, well-characterised NP suspensions, which is a prerequisite for each approach. Moreira-Alvarez et al.290 proposed an alternative to these methods where size calibration is carried out using the target NP itself, measured under different instrumental conditions, which varied sensitivity, without the need for the transport efficiency to be quantified. The method, which was described in detail and included a good discussion on LOD estimation, relies on the linear correlation between the normalised signal drop for Au+ standards with sensitivity change and the shift of the median values of the Gaussian curves fitted to the corresponding Au NP histograms obtained under the sensitivity conditions. The slopes of these curves, being specific for each NP size/mass analysed, are then used to calculate the volume, and hence other metrics, of NPs of unknown size if an assumption of sphericality is made. The NP sizes determined by this method were in good agreement with TEM size data. In addition, a single set of the calculated slopes was shown to be useable for a period of 8 months without the need for re-measurement and the size data from these experiments was consistent with that obtained using conventional transport efficiency measurements estimates made on the day of analysis. The relative approach of the method was also applicable to accurate sizing of NPs that had other molecules, such as proteins, intentionally conjugated or attached due to biological processes to the NP surface. Murphy et al. systematically compared three methods for the measurement of transport efficiency: the particle frequency (TEF), particle size (TES), and dynamic mass flow (DMF) methods.291 The TEF and TES methods provide a direct measure of transport efficiency but require a NP reference material of which few are available. The DMF method provides an indirect measure of transport efficiency, as it measures the mass difference between the amount of sample solution introduced to the instrument and the amount of effluent exiting the spray chamber and thus can provide an SI traceable measure of transport efficiency. The work used three different spray chamber types, Scott-type double pass, conical impact bead and baffled cyclonic spray chambers, either at cooled (2 to 10 °C) or ambient (19 to 21 °C) temperatures, and they were equipped with different nebulisers and fitted to different ICP-MS instruments with Au NPs as the analyte. The authors reported that at ambient spray chamber temperatures, the DMF method yielded systematically higher measures of transport efficiency than the TEF and TES methods regardless of nebuliser type, spray chamber type or ICP-MS platform. Better agreement was observed between the three measures of transport efficiency was achieved using cooled spray chambers at 2 °C, although the repeatability for the DMF method was poor. The deviation of particle number concentration, the % difference from the known value, ranged from −18 to +8% (TEF) and from −1 to −70% (DMF), whereas the deviation for particle size ranged from −3 to +1% (TES) and from −4 to +44% (DMF) across all conditions. The authors concluded that the biases observed for the DMF method were not fully understood for all studied conditions and that only the TEF and TES estimates yielded acceptable results across all use conditions. This paper prompted a Comment article292 from the lead author of the DMF method paper cited,293 stating that this approach is not intended for use with any set up or condition but, if used under specified optimal operating set up and conditions, the method is invaluable for applications where use of the TEF and TES methods to determine transport efficiency is constrained by their reliance on reference materials that are limited or unavailable. A further Comment paper from Murphy et al.294 acknowledged this and stated that “The intent of our study is not to condemn the dynamic mass flow (DMF) method but rather to explore the use parameters under which it can and cannot be applied”. Each of the comment papers goes into further detail on the work covered and readers are urged to read all three papers in conjunction, and to look out for further Comment papers on this topic as the reply from Murphy et al. was published shortly before the cut-off date for papers used in this ASU. Torregrosa et al. have also been investigating transport efficiency with the aim of evaluating the role of aerosol transport on NP characterisation by sp-ICP-MS.295 To this end, 70 nm Pt NPs and ionic Pt (10 ng mL−1) were analysed under differing instrumental operating conditions with variation of both the sampling depth and nebuliser gas flow. The plasma and tertiary aerosol characteristics (i.e., drop size distribution and transport rate of solvent, Pt NPs and Pt ions) were also determined in order to explain the experimental findings. Unsurprisingly, the results showed that the number of particle events, Pt NP event intensity and Pt ionic signal depend on both aerosol transport and plasma operating conditions. The tertiary aerosol characterisation revealed that NP, ionic and solvent transport efficiencies differ significantly, which is problematic as the sensitivity calibration in sNP-ICP-MS is usually undertaken with ionic standards. The transport efficiencies followed the order solvent > ionic Pt > Pt NPs irrespective of the nebulisation conditions. For instance, when operating a nebuliser gas flow of 0.9 L min−1 and a sample uptake rate of 300 μL min−1, transport efficiency values were 5.77 ± 0.03, 3.89 ± 0.12 and 3.35 ± 0.06%, for solvent, ionic Pt and Pt NPs, respectively. Similar results were observed for other NPs, 50 or 150 nm Au NPs and spray chamber designs (Scott double pass or cyclonic spray chambers). These findings are of fundamental importance for sp-ICP-MS metrology, since some strategies for evaluating transport efficiency are based on assuming that the above-mentioned species are transported similarly into the plasma. The authors concluded that the particle frequency approach for determining transport efficiency seems to be the best approach as it is based on actual NP transport into the plasma. All six of the papers covered in this paragraph contain a wealth of information and detail and are recommended reading for those involved in this field.

Three papers report advances in data processing for sp-ICP-MS this year. The first of these covered here aimed to improve detection thresholds, which is often a challenge when small sized particles and an elevated ionic signal co-exist, and robust event filtering in both single-particle and single-cell ICP-MS analyses.296 The main focus of the work was the proposal of a modular workflow, involving Gaussian and Poisson distributions that describe the background signal, which could improve data analysis in single cell and sp-ICP-MS through separation of the tasks required to achieve this. Outlier analysis was also separated from the subsequent event detection and a data-dependent decision criterion based on Gaussian and Poisson modelling was developed to effectively address extra-Poisson variance in experimental data. In addition, a gate filter, based on peak height, was developed to reduce the amount of excess false-positive events in single cell ICP-MS. Two different approaches, based on a numerical approximation via the LOD, and critical values of the Gaussian and Poisson distribution were evaluated to calculate the gate filter level. The statistical arguments for the basis of these approaches are well-presented in the paper and the authors make the not uncommon conclusion that single particle and single-cell ICP-MS data should be treated with special care when approaching the detection limit. In sp-ICP-MS, small NPs produce low-count signals that are below the LOD and are therefore not included in the subsequent data processing stage, which can mean that the particle number concentration is underestimated and the mean particle mass, and hence the particle diameter, is overestimated. This can also occur when high ionic background signals are present. Suzuki et al. have addressed this by proposing a Bayesian estimation to deconvolute sp-ICP-MS data with a mixed Poisson distribution.297 In this work, Ag and SiO2 NPs were the target analytes and the signal distributions obtained were parameterised on the assumption that they could be described by mixed Poisson distributions. The results were then compared with results obtained with the commonly used LOD estimation of the blank signal plus 3σ of the blank and the dwell time was set so that a particle-event duration was obtained in 2 readings. The results obtained with the Bayesian estimation method were stated to be better than those for data based on the conventional LOD estimation, particularly for samples with high particle number concentration. Reasonable results were also obtainedfor analyses when the signal counts were 6 or less and the background counts were high. The authors point out the limitations of the approach, such as multiple NP coincidence during one dwell time period and multi-diameter particle system, and that the model needs extending to account for these occasions and that Bayesian algorithms that can detect particle events that span more than one reading, as is the case when μs dwell time are used, also need to be developed. The detection of co-incident small particles in sp-ICP-MS can lead to biased particle number concentration and particle size distribution estimates. Bevers et al. reported the use of analysis of serial dilutions of the same sample, to identify distortion-free segments of the particle size distribution, which are then combined and modelled using a power law to mitigate these effects.298 The relevance of the parameters derived from this modelling were then demonstrated using suspensions of two different metal tagged nanoplastics and then applied to the analysis of Al-bearing colloids sampled during a storm event. The paper and the supplementary information contain a full description of the theory and process involved in the work and deserve careful reading to gain an in depth understanding of this.

The modification of instrumental settings or hardware is sometimes undertaken to achieve a specific analytical goal, and this year is no exception. Schardt et al. modified a commercially available ICP-MS instrument such that data could be acquired with very short dwell times.299 An in-house built data acquisition system with nanosecond time resolution, named nanoDAQ, was fitted to continuously sample the electron multiplier detector signal at a resolution of approximately 4 ns. The data processing method is based on determining the temporal distance between detector events that is denoted as event gap (EG) and it was found that the inverse log of EG is proportional to the particle size and that the number of detector events corresponding to a particle signal distribution could be used to quantify the particle number concentration of a Au NP suspension with a run time of 60 s. At the current stage of development, the data processing method provides average information on complete data sets only and further work is required to enable particle-by-particle analysis with future hardware/software revision. The approach enabled the detection of Au NPs down to 7.5 nm in size. With sequential quadrupole-based ICP-MS instruments it is generally not possible to use an internal standard in sp-ICP-MS without degrading the quality of the data acquired. One way to overcome this is to reduce the resolution of the quadrupole mass filter, by modifying the standard RF and DC voltages applied, such that more than 1 m/z is transmissible. This does of course mean the signals from each m/z transmitted are simultaneously detected. Bazo et al. used this method to allow both Au and Pt ions through the mass filter and used the signal from Pt NPs (70 nm ϕ) as an internal standard to correct for signal variations when analysing for Au NPs (100 nm ϕ).300 Due to the size of the NPs used, the detector cross-calibration factor was set daily, otherwise normal tuning procedures applied. Data processing was by an in-house developed script, and the two different metal NPs were differentiated by plotting the count frequency against signal intensity, which was significantly different due to the differing Au and Pt particle sizes, first ionisation potentials and isotopic abundances. The approach was successfully applied to Au NPs suspended in three different matrices: high-purity water, aqueous 5 g L−1 NaCl and aqueous and 2.5% (m/v) TMAH/0.1% Triton X-100, all diluted for analysis into 1 mmol L−1 trisodium citrate. The authors noted that this approach might not readily transfer to other NP systems due to the potential for other elemental interferences between those of the two target analytes, and it is difficult to see how it could be applied to natural samples where matrix effects are likely to be of more concern due to the sample types involved, which could easily have overlapping particle size distributions. For these types of analysis, a standard-sample bracketing approach could be a better option to account for instrumental instability and matrix effects.

There are two reports on the use of different sample introduction hardware for sp-ICP-MS work in this review period. In the first of these, the development of a 3D printed single-pass spray chamber, capable of being fitted with a microconcentric nebuliser and with an additional sheath gas flow to facilitate the transport of larger droplets or particles, was described.301 The system geometry was optimised using computerised numerical simulations whilst the characteristics of the produced aerosol and operational conditions were studied via optical particle counting and sp-ICP-MS measurements, and analysis of cell suspensions. In a comparison of the performance of the new system and the standard instrument fitment (quartz microconcentric nebuliser plus a double-pass spray chamber), it was found that the new device gave a four-fold improvement in particle detection efficiency (presumably this is an improvement in sample transport efficiency) and, due to a larger s/n ratio, a 20% reduction in the lower NPsize LOD. In addition, it enabled the extension of the upper limit of transportable particle diameters to about 25 μm. From a plot of signal vs. particle mass for Ag NPs, the sensitivity of the new system is also about double that of the standard system fitted to the instrument. Membrane desolvation devices are often used in conventional ICP-MS to reduce plasma solvent load, and hence decrease the potential for polyatomic interference production whilst generally improving sensitivity. However, it is unusual to see their use reported for analyses involving NPs. So, it is good to see a paper describing their use, for the characterisation and quantification of TiO2 NPs in food simulants, published. Data were acquired using a TQ ICP-MS instrument operated in mass shift mode, with an O2 + H2 mixture as the reaction gas, to negate the isobaric 48Ca interference on the major Ti isotope. For analysis of suspensions of large (>300 nm ϕ) TiO2 NPs, the detector was operated in analogue mode. The performance of two membrane desolvators, the ESI APEX™ Q and APEX™ Omega, was compared with the ICP-MS manufacturer fitment sample introduction system in terms of sensitivity and trueness for the detection of TiO2 NPs in various organic food simulants (5, 10, 20 and 50% v/v EtOH and 3% v/v acetic acid). The best results were obtained using the APEX™ Omega, with a significant reduction in LODsize compared with that obtained with the standard sample introduction system and the particle sizes determined were in statistical agreement regardless of the sample suspension matrix.

4.8.4. ICP-TOF-MS analysis. The quasi-simultaneous ability of ICP-TOF-MS for multi-isotope detection with short data acquisition times opens up new possibilities for the analysis of nanomaterials. Of the primary analytical methods, isotope dilution analysis is probably the one most commonly applied by metrology institutes. Aramendía et al.302 reported the use of isotope dilution analysis to determine the mass of Ag NPs, hence NPsize, using sp-ICP-TOF-MS. In this work, ionic 109Ag+ was used as the spike material and added directly to Ag NP suspension and, as the spike/NP signals are transient, a modification of the species unspecific isotope dilution mass spectrometry equation was used for the calculations required and the transport efficiency was determined by the DMF method and data were acquired in continuous acquisition mode with a maximum time resolution of 3 ms. The study involved optimisation of numerous parameters, including the read time, the composition and concentration of the spike and how to account for split particle events, to allow the optimal isotope ratio measurement precision to be obtained, and all of this is discussed in detail in the paper. The developed method was used to determine the average Ag NPsize in NIST RM8017 and the results obtained, 69.8 ± 2.7 nm, were in good statistical agreement with the certified values, 70.1 ± 6.0, 74.6 ± 3.8, 67.9 ± 0.8 and 105.6 ± 4.6, measured using AFM, TEM, USAXS and DLS, respectively. As isotope dilution analysis is suitable for estimating a complete uncertainty budget, it would be good to see a breakdown of the individual uncertainty contributions of each stage of the method for this work. The accurate quantification of microplastics, which are now globally dispersed due to human activities, can prove challenging and laborious with currently available techniques but there is now a growing body of work whereby ICP-MS is being utilised to achieve this. This also presents a challenge due to the large background signal from C in laboratory reagents or inherently present in natural samples. Two papers report the use of sp-ICP-TOF-MS, with 12C-based detections, for quantifying microplastics in various sample types. In the first of these, Hendriks and Mitrano reported the optimisation process required to allow the quantification of 4 μm polystyrene beads spiked into waters containing up to 20 mg L−1 dissolved organic carbon, or mixtures of algae cells and microplastics, at ratios of 1[thin space (1/6-em)]:[thin space (1/6-em)]9 and 5[thin space (1/6-em)]:[thin space (1/6-em)]9, respectively.303 For the latter, where signals for 31P and 24Mg coincided with 13C signals in the time spectrum, these events were ascribed as arising from algal cells, whilst 13C-only events were attributed to microplastics. For the spiked simulated water sample suspensions, constant particle number concentrations were found in all solutions, indicating that the microplastics created a pulse intensity that was not masked by the background dissolved organic carbon concentration and the microplastic transport efficiency was not affected with increasing dissolved organic carbon content. The authors noted that whilst the type of microplastic cannot be determined by this technique, nor yet nano-sized particles, the method offers a rapid screening possibility for laboratory-based ecotoxicological experiments. The second paper, from Harycki and Gundlach-Graham, to use the same type of platform reports its use for NPs, submicron particles and microplastics in seawater.113 An on-line microdroplet system was used for sample introduction and it was shown that one multi-elemental standard could be used for calibration for a wide range of target analytes, metal NPs, polystyrene microplastic beads doped with rare-earth elements and metal-oxide sub-micron particles in artificial seawater, independent of the sample matrix. The results obtained demonstrated mass recoveries of 90–98% for Au NPs in high-purity water and 99% in seawater. For food-grade TiO2 submicron particles, an accurate mass of Ti per particle was achieved even with a seawater matrix causing a signal attenuation of up to 80%. Accurate particle diameter measurements were also obtained for 3.4 μm polystyrene beads suspended in up to 80% simulated seawater. It was also possible to accurately quantify REEs in the doped particles up to a seawater concentration of 33% but above this point the loss of sensitivity hampered the analysis such that a bias towards larger particles was observed. All three of these papers give a good account of the theory of ICP-TOF-MS along with detailed discussions, including those steps required for data processing. Hendriks et al. developed a single cell-ICP-TOF-MS method for the quantification of metal-doped model nanoplastics in human cells.304 In a proof-of-concept study, two different human cell lines relevant to inhalation exposures (A549 alveolar epithelial cells and THP-1 monocytes) were exposed to Pd-doped nanoplastics. The single-cell ICP-TOF-MS analysis revealed a similar dose-dependent endocytotic capacity of THP-1 and A549 cells for nanoplastics uptake, and particle internalisation was confirmed by TEM. Single-cell quantification showed that a large proportion of the exposed cells (72% of THP-1; 67% of A549) did not associate with any nanoplastics after exposure (50 μg L−1) for 24 h.

Continuing on the theme of ICP-TOF-MS for multi-isotope detection Lockwood et al. described two strategies to enhance figures of merit in ICP-TOF-MS.305 First, instead of analysing full mass spectra, the mass range focussed on that between 152Gd and 176Yb, with blanking and equilibration regions before the lower mass and a second blanking region after the upper mass, being excluded through the use of a Bradbury–Nielsen gate. The resulting restricted mass range was acquired up to 5 times faster thus increasing duty cycles and sensitivity. Secondly, isotopes of poly-isotopic elements recorded simultaneously were accumulated to increase s/n ratios. As proof of concept, the approach was applied with sp-ICP-TOF-MS to characterise the up-conversion nanoparticles that contained Gd and Yb. Both signal amplification strategies were combined and s/n ratios and LOD values were compared with those obtained when the instrument was operated with detection of the full mass range. Sensitivities were increased up to 27-fold when accumulating all Gd and Yb isotopes at 177 kHz, and NPsize LOD values decreased by a factor of approximately 3. The improved figures of merit promoted more accurate investigations of up-conversion nanoparticles, which were characterised regarding size distributions and composition. The approach was also applied using LA-ICP-TOF-MS, with Mo and Se as the analytes, of rat brain tissue. Again, the improved s/n ratios enabled the mapping of both elements whilst also acquiring data for other elements such as Fe and Zn. Manard et al. described an automated, high-throughput system for the quantitative sizing and isotopic analysis of Au and Au/Ag core shell NPs by sp-ICP-TOF-MS.306 A microFAST SC sample introduction system, which incorporates the ability to stir sample and is designed for single-cell work, was used for this work and isotopic ratio measurements were also made by MC-ICP-MS. After optimisation, the transport efficiency was >80% and 50 samples (including blanks/standards) were analysed over an 8 h period and operated for 5 days to assess within run and longer-term repeatability, which were found to be 3.5 and 9.5% as RSD, respectively. The measured Au NPsize and particle number concentration showed <5% relative difference from the certified values over these two time periods. The 107Ag[thin space (1/6-em)]:[thin space (1/6-em)]109Ag isotope ratio in the Ag particles (n = 132[thin space (1/6-em)]630) over the course of the measurements was determined to be 1.0788 ± 0.0030, which was a 0.23% relative difference when compared with the MC-ICP-MS data.

Analysis for NPs using ICP-TOF-MS can generate very large data sets and advances in data processing for this technique are also reported this year. It has been reported that sp-ICP-TOF-MS data can show a systematic bias in the detected elemental compositions of particles as a function of particle size, composition, and analytical sensitivity. To overcome this bias for the classification of NP types, a multi-stage semi-supervised machine learning (SSML) strategy was developed.307 The strategy first finds the systematic particle mis-classifications, which are then incorporated into the SSML model for the development of a second, more robust classification model. In a case study, CeO2, ferrocerium mischmetal, and bastnaesite mineral NPs were used to represent engineered, incidental and natural particle types, and particles were classified in mixed samples based on the final SSML model. False-positive rates of 0.030, 0.001 and 0 for engineered, incidental and natural nanoparticles, respectively, were found. These low rates allowed for accurate particle-type classification of the NP mixtures with varied particle number concentrations. Gundlach-Graham et al. developed and tested a set of sp-ICP-TOF-MS data analysis programs, named “time-of-flight single particle investigator” or “TOF-SPI” and full details are given in a comprehensive technical note.308 The programs, written in LabVIEW for the direct analysis of HDF5 data files from the TOFWERK icpTOF instrument, are now available for use as a Windows executable program. The software can provide accurate single-particle finding, split-event correction, quantification of number concentrations, quantification of element mass amounts per particle, generate user-readable output reports and performing batch analyses of data calibrated with either the particle-size method or on-line microdroplet calibration. Data analysis using clustering approaches has been applied to extract elemental fingerprints from multi-element nanoparticle data obtained by sp-ICP-TOF-MS.309 Hierarchical clustering, spectral clustering and t-distributed stochastic neighbour embedding coupled with density-based spatial clustering of applications with noise (tSNE-DBSCAN) were the methods used in this work with the performance evaluated by comparing the size of the extracted clusters and the similarity of the elemental composition of nanoparticles within each cluster. As hierarchical clustering is sensitive to outliers this method often failed to achieve an optimal clustering solution for sp-ICP-TOF-MS, whilst spectral clustering and tSNE-DBSCAN extracted clusters that were not identified by hierarchical clustering. It was found that although all three clustering approaches successfully extract useful information from sp-ICP-TOF-MS data, spectral clustering outperformed hierarchical clustering and tSNE-DBSCAN by generating clusters of a large number of nanoparticles with similar elemental compositions.

4.8.5. Sample extraction. As is the case with elemental speciation, where species information must not be modified for accurate results, extracting NPs from a sample without modification of NPsize, the particle number concentration or particle mass concentration can be challenging. To this end, Sajnóg et al. conducted a comprehensive study of ten possible extraction procedures for Se NPs in selenised yeast, S. cerevisiae.310 The extraction procedures studied were: enzymatic with protease, mechanical cell lysis by vigorous shaking with glass or metal beads or sand, and chemical with various combinations of SDS, NaOH and TMAH. Analysis of the extracts was by sp-ICP-MS, with H2 as the cell gas to allow 80Se to be monitored with a 100 μs dwell time. The results obtained showed that the enzymatic procedure provided the highest recoveries, i.e., the mass of SeNPs extracted, for 3 out of 4 studied samples, although broader size distributions were obtained compared with the other methods, which the authors attribute to partial agglomeration due to surface modification of the NPs by protease. The mechanical procedure used was faster and provided more realistic size distributions, but extraction efficiency was lower in terms of mass recovery. The chemical procedures studied did not provide satisfactory results due to the partial dissolution of the Se NPs in the sample. In another study, three different procedures, with high-purity water, TMAH or tetrasodium pyrophosphate, were evaluated for the extraction of Ag species from faeces of animals fed with Ag-based nanomaterials.311 For extractions with TMAH, both cysteine with Triton-X100 and CaCl2 were evaluated for their stabilisation effect on the Ag species in the samples. Samples were analysed using sp-ICP-MS, hydrodynamic chromatography coupled with ICP-MS and AF4-ICP-MS. The data obtained showed that extraction with water was inefficient and unsuitable for further use, extraction with TMAH, cysteine and Triton X-100 followed by a sedimentation procedure gave the highest total Ag recoveries (ca. 40% of total Ag present), and that although the tetrasodium pyrophosphate method gave lower recoveries (<30%) than the TMAH method, it proved useful in allowing the different Ag species in the sample, e.g., AgCl and Ag2S NPs, to be qualitatively determined. It was also found that the sample matrix affected Ag NP spike recoveries compared with those from controls, 40 and 65 to 78%, respectively, showing the limitations of using spike recoveries for method validation purposes. The paper also contained a good discussion of the possible mechanisms for ionic Ag transformations to NPs during extraction. A surfactant-assisted DLLME approach for the selective extraction of Ag and TiO2 NPs from tap water, with analysis using sp-ICP-MS, was reported this year.312 The optimal pre-concentration conditions were found to be 10 mL of sample, 1.0 mL of the Triton X114/1,2-dichloroethane (49[thin space (1/6-em)]:[thin space (1/6-em)]1), vortex mixing followed by centrifugation and dilution of the collected organic phase with 1% (w/v) glycerol, which gave an enrichment factor of between 5- and 10-fold. The LODsize (5 sigma criteria) was 66 and 17 nm for TiO2 and Ag NPs, respectively. The LODPNC was 6 × 106 L−1 for Ag NPs and 2.9 × 106 L−1 for TiO2 NPs, whereas the limits of detection for dissolved Ag and Ti mass concentrations were 0.435 and 9.82 ng L−1, respectively. Recoveries for the two forms of NPs studied ranged between 88 and 114% for spiked extracts and 44 to 58% for spiked tap water subjected to the entire DLLME procedure. No agglomeration was observed for spiked Ag NPs, but did occur for TiO2 NPs, as shown by the increase in the NPsize range detected compared with the raw material.
4.8.6. Separations. In sp-ICP-MS, some form of separation is often used, either to distinguish between co-extracted ionic species and the target NP or to remove interfering species that affect LOD estimates. This latter aspect of separation science has been reported on this year, whereby 3D-printed polymers have been fabricated to selectively remove Ag+ ions from samples to enable an improvement in the accuracy of sp-ICP-MS of Ag NPs.313 The porous 3D scavengers used SiliaBond® tosic acid on a polystyrene support and the macroporous structure allowed Ag NPs to pass through without affecting their original properties. The scavengers were fitted into a 10 mL syringe and the samples (100 mL) passed through with dissolved Ag+ ions being efficiently removed (>98%). The retained dissolved silver could be eluted with a 0.5 mmol L−1 solution of sodium thiosulfate with recoveries of >99%. The competitive adsorption of other elements commonly found in natural waters (e.g., Ca, K, Mg, Na, S, Si and Sr) did not affect the Ag extraction efficiency. The developed procedure was then applied to the determination of 30 nm Ag NPs spiked into high-purity and clear environmental waters, which also contained dissolved Ag+ (0.2 μg kg−1), followed by sp-ICP-MS analysis. Measurements of the samples prior to any Ag+ extraction exhibited a bias in NP sizing (up to +12%) and particle number concentration (down to −51%), whilst the use of the scavengers eliminated the interfering effect of dissolved Ag+. This resulted in a mean Ag NPsize of 30 ± 1 nm and a particle number concentration recovery of 87%. The scavenger units were washed with water between use, after elution of the retained Ag+, and could be reused at least five times. In the same vein, Cen et al. evaluated three different approaches to reduce metal salts in aerosol samples by sp-ICP-MS.314 Firstly, a rotating disk diluter was used for the on-line dilution of the aerosol, while a differential mobility analyser and a centrifugal particle mass analyser were used separately for the on-line fractionation of specific-sized NPs in the aerosol. The results from the analysis of 100 nm Au NPs mixed with Au3+ (1[thin space (1/6-em)]:[thin space (1/6-em)]25 mass ratio) showed that the LODsize decreased from 78 nm for direct analysis of the aerosol to 61, 50 and 33 nm by using the rotating disk diluter, centrifugal particle mass analyser and differential mobility analyser devices, respectively. It was also found that the separation performance was in the order of differential mass analyser > rotating disk analyser > centrifugal particle mass analyser. An HPLC-ICP-MS method was reported for the separation of Ag NPs and Ag+ ions in surface water and algal cells. Different compositions of the mobile phase, which included surfactants to minimise agglomeration or aggregation of the AgNPs, pH buffers and Ag+ complexing agents, were investigated to ensure complete elution and stability of the silver species. The optimal conditions were: a mobile phase of 10 mmol L−1 sodium dodecyl sulfate, 2 mmol L−1 citrate buffer and 2 mmol L−1 tiopronin, flowing through a C18 then a polystyrene/divinylbenzene column at a flow rate of a 0.5 mL min−1 for 7 minutes. The LOD values of 2.0 ng L−1 for Ag+, 3.1 ng L−1 for 10 nm Ag NPs, and 2.2 ng L−1 for 30 nm AgNPs were reported. A linear relationship was found between the retention time of AgNPs and logarithm of particle diameter up to 40 nm. The HPLC-ICP-MS method was applied to study interactions of Ag+ and citrate-stabilised Ag NPs (10 nm and 30 nm ϕ) with a green microalga, Acutodesmus obliquus in river water. The Ag biosorption was found to be in the range of 77.2–82.7% and, due to interactions with the biomass, biosynthesis of small nanoparticles (with a size of 6 nm) from Ag+ ions and partial dissolution of both the 10 and 30 nm AgNPs was observed. Paton et al. took advantage of the simultaneous multi-element capabilities of TOF-MS to investigate the nature of HgSe NPs formed by cetaceans.315 Liver tissues from stranded sperm whale were extracted with either proteinase K or water, with both using SDS. The enzymatic extraction was used for subsequent experiments as a greater number of NPs were extracted using this method. Sample analysis was firstly by AF4-MALS-ICP-TOF-MS to identify HgSe NPs, along with a bootstrap statistical analysis of the data, and then by sp-ICP-TOF-MS to confirm the association of Hg with Se on a single-particle basis. Other elements, Cd and Sn, were also detected on particles alongside the Hg and Se. The authors state this indicated that the detoxification process resulting in Hg/Se nanoparticles may not be specific to MeHg and that this data indicates nanoclusters of toxic elements, bound to selenium, make up a nanoparticle core that is surrounded by a larger non-metal(loid) corona. The paper gives a good description of the development of the methodology, as well as some insights into biotic detoxification processes and is well worth reading. The final paper covered in this section used Taylor dispersion analysis in capillaries hyphenated to ICP-MS to characterise NP mixtures.316 Experiments were conducted with carboxylated magnetite (Fe3O4@COOH) and Ag NPs in the presence of Au NPs under various conditions, and realised data for the hydrodynamic diameter, size distribution, concentration, elemental composition, isotope ratio and behaviour in the presence of other NPs. The authors concluded that there is extraordinary potential in the method, which has a wide applicability for addressing numerous questions in the fields of nanomaterial chemistry, nanotoxicology and medicine, giving examples such as studying self-assembly processes and metal-based nanorobots freely moving in living organisms or the environment.
4.8.7. Laser ablation studies. One of the challenges with using LA to mobilise NPs from solid samples is that the NPs present must not be disintegrated by the LA process. Yamashita and Hirata have evaluated the fluence level below which Ag and Au NPs, ranging in diameter from 20 to 60 nm, adsorbed onto filter paper remained intact during LA-ICP-MS.317 A Nd:YAG laser was used in the study and operated with fluences from 0.2 to 3.0 J cm−2 and a beam size of 5 μm. Disintegration of both NP types was observed when the laser fluence was greater than 1.0 J cm−2, whereas no disintegration was observed when the fluence was lower than this value. When compared with solution-based sp-ICP-MS and TEM data the mean NPsize and its associated SD obtained by LA-sp-ICP-MS were in good agreement within analytical uncertainty. In addition, plots of signal intensity vs. NPsize were linear, suggesting that NPsize calibration is possible by the approach used. The authors also concluded that, as biological materials are also successfully ablated with fluences of 1 J cm−2 or lower, that the methodology is suitable for detecting and quantifying NPs in these materials. Seiffert et al. have undertaken such a study, determining the size and localisation of multi-shaped NPs in tissue sections by LA-sp-ICP-MS.318 A 193 nm excimer laser, with the fluence set to 0.24 J cm−2, was used in this work and initial studies involving Ag NPs suspended in gelatine and aqueous standards, LA-sp-ICP-MS, solution sp-ICP-MS and TEM, gave data for each method that was comparable, with an LODsize of 10 nm. For CeO2 NPs the same statistical agreement was found for each analytical method with, as these NPs are roughly cubic, the obtained Feret diameter being compared. Finally, the developed methodology was applied for the detection of CeO2 NPs in spleen sections. Different mean particle sizes were detected, which depended on the time taken for the NPs to reach the spleen after inhalation, but the overall mean was in agreement with that for the same CeO2 NPs when suspended in gelatine. Luo et al. have used a different approach for producing NP calibration standards for use with LA-sp-ICP-MS.319 The procedure involved coating a nano-flat indium tin oxide slide with a polymethyl methacrylate (PMMA) layer followed fabrication of a microhole array in this layer and the slide, spreading and drying of an Au NP suspension on the PMMA layer and removal of the PMMA layer to leave the Au NP suspension residue distributed in the microholes on the slide. Slides were prepared with either Au rods (length 99 nm, diameter 34 nm) or spherical Au NPs of 100 or 160 nm ϕ. Calibration with ionic Au solutions was also performed. The results obtained showed that Au ions and AuNPs have different signal transduction efficiencies in LA-ICP-MS and the authors concluded that NP quantification with LA-ICP-MS should not be performed using calibration with ionic standards, demonstrating the necessity of developing particulate standards.
4.8.8. LIBS studies. Reports where LIBS is used for the detection of NPs are not common due to the challenges involved, but these are being addressed in recently published research. Cárdenas-Escudero et al. investigated the use of a chemically functionalised glass support for Ag and Au analysis with LIBS.320 Glass microscope slides were silanised with 3-(mercaptopropyl) trimethoxysilane followed by the deposition of 4 μL drops of Ag or Au NP suspensions in the concentration range of 0.6 to 3 μg mL−1. The μ-LIBS instrument consisted of a multimode 1064 nm IR laser source with a pulse duration of 8 ns, a maximum pulse energy of 2 mJ, and a repetition frequency of 100 Hz and operated with a pulse resolution of 35 μm with 31[thin space (1/6-em)]000 spectra acquired per sampling point. Using this methodology, calibration curve correlation statistics were improved, particularly for Ag NPs, when compared with data from non-silanised slides used as a substrate, which can be attributed to the obtained improved sensitivities of 330 and 5% for Ag and Au NPs, respectively. The authors suggested that this work opens the way for sensor-based applications involving LIBS, which may be the case for pure suspensions, but would be much more problematic to achieve for, say, biological tissue samples. Burgos-Palop et al. reported the use of ultrafast laser excitation to improve the performance of LIBS for the analysis of optically trapped NPs.321 A full description of the instrument used was given in cited literature in the paper, and the instrument uses three lasers, one each for sample ablation, optical trapping of individual micro- and nanoparticles produced and finally for producing the LIBS plasma. The NPs studied were Cu of 25, 50 and 70 nm ϕ, which were deposited as loose powders onto a glass slide and covered with quartz cuvettes of 10 mm path length and excitation pulses of nano- and pico-seconds were used. Particle signals were discriminated from the background on the basis of signal intensity. It was found, from plasma imaging, that ps-plasmas were more spatially confined than those from ns pulse duration and these were used to provide qualitative data for the NPs under study, which led to a mass detection limit of 27 ag, equivalent to Cu NPs of 18 nm ϕ. Liu et al.322 investigated the use of a photophoretic optical trap to improve repeatability in LIBs analysis using carbon black particles. A Q-switched Nd-YAG laser, pulse duration 8 ns, and beam size of 7 mm was used in the study and full details of the instrumental set up are given in the paper. In this work, the photophoretic force required was theoretically modelled and the size selectivity in a Gaussian trapping beam was analysed followed by statistically analysis of the NPsize examined under different powers of the trapping Gaussian beam. The results obtained showed that, in accordance with theory, as laser power increases, larger particles are trapped. When comparing the particle size distribution of hollow beam trapping with that of Gaussian beam, it was found to be broader in the former, and the authors stated the repeatability was improved. However, this reader could not find any data in the paper to confirm this statement.

Table 3 shows other applications of nanomaterial characterisation and/or detection presented in the literature during the time period covered by this ASU.

Table 3 Applications of nanomaterial characterisation and/or detection
Analyte Technique Comments Reference
Pd nanowires GI-SAXS, XRF Study of the structural evolution of electrochemically fabricated Pd nanowires. Competition and interactions between electrodeposition and the hydrogen evolution reaction observed. 323
MWCNTs DBD-OES, XPS Study into amination of MWCNTS. Various N-based functional groups detected. 324
Ag NPs Hollow fibre FFF-ICP-MS Study of the sulfidation of Ag NPs in aqueous suspension. Particle coating dictated stability of NP size distribution. 325
Zr-based metal organic framework (MOF) XRD, FI-ICP-MS, AF4-ICP-MS In depth physicochemical investigation during production of cisplatin-loaded core–shell UiO66/CdMO oligomer MOFs. 326
Gelatin-coated Au:Si core–shell NPs TOF-SIMS, LC-ESI-MS/MS, sp-ICP-MS, XPS Study into gelatin biodegradation on the surface of Au:Si core–shell NPs by Alteromonas macleodii. Confirmed the presence of and activity of extracellular proteases. 327
Nd2O3, Pd, Y2O3 NPs sp-ICP-MS Study into particle-generated spectral interferences in sp-ICP-MS. YO+ interferences mitigated with CCT and NH3 as the cell gas. 328
Si NPs doped with tris(2,2′-bipyridyl)ruthenium(II) sp-ICP-MS Study into dopant distribution through Si NPs. Average concentrations of dopant on NPs agreed with bulk analysis but was variable on a particle-by-particle basis. 329
NbC NPs in steels AF4-ICP-MS Method development for the analysis of NbC precipitates in steels. Found that long-duration heat treatment increased the number of nm-sized NbC particles. 330
Graphene nanoflakes XAS Study into the stability of iron single atoms in graphene structures. 331
Pd–Pt NP clusters sp-ICP-MS Study into Pt–Pd NP cluster formation during synthesis by gas diffusion electro-crystallisation. 332
Si nanowires SIMS Development of SIMS method for the quantification of B distribution in silicon nanowires. 333
Ag–Co, Ag–Au, Au, Cu, Sn–Cu NPs LIBS The detection and characterisation of mono- and bimetallic NPs produced by electrical discharge plasma generators. 334
Ag NPs FAAS Separation and analysis of Ag+ and Ag NPs by LLE and FAAS to show that cost-effective methods can be used for this purpose. 335
C, Cu, CuFe2O4, NPs LIBS Study into the diffusion dynamics and characterisation of optically trapped NPs. 336
LiNbO3:Au core–shell NPs XPS, TOF-SIMS Study into the growth of an Au raspberry shell on LiNbO3 NPs. 337
Coated Ag and Au NPs FFF and sp-ICP-MS Characterisation of variously coated Ag and Au NPs after separation by EFFF. 338
Polymer-coated Si NPs TOF-SIMS The depth profiling of NP coatings. 339
Ag+ and Ag NPs sp-ICP-MS Development of an imprinted magnetic adsorbent for the separation of Ag+ and Ag NPs. 340


5. Forensic analysis

The most popular area of forensic analysis over this review period has been gunshot residue. The use of a portable LIBS analyser that can be taken to crime scenes and can give data within minutes rather than hours is a notable development. In last year's update there were several references that described the analysis of drugs. This has not been the case during this review period, with no drug papers at all.

There have been two reviews in the general area of forensic science published during this review period. One presented by Kobylarz et al.341 contained 31 references and gave an overview of the use of portable XRF instruments in forensic science. Portable XRF ticks most boxes required for successful forensic analysis in that it is rapid, non-destructive, can be used on relatively small samples, may be used in situ and is cost effective. However, it is known that the in-built calibration algorithms are not suitable for all sample types and hence the analyses are often regarded as being only semi-quantitative. The review was split into sections covering an introduction, counterfeit medicines/pharmaceuticals, the analysis of bones, banknotes, forensic geoscience, glass, paint and gunshot residue. Conclusions and future perspectives were also presented. Many of the applications were conveniently tabulated for easy reference.

The other review of general forensic science rather than a specific sample type was presented by Jackson and Barkett.342 These authors provided a review entitled “Forensic Mass Spectrometry: Scientific and Legal Precedents”. The review contained 222 references, most of which focussed on the use of organic mass spectrometry. Section titles included drugs, arson, gunshot residue and explosives, trace fibres and hair and glass. Although the majority of references were organic mass spectrometry, i.e., out of the scope of this update, several did cover atomic spectrometric techniques.

Several other reviews more specific to a certain sample type were also produced. These will be discussed in the relevant sections below.

5.1. Forensic applications of glass analysis

Two papers of note are suitable for this sub-section. Forensic samples obviously do not necessarily conform to the ideal sample shape and size for analysis. Therefore, being able to analyse tiny fragments of glass would be a clear advantage for a technique. Polychromatic X-ray analysis can lead to Bremsstrahlung scattering and noise, which can affect the detection of analytes. A paper by Wang et al.343 described an XRF instrument capable of both micro-analysis and monochromatic X-ray analysis. The authors termed it a Mμ-XRF instrument. A detailed description of the instrument was given in the paper. Briefly, it comprised a low-power molybdenum target X-ray tube, two polycapillary X-ray lenses, a flat crystal and a Si drift X-ray detector. The maximum power of the X-ray tube was 5.3 W, the maximum voltage was 90 kV, and the focal spot size of the X-ray source was 8 μm. The two lenses were a polycapillary parallel X-ray lens (PPXL) and a polycapillary half-focusing X-ray lens (PHFXL). Ten types of glass fragments (including a flash-light glass cover, tempered glass, mobile phone screen glass layer, cosmetic glass bottle, flat mirror, glass slide and four common beer bottles) were analysed for a period of 900 s per sample. No chemometric analysis of the data was undertaken, but the authors concluded that differentiation between sample types was possible. With its two modes of analysis, the instrument could analyse glass particles with a size of tens to hundreds of μm using PHFXL and up to mm-sized fragments when the small aperture was used. The authors hoped that a modified version of their instrument could be used as a portable device.

The second paper was presented by Tran et al.159 These authors analysed a series of automotive glasses using neutron activation analysis (NAA) and PIXE and then inserted the analytical data from 29 elements into PCA and an agglomerative hierarchical clustering method. Grouping of the materials according to car brand could only be achieved using NAA, with the rare earth elements contributing highly to the differentiation. However, the major analytes showed a great deal of clustering and could be used for differentiation between different models of cars.

5.2. Forensic analysis of organic materials

5.2.1. Skin and fingerprints. In a fascinating application, Tambuzzi et al.344 described how portable XRF could be used on skin samples and on adhesive graphite tape to assess electric marks. A man's body had been found on electric railway lines and there were clear areas of scorching indicating electrocution. Seven small skin fragments (1 cm × 1 cm × 0.5 cm) were taken at autopsy (using a ceramic knife) and then portable XRF was used to determine trace metals in these scorch marks to determine whether or not their composition agreed with that of the electrified rails. As an alternative method of sampling, adhesive graphite tape was also used to pick up any metal fragments from the skin. Tape samples were used to “swab” un-scorched areas of skin and these acted as blank samples. Traces of Ca, Cu, Fe, Ti and Zn were detected in the negative control but at generally lower concentrations than in positive samples. Clear increases in Fe were observed in most of the skin samples that exhibited scorch marks. Some samples also demonstrated increased levels of other elements, e.g., Cu, Mn, Ti and Zn. The authors then determined the ratio of Fe to Rh signals, because Rh is the anode material of the X-ray tube and therefore gives a constant signal for all samples. The negative control gave a Fe/Rh of 0.2. Some areas of skin reached ratios of 7–9.6, indicating large quantities of Fe. The ratio was not as high for the graphite tape but still reached levels of 0.4–1 for many samples.

The analysis of fingerprints using TOF-SIMS is not a new method. However, two studies have contributed to the advancement of knowledge in the area. In one, by Charlton et al.,345 the technique was used to visualise fingerprints deposited on polyethylene or stainless steel. Ten donors placed fingerprints on the materials, which were then aged for up to five months in ambient air, in rainwater or underground. A total of 120 fingerprints were then treated prior to analysis. The fingerprints aged in ambient air were treated with cyanoacrylate fuming and Basic Yellow 40 staining. Those fingerprints treated with rainwater or kept underground were treated using black powder suspension. After treatment, the fingerprints were analysed, with each analysis taking 36 minutes and 48 seconds to complete. The best images were obtained using 10 or more ions, but some samples could readily be visualised using just one ion. Both positive and negative SIMS spectra were obtained so that a comparison could be made. A fingerprint expert then inspected the fingerprints and stated that the TOF-DIMS method provided a significant improvement in clarity in over 50% of the samples compared with the standard methods. The TOF-SIMS was at least equivalent to the other methods in over 90% of cases, irrespective of the aging time or process. There were advantages in obtaining samples in both positive and negative modes. The authors admitted that the instruments are expensive and therefore not available to all police forces. The time taken to obtain a sample is also comparatively long, meaning that even if a police force has such an instrument, they are likely to use it as a last resort. The fact that fingerprints can be obtained after five months after all of the aging processes is impressive.

The second paper to analyse fingerprints was presented by Liu et al.346 This study used transfer films in an attempt to obtain fingerprints from challenging substrates, i.e., ones that do not necessarily have a smooth surface. Preliminary experiments identifying transfer strips that yielded optimal results were undertaken. Once identified, they were used to transfer the fingerprints from objects that could not be analysed directly. Inevitably, the best results were obtained from samples that had a smooth surface, but other samples also yielded data. In some samples it was even possible to detect residues of morphine and the drug MDMA. It is therefore possible to not just visualise the fingerprint, but also to further characterise a suspect. It was concluded that this work exhibited the unique abilities of TOF-SIMS and demonstrated that its known applications could easily be expanded.

5.2.2. Paints. Only one paper of note has been published in this area during this review period. This was presented by Furukawa et al.,347 who used a series of instrumental analytical techniques, including FTIR, ICP-MS and headspace GC-MS, to obtain data before inputting it to a new chemometric method for classification. Various analytical data were therefore obtained, each with its own intensity and time axis variables that cannot be handled equally. The data were therefore normalised as an angle parameter using Arctangent. Once all of the data were in one useable form, it was input to a modified PCA program where multivariate analysis was undertaken. The modification was called PCA-Merge. This method merges the PCA data groups of each time-component axis into that of specific-component axes and distinguishes each sample by utilizing the shift in the barycenter of the PCA score plot. The proposed method enabled the simultaneous utilisation of various data groups that contain information about static and dynamic properties.
5.2.3. Polymers. Polymer fibres are a common material to be determined in forensic analysis. They can arise from clothes rugs/mats, cords, etc., and often contain metals that originate from polymerisation catalysts, matting agents, transesterification catalysts, etc. Komatsu et al.124 described a methodology that could discriminate single fibres of white polyester. Their methodology adopted X-ray absorption fine structure (XAFS) to determine the chemical state of the analytes, semi-microbeam XRF for spot analysis and nanobeam XRF for imaging. The semi-microbeam XRF identified Co, Ge, Mn, Sb and Ti to be present in the sample. X-ray intensity ratios of Ti/Mn, Ti/Ge, Ti/Sb and Mn/Sb gave good reproducibility. However, those involving Co, e.g., Ti/Co, Mn/Co and Co/Sb, did not. The results for the nanobeam-XRF explained this. The Co was not uniformly distributed and hence yielded irreproducible data. The XAFS data indicated three types of Ti compound and two types of Co compound. The combination of the different techniques yielded data that enabled discrimination of the single fibres.

de Bruin-Hoegee et al.123 emphasised the lack of suitable certified reference materials for ensuring that accurate data are obtained for polymer analysis. They devised a scheme by which suitable materials could be made for polyethylene, polystyrene and polyvinyl chloride standards. Every standard would be certified for 23 analytes at three different concentrations. The paper explained in detail how the three types of material were made. Analysis was achieved using XRF and LA-ICP-TOF-MS. The XRF could measure the bulk concentrations and the LA-ICP-TOF-MS could give an indication of the homogeneity. Unfortunately, the raw materials contained significant amounts of Cl, S and Zn and so these analytes were not certified. The calibration graphs obtained using LA-ICP-TOF-MS were >0.99 for virtually all analytes in all three sample matrices. The homogeneity of all sample types was regarded as being acceptable, although that of the polystyrene was the worst. When a series of polymer pieces (Jerry cans, tapes, electrical wiring, etc.) relevant to forensic science were analysed, the data from them were input to PCA where data reduction was performed and then into one-way ANOVA. The statistical treatment used was discussed in the paper. The overall methodology enabled a very good success rate of classification to be achieved.

5.3. Inorganic materials

5.3.1. Gunshot residue. This has been, by far, the most popular research topic in the forensic science area, with several different aspects reported. These included the determination of shooting distance, classification of ammunition, Pb isotope ratio measurements, the use of portable LIBS instruments, etc. As well as research papers, there has also been a review/overview presented. This review, by Krishna and Ahuja,348 contained 89 references and covered many of the detection techniques that have been used. The review was split into numerous sections. The first of these sections covered the origin of many of the metals commonly detected, e.g., lead styphnate (primary explosive initiator), barium nitrate (an oxidiser) and antimony trisulfide (a fuel). A section followed on the effects that the use of lead-free ammunition has had on the ease of forensic analysis. Most of the review was given to the description of analytical tests that have been developed. The “spot” tests were the first to be developed and these were discussed and summarised in a table in the paper. Other sections covered analysis using NAA, Raman and SERS, SEM-EDS, ICP-MS (plus LA-ICP-MS), Ion Mobility Spectrometry and LIBS. In many cases, examples were tabulated for easy reference.

Single-particle ICP-TOF-MS is a new technique and is a development on sp-ICP-MS that has been around for several years. Two papers by Szakas and Gundlach-Graham have discussed the use of sp-ICP-TOF-MS and have given its relative advantages over more conventional techniques.349,350 In the first of these papers, the problems arising from natural or other particulate matter interfering during the analysis of inorganic gunshot residue (GSR) were discussed.349 Materials such as nanoparticles from car brake pads, fireworks or even some sunscreens can potentially interfere with the analysis because their particulate size range is similar. The sp-ICP-TOF-MS system enables analysis of materials with diameter 10–100 s of nm in a rapid time-frame (typically 2 minutes per sample) at attogram levels. Through the analysis of thousands of nanoparticles from each potential source, the authors developed probabilities for detecting GSR in the presence of other nanoparticles. Based on these analyses, robust sample-specific ‘characteristic’ particle types can be used to classify leaded and unleaded GSR particles, even in the presence of interfering particles. Particles from brake pads and fireworks were the most similar to the leaded GSR ones. However, the GSR particles could unequivocally be identified by the presence of both Pb and Sb. For lead-free ammunition, the most interfering particles originated from sunscreen lotions. However, the GSR particles could easily be differentiated from them because of the unique fingerprint from Ti–Zn–Cu. Overall, very few false positive results were noted (i.e., interfering particles identified as GSR), even when the interfering particles were present at a concentration in excess of 200. The second paper of this type analysed particles produced from leaded and unleaded ammunition and compared results with those obtained from SEM-EDS.350 The leaded munitions could be identified by the presence of Ba, Pb and Sb (just as with SEM-EDS). However, the sp-ICP-TOF-MS could detect and analyse much smaller particles, e.g., in the nm range. However, the average particle size was 180 nm for leaded and 320 nm for unleaded materials. This meant that nearly twice as many particles could be measured per mL of solution compared with SEM-EDS. It was noted that approximately 30% of the particles from leaded ammunition contained particles of Pb–Cu. Unleaded ammunition particles mainly comprised Zn–Ti at a ratio of 1.4[thin space (1/6-em)]:[thin space (1/6-em)]1, but Cu was also a component in many.

The use of LIBS has garnered significant research interest because it offers the possibility of measuring both elemental concentrations and also gives molecular information of organic materials. A paper by Khandasammy et al.351 used LIBS as well as Raman spectroscopy to analyse and then classify different organic gunshot residues produced from the same weapon, but with ammunition from different manufacturers. Once the data had been collected, they were input to Matlab for pre-processing. All spectra were normalised by the total area to enable a better spectral comparison between the different samples. The data were also mean centered before modelling. The modelling strategy involved three steps: (1) genetic algorithm-partial least-squares (GA-PLS)-based feature selection; (2) PLS-discriminant analysis (DA) component construction for further dimension reduction; and (3) support vector machine (SVM)-DA for the prediction/classification. The success rate of the classification was excellent, with only one out of 39 samples being misclassified. This demonstrated the power of a quick and simple technique such as LIBS when combined with statistical analysis.

Historically, the rifling (striation) marks on bullets is used to link bullets with guns and possibly the shooter. However, there are times when the bullet is too badly damaged for this to occur. It may then be possible to use the trace elements present in the gunshot residue. An additional possibility is that the determination of Pb isotope ratios may be used to distinguish between different ammunition. This is the subject of a study by Guo et al.352 who used LA-MC-ICP-MS for the determination of Pb isotope ratios on a bullet fragment found in a victim's head. There had been two suspects, a driver and his boss. Using LA-MC-ICP-MS with 203Tl and 205Tl as mass bias correction and NIST 981 to correct for instrumental mass fractionation, the Pb isotope ratios were determined. The experimental 206Pb/204Pb, 207Pb/204Pb and 208Pb/204Pb values of SRM 981 were 16.9364 ± 0.0009, 15.4924 ± 0.0016 and 36.7078 ± 0.0146, which were consistent with certified values of 16.9371, 15.4913 and 36.7213, respectively. As well as the shot from the two guns and the one recovered from the victim's head, the authors also studied 50 munitions from 14 different manufacturers. Analytical data were then input to the multivariate likelihood ratio algorithm. Results were not conclusive with only limited support for the munition coming from one suspect. However, there was a very strong likelihood that it did not come from the other. Despite the flaws, this evidence was strong enough for the boss to admit that he had accidently shot the victim.

Atomic spectroscopy may also be used to assess the distance from the victim to the shooter. This was the subject of a study by Doña-Fernández et al.353 In this study, rounds from three different manufacturers were fired into white cotton fabric from distances varying between 8 and 200 cm. These ammunition types were selected to ensure variations in the lead-free primer and also because all of them used different projectile types and jackets. Previous SEM-EDS results had indicated that approximately 40% of the particles produced contained Cu. This was therefore chosen as the analyte. An Indra System iForenLIBS V.1 instrument (capable of determining 48 analytes) along with the ballistic module was used to measure the Cu concentration and estimate shooting distance. The system scanned the surface of each specimen around the region at which the bullet penetrated while moving around eight axes, resulting in a total of 2917 laser shots. Once the scanning process was complete, the system generated a map of the relative concentrations. Artificial intelligence algorithms then provided an estimate of the shooting distance by comparing the data obtained with the internal patterns in the system library. The results were very encouraging in that the iForenLIBS system correctly identified whether the shot was from short, medium or long range. It could then be used further by giving a percentage likelihood of it being fired from certain distances. Although these results were less clear cut, they still gave a reasonable estimation of shooting distance. For instance, one shot fired from a distance of 200 cm gave a percentage likelihood of 36% for 140 cm and 51% for 200 cm. The system worked for lead-free ammunition regardless of the primer composition.

The use of portable LIBS instruments is useful for forensic applications in that they may be taken to crime scenes and provide results within minutes rather than hours with traditional lab-based instruments. An example of the use of a mobile instrument was presented by Rodriguez-Pascual et al.354 These authors used the iForenLIBS instrument, discussed previously, on a variety of surfaces, e.g., walls, furniture and assorted fabrics. Since the instrument ablates the GSR particle as well as the sample substrate, the authors emphasised the necessity to analyse an unexposed part of the sample. This was to reduce the number of false positive results. Overall, the analysis was very successful with the most common analytes in GSR particles (Ba, Pb and Sb) being detected in all samples as well as many of the elements that comprise the jacket or core of the projectile (Cu, Pb and Zn). Since the analytes could be detected directly, there was no need to do any “sampling”. A time-based study was also undertaken. The GSR could still be detected close to the bullet hole after one month in virtually all of the fabrics and many of the other sample types.

In another study presented by Vander Pyl et al.,355 a mobile LIBS instrument was developed especially with the analysis of GSR in mind. It had a CMOS detector, a sampling chamber that could hold up to six typical GSR collection devices with separate gas flows to prevent cross-contamination and an image magnification device enabling the small particles to be identified and analysed more easily. The magnification also enabled the particulate materials to be visualised so that morphology could be noted. The newly developed device was used to analyse samples collected from the hands of shooters (100 samples) and from non-shooters (200 samples). The same samples were also analysed using a fully validated laboratory-based instrument, identification of the shooters had an accuracy rate greater than 98.8% for both instrument types. The excellent success rate along with its ability to obtain results in a few minutes, compared with several hours for SEM-EDS, was considered to be a big advantage.

A study by Donghi et al.356investigated how long residue from one type of ammunition lasted in the gun even if it is cleaned or subsequently used with a different type of ammunition. This is an important question, because it makes interpretation difficult and could cause confusion. Six different tests were undertaken. These were: (1) standard gun cleaning with copper brushes, patches and gun oil, (2) use of a dishwasher with alkaline detergents, (3) use of a glassware washer with alkaline detergents, (4) use of an ultrasonic bath and methanol, (5) all four of the previous treatments in sequence, and finally, (6) firing new ammunition (with 5, 10 or 15 rounds). The experiments were conducted using two brand new pistols. Specimens, collected both from the shooters' hands and from cotton targets set nearby the gun muzzle, were analysed using SEM-EDS and using ICP-OES. The residue emerging from the gun barrel showed an asymptotic decrease to zero in the number of particles from the first ammunition. Results from the shooter's hands were less predictable. The authors concluded that residue causing the memory effect was also present in the other parts of the gun rather than just the barrel.

5.3.2. Explosives. There have not been any research papers in this subject area during this review period. However, a review has been presented by Ding et al.357 The review entitled “Recent advances in laser-induced breakdown spectroscopy for explosive analysis” contained 144 references. After an introduction describing the assorted methods of detection, the main parts of the review focussed on LIBS. A section on basic principles was followed by sections on instrumental setup and experimental conditions and then a relatively brief section on data processing. The section on explosives detection was split into several sub-sections that included inorganic oxides, high-energy organic explosives and standoff analysis. A final section discussed how LIBS is being used to assess the performance of explosives.

6. Cultural heritage samples

There is usually a necessity to inflict as little damage to these samples as possible and certainly not to destroy them completely. This is because they may be valuable both in terms of them being rich in history, and so can tell us more about ancient civilisations, and because they may be worth a lot of money. Therefore, it is rare that an application will report the use of an acid digestion or fusion to prepare the sample ready for nebulisation into an ICP instrument. Instead, techniques capable of analysing the materials directly predominate. This may be ones that are regarded as being generally non-destructive, e.g., X-ray-based techniques, or those that are known to be minimally destructive, i.e., LIBS or laser ablation. The latter two do ablate some sample away, but the craters left are usually so small that they cannot be seen with the naked eye. Once the data have been obtained, they are sometimes treated using a chemometric technique to help elucidate provenance, which may also assist in determining trade routes, the characterisation of manufacturing methods, etc.

Several reviews are of interest in this section of the update. One by Manhas et al.358 reviewed (with 72 references) the application of XRF in forensic archaeology. The introduction explained forensic anthropology (the search for clandestine graves and identification of suspects through facial recognition from closed circuit TV, etc.) and forensic archaeology, which covers the methods used for analysis of skeletal remains, artefacts, etc. A description of the capabilities of XRF and the advantages over many other techniques was then provided. The largest section of the review was that of the applications. This was split into the two main sub-sections of anthropological remains and archaeological remains. The former sub-section contained: human skeletal remains, forensic dental examination, animal remains and marine remains. The archaeological remains section covered soil, ceramics, paintings and coins. Factors affecting the forensic investigation of archaeological evidence and the limitations of XRF were also discussed.

Confocal micro-X-ray fluorescence analysis for the non-destructive investigation of structured and inhomogeneous samples” was the title of a review by Heimler et al.359 This review contained 75 references and, although it covered many aspects including theory, its main focus was on applications. It has been used for many areas of research including materials science, biology and geology/environmental science. All of these were covered, as was a section on archaeometry. It was therefore not a specialised review on heritage-type samples. However, the section is interesting and covered the confocal μ-XRF analysis of paintings, manuscripts and ceramics.

Portable XRF is an invaluable tool for archaeologists as it can provide results within minutes, rather than having to wait hours for samples to be sent to a laboratory, and it is non-destructive. Johnson et al.360 discussed how this indispensable tool, which has increased rapidly in popularity over the last 10 years, produces data but that there is no set methodology and so the sample preparation methods, analysis settings and reporting methods vary. They therefore proposed a “best practices” protocol for reporting the data from portable XRF measurements. This includes all of the information that should be presented alongside the data. Different instruments are known to produce different data depending on how the calibration is prepared, the X-ray source and power, etc. The paper provided a useful table that itemised many of the variables that affect the accuracy between different instruments and possibly different users of the same instrument type. Another table itemised the 18 different questions that should be answered or have the relevant information provided when publishing the data. In this way, the authors hoped that the data will become more transparent and more easily inter-relatable assisting journal editors and peer reviewers.

As well as review papers, there are several other papers relevant to multiple areas of cultural heritage. Following on from the paper discussed above, a paper by Chiti et al.361 also discussed how data can vary between XRF instruments. It then described the development of two portable XRF units that had been designed specifically for the analysis of heritage materials. One (named Frankie) had a spot size of 300 μm and therefore specialised in the analysis of very small features or “the detail” of painted materials. The other (named F-70) was specifically designed for the analysis of copper-based materials, especially for the analytes Ag, As and Sb. The abilities and specifications of both instruments were described. They were then applied to the analysis of different artefacts. Frankie was used to analyse small glass beads with confirmation of the results being obtained using LA-ICP-MS. The F-70 was used to analyse copper-based materials such as arrow heads. The LOD obtained were discussed, with the F-70 providing lower values (typically 10−2–10−3 %) compared with 10−1–10−2 % for Frankie. However, Frankie provided superior LOD for elements with lower atomic number, e.g., Co, Fe and Mn.

Radiocarbon dating is a relatively common analysis for archaeological materials, Moreau et al.362 presented a paper that described their participation in an inter-laboratory exercise called MODIS, in which historical mortars were dated using AMS. Their approach was to heat all of the candidate materials to 550 °C for 30 min to remove organic contaminants and layered double hydroxides. These may otherwise contaminate the sample with “young carbon”. After this pre-heating, the samples underwent a thermogravimetric analysis at temperatures that were sample dependent, but typically in the region 550–740 °C. Finally, the samples were heated to 840 °C, the carbon dioxide collected and analysed using AMS. Five materials had been supplied and the results of four were in good agreement with those from other laboratories. The errant sample, MODIS2-3, did not give reproducible results from any laboratory, with most using different methods (thermal decomposition, sequential dissolution, etc.). Calculated dates were very often much older than the expected age. This was attributed to the presence of dolomite, a natural geological carbonate.

6.1. Analysis of metallic cultural heritage samples

As explained previously, there is usually an attempt to cause the minimum damage to the sample. An interesting paper presented by Porcinai et al.363 determined if there was a difference in the analytical results obtained using portable XRF when the sample surface was analysed compared with those obtained when analysing shavings. This is obviously an important question since if a completely different answer is obtained, then at least one of the methods is giving incorrect data. The authors obtained 22 copper-based certified materials and then determined the accuracy, precision and repeatability of the measurements obtained from the bulk shiny surface and from shavings of the same materials. Results for Ag, Bi, Co, Cu, Fe, Mn, Ni, Pb, Sb, Sn and Zn were obtained. Of the 22 materials, 13 were used to construct calibration curves and the other nine were used for validation. Two regression models were tested. One simply exported the raw counts into a program called PyMCA and obtained fully quantitative data by plotting simple calibration curves. The other model used a plot of nominal concentration (independent variable) against the concentration obtained using PyMCA (dependent variable). Results using the first method were somewhat varied, but Pb in particular provided some poor data when shavings were used. The other regression model provided comparable accuracy between surface analysis and shavings, although the uncertainties associated with the analysis of shavings was slightly higher. It was concluded that accuracy need not be compromised when using different analysis methods as long as a suitable regression model is used and a calibration is made from certified materials in the same form as the samples under investigation.

A similar paper was presented by Bliujiene et al.,364who compared the data obtained from a portable instrument with those from a commercial EDXRF instrument. Again, a comparison was made between the analysis of the bulk surface and shavings from cultural heritage alloy reference materials – the so-called CHARM series. The fundamental parameter correction method was applied, assuming that all analysed elements added up to 100%. The copper alloys were characterised according to their principal alloying elements (Pb, Sn and Zn) and the presence of small amounts of the deliberately added alloying elements. These authors also found that the shavings produced significantly less precise data than those obtained from the surface. A little surprisingly, the data obtained using the portable instrument were both more accurate and precise than those from the EDXRF instrument. This was true for both surface analysis and for the shavings.

The quality of data obtained from handheld XRF instruments was also the subject of a paper presented by Konstantakopoulou et al.365 These workers also used copper-based alloy certified materials to compare the accuracy of three different regression models. One model was the “built-in” calibration in the commercial instrument used and the second was a customised calibration that the authors had prepared following the Bruker procedure. Both of these were based on the empirical coefficients approach. The third model was used off-line, again using the software PyMCA. The regression models were all discussed in the paper. In virtually all cases, the “built-in” calibration produced the poorest regression coefficients sometimes, as in the case of Fe, being as low as 0.86. The customised regression improved the regression coefficients significantly, with R2 values being typically 0.95–1. In many cases, the PyMCA regression provided the best results of all. For the customised and PyMCA regression models, a total of 26 certified materials of the CHARM series were analysed so that the models could be constructed and tested. This work shows that simply relying on the “in-built” calibrations in portable instruments can lead to errors and that for best accuracy, the regressions should be built on the analysis of matrix-matched materials.

Many other applications have been reported and these are summarised in Table 4.

Table 4 Applications of the analysis of metallic cultural heritage samples
Analytes Sample Technique Comments Reference
Various including Ag, Au, Cu, Fe and Zn Silver coins from 5th to 3rd century BCE Italy Macro-XRF and XRF Elemental maps of corroded and un-cleaned samples were produced using non-negative matrix factorisation, region of interest imaging, k-means and deconvoluted maps. 366
Various including Ag, Au, Bi, Cu, Fe, Pb and Si 18 double relief silver coins from Southern Italy pXRF Analytical data interrogated using PCA, hierarchical cluster analysis (HCA) and graph analysis. The models enabled the coins to be classified according to historical period. 367
Various 13 bits of armour and arms from medieval Aegean region LA-ICP-MS Analytical data interrogated using PCA, LDA and PCA followed by HCA. The armour had considered to be of Italian origin. The statistical analysis indicated several different origins. 368
Various including Cu, Pb and Sn Bronze objects from Bronze age Bulgaria LIBS; XRF 60 artefacts from 15–11th century BCE analysed. Method validation using four bronze certified materials. Data inserted to PCA. The samples were classified according to their date of manufacture. 369
Various including Ag, Au, Ca, Cr, Cu, Pb, S and Zn Bimetallic alloy from Ancient Moche, Peru XRF XRF integrated with Monte Carlo simulation used to analyse a thigh plate split into two halves: one seemingly silver and the other gold. Analysis identified the “silver” part to be a silver–copper alloy with a patina layer. The “gold” half was a copper–silver–gold alloy called Tumbaga. 370
Various, up to 11 111 copper-based and 11 silver alloy coins from Rhodes covering the period 4th century BCE to 2nd century CE μ-XRF Method validation through the analysis of CRMs BCR 691, BAM-374 and NIST 1107 for the copper-based coins. Silhouette analysis for k-means clustering identified four distinct groups for the copper-based coins. Four groups also identified for the silver coins based on the grade of silver 371
Various including Ag, Au, Bi, Fe, Hg and Zn 213 coins from Poland XRF; NAA A series of statistical tests used to classify the coins into groups. These tests included HCA (7 groups), Kohonen neural network (6 groups) and k-means cluster analysis. 372
Various including Al, Ca, Cu, Mg, Na, Ni, Sn and Zn 13 old Indian coins from between 1922 and 1986 Calibration-free LIBS Data input to PCA and SIMCA. Four distinct groups of coins detected that are prepared from four different alloys. Surface hardness measurements also undertaken and correlate with the LIBS elemental data. 373
Various including Ag, Cu, Pb, Sn and Zn 12 corroded Roman copper-based coins X-ray computed microtomography; XRF A largely successful attempt to analyse corroded and illegible coins without recourse to cleaning. The features became clearer in most coins. The coins were then cleaned and analysed using XRF to verify the identity of the denomination through the alloy used. 374
Various including Ag, Au, Hg, Pb, Pd and Pt Gold coins from the Roman empire XRF; XANES An X-ray beam of 100 × 100 μm2 was used as a compromise between speed of analysis and resolution. Elements found in the hollow parts of the coin, e.g., Ca, Fe, Ni, etc., were assumed to arise from soil. This was confirmed using XANES. XRF data were input to PyMCA for background subtraction and calibration and then inserted into T-distributed stochastic neighbour embedding for group classification. 375
Various including Au, Ir, Os, Pd, Pt, Rh and Ru Silver coins of different age (from Ancient Greece to 18th century CE) and from different places, e.g. India, Europe, Mexico and South America ICP-MS Samples were cut with pliers, cleaned and acid digested prior to ICP-MS analysis. Pt, Pd, Rh, Ir and Ru were determined for the first time in silver coins using quadrupole ICP-MS in KED mode. 376
Various including Cu, Fe and Pb Gold coins MC-ICP-MS Provenance study using isotope ratios. Pieces of coins cut, acid digested and the gold separated using anion exchange. The analytes then determined using MC-ICP-MS. 377
Various and Pb isotopes 37 Roman lead artefacts from the Arade river in Portugal ICP-MS and MC-ICP-MS A provenance study of the artefacts. Multi-analyte determination using ICP-MS followed by cluster analysis of the analytical data classified the artefacts into groups. The Pb isotope ratios obtained using MC-ICP-MS then helped determine provenance. Some samples may have originated in Africa, indicating long-distance trade. 378


6.2. Cultural heritage samples of organic origin

Cultural heritage samples of organic origin include samples such as manuscripts and documents, inks, paints and paintings. This has been quite a popular area of research during this review period. A review by Perino et al.379 contained 106 references and was entitled “New Frontiers in the Digital Restoration of Hidden Texts in Manuscripts: A Review of the Technical Approaches”. Old documents can contain writing that is not visible to the naked eye because it has been erased or has faded. Several techniques were discussed in this review that help to enable these things to be seen again. These included: UV photography, multi- and hyperspectral imaging, macro-XRF, X-ray tomography, IR thermography, terahertz imaging and photoacoustic imaging. Unsurprisingly, most of the techniques were not related to atomic spectroscopy. However, the sections that are, provide useful information and references.

Cortea et al.380 described the development of an open-access database called INFRA-ART. The database contains over 1000 spectra obtained using attenuated total reflection FTIR, XRF and Raman of pure substances used as artists' materials, e.g., paints. Such data are required to assist curators determining the type of painting materials used, techniques and possibly provenance. This paper presented a summary of the database structure and design as well as its functionality. A keyword search function has been introduced to aid workers find the relevant spectra. Future work will continue to expand the database with types of mixtures, e.g., binder–pigment, pigment–pigment, and pigment–dye.

A paper entitled “A cloud-native application for digital restoration of cultural heritage using nuclear imaging: THESPIAN-XRF” was presented by Bombini et al.381 It described how a trained neural network can be employed as a web application for XRF raw data real-time analysis. The authors described its development and commented on the outcomes.

Using a combination of techniques can aid the identification of the materials used for painting, reveal under-paintings, painting techniques and can play a key role in art conservation and research. The advantage of having a combination of complimentary techniques on one platform is that more than one analysis may be undertaken on exactly the same spot simultaneously. This is not possible when using two separate instruments. Occhipinti et al.382 described an instrument called IRIS, which combined macro-XRF with visible near IR and short wave IR hyperspectral imaging. The full details were given in the paper, but it comprised a motorised XYZ stage and an analysis head weighing 3.5 kg. The XRF component employed a rhodium-target X-ray tube that used a voltage ranging between 10 and 50 kV, current 5–200 μA, for a maximum power of 10 W and up to five X-ray filters. The system also had three collimators of 0.5, 1.0 and 2.0 mm. The parameters were all changeable giving the instrument a wide range of functionality. The IR hyperspectral imaging system could operate between 400 and 2500 nm. The instrument was designed to be mobile, i.e., it may be taken out of a laboratory, to be quick, simple and non-invasive. It could have a high sample throughput whilst retaining good spatial resolution (the best spot size was 0.5–0.5 mm). The capabilities of the instrument were illustrated by the analysis of a painting using a scan speed of 27.8 mm s−1. Trace elements could clearly be associated with certain colours, e.g., Cu with light blue, Fe with browns and yellows and Hg with red.

Another paper that reported the use of two complimentary techniques was presented by Kechaoglou et al.383 These authors combined macro-XRF and fs-LIBS (albeit not on the same instrument platform) to analyse the stratigraphy of mock paintings. The paintings were prepared using either a single layer of assorted organic or inorganic pigments on a layer of calcium carbonate substrate, or two or three successive layers placed on the substrate. The LIBS was used by firing 10 laser shots on the same spot and monitoring any change in signal. The crater size remained below 100 μm in all cases – an advantage offered over the use of ns-LIBS. Although this is a recognised method of depth-profiling, it is not without its problems. Therefore, the complimentary method of macro-XRF was ideal in completing the knowledge picture and so it was concluded that it was very promising for the future analysis of layered paintings.

An interesting paper by Swiecicka et al.384analysed iron gall inks indirectly. This was achieved by soaking some indicator paper in 4,7-diphenyl-1,10-phenanthroline and then resting it on some model gall inks prepared from gall nut extracts, copper sulfate, iron sulfate and gum arabic. The trace metals bound with the complex and were extracted from the ink onto the indicator paper. This was then analysed using LA-ICP-MS, portable XRF and SIMS. The portable XRF instruments used had a resolution of 1 mm2, whereas the laser beam for LA-ICP-MS had a spot size of 100 μm. A total of seven different types of paper (and with different density) were tested along with four model gall inks. The XRF could clearly identify the Cu, Fe and S where present in the gall ink. The LA-ICP-MS demonstrated that numerous other analytes were also extracted but at concentrations below that at which the XRF instruments could measure. The SIMS results demonstrated that the fragment ions were typically ML2+, ML+ or M+ where L is the ligand and M is either Cu or Fe. The SIMS analysis is particularly interesting because many samples would not fit into the analysis chamber. Therefore, this indirect approach could potentially be of great use.

Artificial intelligence (machine learning) techniques are often used to help interpret analytical data. Andric et al.385 reported the development of an autoencoder neural network to be used as a dimension reduction tool for data that were initially of dimensions 40 × 2048 that had been collected using EDXRF analysis of a painting. The autoencoder enabled the best reconstruction of the original EDXRF spectrum by extracting the most informative features and discarding redundant information. In this way, the data input to classifying algorithms is much decreased, enabling much faster processing and requiring less computer power and expert input.

A technique entitled gas optical emission enhanced by solid initiator (GENS) as a support to LIBS analysis was described by Bai et al.386 This technique produces plasma in the atmosphere using a metal target, which increases sensitivity while lowering laser irradiance. It is used to monitor gas emissions from old materials during intense radiation analysis, particularly under ion beam irradiation. This study focused on determining hydrogen gas emissions from lead white pigments mixed with linseed oil as a binder in paintings. The lead-white-containing paint layers were exposed to 10 to 40 μC cm−2 of 3 MeV protons in a specially designed sealed cell and the H given off was determined using the technique. The authors stated that as well as H, other light elements, e.g., N and O, could also be determined and concluded that it was a safe and non-destructive method of analysing valuable cultural heritage samples.

Several other applications of the analysis of organic materials have been presented that offer some novelty with regard to the atomic spectrometry or handling of atomic spectrometric data. These are summarised in Table 5.

Table 5 Applications of the analysis of organic cultural heritage materials
Analyte Matrix Technique Comments Reference
As Paintings XANES; μ-XRF The degradation of orpiment (As2S3) was investigated. The use of light only led to the formation of As2O3. Sunlight plus a medium of egg white suspended in water led to the formation of AsV. The egg white/water mix in the dark also led to the formation of AsV. 387
Pb Paintings Portable XRF; macro-XRF, ICP-MS; MC-ICP-MS and XPRD Some lead-white pigments darken with time and others do not. A study of this phenomenon was made. Degradation products, e.g., anglesite formed with sulfur-containing materials present during historical paper production (e.g., gypsum), if stored incorrectly (exposed to hydrogen sulfide) galena forms. Provenance of Pb had no influence on discoloration. 388
Various Brazilian paintings of the 20th century Macro-XRF; XRF; FTIR and optical microscopy XRF data followed by PCA and spectral deconvolution could distinguish between genuine and fake paintings based on the materials used. Macro-XRF also demonstrated a lack of a polychrome preparation layer in the fake paintings. 389
Various (12) 23 papyri from 4–7th century CE Egypt Portable XRF Data input to PCA. The inks used for documents were carbon-based whereas iron gall inks were used for literary texts. 390
Various Harward's Almanac pages Portable XRF Raw data exported to PyMCA for semi-quantitative calibration. Analytical data input to self-organising maps in an attempt to identify the correct order in which 19 poems were written. Two groupings observed: 1674–1693 and 1699–1711. Other techniques, e.g., palaeography and codicology, also used. 391
Various Flower still life paintings from The Mauritshuis, The Netherlands Macro-XRF The use of macro-XRF led to many insights regarding the ways painters prepared their compositions. Many of the colours were used by most artists. The under-drawings could clearly be seen. 392
Various Ancient architectural paintings Portable XRF Mock paintings prepared and analysed. Data input to PCA enabled specific analytes to be associated with certain colours. The effects of a dust layer on the analytical accuracy were also determined. Al, Ca, Fe and Si results were compromised. Authors recommended that paintings be blown clean prior to analysis. A formula was derived to obtain accurate data from layered materials. 393
Various The painting “The Night Watch” by Rembrandt XRF; X-ray ptychography; SEM-EDX An unexpected lead-containing layer was discovered. This was thought to act as a protective impregnation layer. The ptychography visualises and maps the elements in 3D. It is also capable of determining components not detectable using XRF, e.g., organic materials and quartz. 394


6.3. Ceramic materials of cultural heritage

An overview, containing 107 references, of the possibilities and limitations of various X-ray fluorescence techniques for the study of the chemical composition of ancient ceramics was presented by Chubarov et al.395 Wavelength dispersive, energy dispersive, portable and spectrometers with polycapillary optics (μ-XRF) instruments were all discussed. Total reflection geometry was also discussed. The different techniques all have different requirements for sample preparation, and for calculating the elemental concentrations, and so this was also discussed. The authors presented their own use of each of the techniques for the analysis of archaeological materials from the stone-age from Siberia.

A paper entitled “Advancements in characterisation of ancient potteries from south east Asia: a review of analytical techniques” was presented by Sirisathitkul.396 The “review” had only 49 references and omitted techniques such as LA-ICP-MS, LIBS and several other commonly used techniques. However, it did have sections covering XRF and XRD, PIXE and NAA, synchrotron X-ray-based techniques (e.g., EXAFS and XANES) and vibrational spectroscopies (IR and Raman). The abilities of each technique were briefly discussed, although no theory was given. A table presented many of the applications that have been undertaken.

Other papers that have described the analysis of cultural heritage ceramics are discussed in Table 6.

Table 6 Applications of the analysis of cultural heritage ceramics
Analyte Sample Technique Comments Reference
Pb isotope ratios Medieval Besztercebanya/Banska Bystrica-type stove tiles MC-ICP-MS A comparison of three sample preparation methods: chipping glaze off and acid digestion; acid digestion followed by separation of Pb from the matrix using extraction chromatography and swabbing the surface with 8 M nitric acid and then soaking swab in 3% nitric acid. Pb isotope ratios were independent of preparation method and therefore the simplest and least destructive one was adopted (swabbing). 397
Pb isotopes Majolica from 15–19th century Florence TIMS; SEM-EDS Provenance study. 1 mg of sample was scraped off each of 10 samples ensuring underlying ceramic body was removed completely. NIST 981 used to gauge precision. Reference material AVG-1 used to determine accuracy of ratios. The lead for the glazes appears to originate in Germany and was not sourced locally. 398
Various 233 Yuan Dynasty blue and white porcelain Portable XRF Calibration graphs made from 13 standard samples. Data accuracy validated using GBW07402 (clay). Total number of samples was 257 with 24 samples from other dynasties (anomalous samples). Machine-learning gradient attention map (GRAM) used to extract anomalous patterns within an underlying variational graph autoencoder (VGAE). 399
Various Ceramic raptors from Northern China (2300–1800 BCE) Portable XRF Analytical data input to PCA. The ceramic chemical composition does not match the local soil. Three sub-groups identified visually. They also have different compositions, indicating different sources of clay. 400
Various (Al, Cu, Mg, Pb and Si) Glaze of Chinese glazed tiles Portable XRF; EDXRF A novel binary linear regression model used for calibration. The quality of body-glaze fitting calibration data was “satisfactory”. 401
Various White decoration of Neolithic pottery from Bulgaria LIBS; ATR-FTIR Semi-quantitative results were input to PCA. Calcite, enriched with diverse fillers such as quartz, clays, feldspars, and metal oxides, was the primary raw material for white decoration. 402
Various (18) Roof tiles and bricks from the Etruscan Domus dei Dolia, Vetulonia, Italy EDXRF; XRD; optical microscopy Samples prepared as fused glass disks. Data input to PCA. Roof tiles and bricks shown to be prepared using different technologies and materials. Roof tiles were probably produced within 4–12 km from the Domus, using two different raw material sources. 403
Various 35 samples of pottery from Oman (3rd century BCE–5th century CE) ICP-MS; XRPD; optical microscopy; SEM-EDS Hotplate digestion of 100 mg of sample using hydrofluoric and nitric acids. Results input to PCA and ternary diagrams. Eight distinct groupings identified: three from Southern Arabian peninsula, two from North-West India, one from Southern India, one from North Eastern Indian subcontinent and one of unknown origin. 404
Various 30 samples of medieval ceramics from Southern Kazakhstan XRF; NAA Results of some major, minor and trace elements input to binary scatterplots, factor and cluster analysis. The places of pottery burial could be related to the production site. 405
Various Roman ceramics from Sinop, Turkiye Polarised EDXRF; XRD; FTIR; TG-DTA Samples presented as a pressed pellet. GBW-7109 and GBW-7309 used for validation. Results input to ternary diagrams, HCA and PCA. Two different raw material sources were identified. Sub-groups were also identified in the Boyabat Roman potsherds. 406
Various (18) Pottery (95 sherds) from five close-by kilns in the Zhoudang town, Hunan Province, China and 57 clay samples WDXRF Data input to PCA and Bi-plots (e.g., Rb against Sr or Rb against Zr). Distinction between pottery from different kiln sites was possible mainly through SiO2, CaO, K2O, Sr, Zr, and Rb concentration differences. 407


6.4. Glass materials of cultural heritage

There have been relatively few applications of note during this review period. Those that worthy of mention are discussed in Table 7.
Table 7 Applications of the analysis of glass cultural heritage artefacts
Analyte Sample Technique Comments Reference
Fe and Mn 5–7th century glass imported to Britain Micro-XANES; μ-XRF Investigation into the effect of Mn/Fe ratios and oxidation states on the colour of glasses, which ranged from green to yellow/amber, with some even having purple streaks. In yellow samples, the analytes were present as Fe3+ and Mn2+. This was also true of the purple-streaked ones, although the purple areas had more Mn. Pale green glasses had a layer of oxidised Mn. 408
Various 23 glass beads from Soba, Nubia LA-ICP-MS Reference materials Corning B and D as well as NIST 610 used for calibration. 29Si used as internal standard. Data input to PCA which identified several groupings: Middle East dating to before the 12th century, Egypt from the 13–14th-century, North Indian glass dated between the ninth and 13th centuries and North Indian origin and dated to 14–17th centuries. 409
Various Late Bronze age Egyptian glass from Amarna PIXE; PIGE The geochemical diorite DR-N sample and the Corning A, B and D glass standards were used as reference materials to calibrate the PIGE data and control PIXE results. Ternary diagrams used to determine groups/classify samples. Other transition metals present enabled provenance of the cobalt colourant to be identified. 410
Various 146 glass samples from South West Taiwan from 500 BCE–700 CE LA-ICP-MS; SEM-EDS Data input to PCA and binary plots. Glasses identified as originating in Mesopotamia or Susanian territories, a Roman glass, and from South and South West Asia. The study indicates the extent of glass trading over as much as 8000 km. 411
Various 4–12th century glass from the northern Venetian lagoon LA-ICP-MS; EMPA Data input to ternary plots. All glass samples identified as soda-lime silica with varying Fe and Mn originating from impure sands. Most materials were of Egyptian origin. Some were Levantine. 412


7. Abbreviations

2Dtwo dimensional
3Dthree dimensional
AASatomic absorption spectrometry
ABSacrylonitrile butadiene
AESAuger electron spectrometry
AFFFasymmetric field flow fractionation
AF4asymmetric flow-field flow fractionation
AFSatomic fluorescence spectrometry
AFMatomic force microscopy
AMSaccelerator mass spectrometry
ANOVAanalysis of variants
ARECVaverage relative error of cross-validation
ASTMAmerican Society for Testing of Materials
ATRattenuated total reflection
BCRCommunity Bureau of Reference
CCDcharge coupled device
CEcapillary electrophoresis
CFFFcentrifugal field flow fractionation
CIGScopper indium gallium selenide
CRMcertified reference material
CScontinuum source
CTcomputerised tomography
CVcold vapour
DAdiscriminant analysis
DLSdynamic light scattering
DLTVdiode laser thermal vaporisation
DP-RLIBSdouble-pulse resonance laser-induced breakdown spectrometry
DMSOdimethylsulfoxide
DRCdynamic reaction cell
EASTexperimental advanced super conducting tokamak
EDIXSenergy dispersive inelastic X-ray scattering
EDSenergy dispersive spectrometry
EDXRDenergy dispersive X-ray diffraction
EDXRFenergy dispersive X-ray fluorescence
ELMextreme learning machine
EPMAelectron probe microanalysis
ESI-MSelectrospray ionisation mass spectrometry
ETAASelectrothermal atomic absorption spectrometry
ETVelectrothermal vaporisation
EXAFSextended X-ray absorption Fine structure
FAASflame atomic absorption spectrometry
FESEMfield emission scanning electron microscopy
FFFFflow field flow fractionation
FIflow injection
FIBfocused ion beam
FI-CVGflow injection chemical vapour generation
FTIRFourier transform infrared
FWHMfull width at half maximum
GCgas chromatography
GD-MSglow discharge mass spectrometry
GD-OESglow discharge optical emission spectrometry
GF-AASgraphite furnace atomic absorption spectrometry
GI-SAXSgrazing incidence small angle X-ray scattering
GIXRDgrazing incidence X-ray diffraction
GIXRFgrazing incidence X-ray fluorescence
HAXPEShard X-ray photoelectron spectroscopy
HDChydrodynamic chromatography
HGhydride generation
HPLChigh performance liquid chromatography
HR-CSAAShigh resolution continuum source atomic absorption spectrometry
IAEAInternational Atomic Energy Agency
IBAion beam analysis
ICAindependent component analysis
ICPinductively coupled plasma
ICP-MSinductively coupled plasma mass spectrometry
ICP-OESinductively coupled plasma optical emission spectrometry
ICP-QMSinductively coupled plasma quadrupole mass spectrometry
ICP-TOFMSinductively coupled plasma time-of-flight mass spectrometry
IDisotope dilution
IL-DLLMEionic liquid-dispersive liquid–liquid microextraction
IPInstitute of petroleum IRMS isotope ratio mass spectrometry
ISOInternational Organisation for Standardisation
ITERInternational thermonuclear experimental reactor
JETJoint European Torus
K-SVM-RFE k-fold support vector machine recursive feature elimination
LAlaser ablation
LASILlaser ablation of sample in liquid
LCliquid chromatography
LDAlinear discriminant analysis
LIASlaser-induced ablation spectrometry
LIBSlaser-induced breakdown spectrometry
LIBS-LAMSlaser-induced breakdown spectrometry-laser ablation mass spectrometry
LIFlaser-induced fluorescence
LIPSlaser-induced plasma spectroscopy
LODlimit of detection
LOQlimit of quantification
LTElocal thermal equilibrium
MALDI-TOFmatrix-assisted laser desorption ionisation time-of-flight
MALSmulti-angle light scattering
MASmagic-angle spinning
MCmulticollector
MEISmedium energy ion scattering spectroscopy
MIBKmethylisobutylketone
MIPmicrowave-induced plasma
MIP-AESmicrowave plasma atomic emission spectrometry
MSmass spectrometry
MSFAmono-segmented flow analysis
MWCNTmulti-wall carbon nanotubes
NAAneutron activation analysis
NAARneutron activation autoradiography
NDneutron diffraction
NDI-MSnear-field desorption ionisation mass spectrometry
Nd:YAGneodymium-doped-yttrium aluminium garnet
NEXAFSnear-edge X-ray fine structure
NISTNational Institute of Standards and Technology
NMRnuclear magnetic resonance
NRAnuclear reaction analysis
OESoptical emission spectrometry
PARAFACparallel factor analysis
PBSphosphate buffered saline
PCAprincipal component analysis
PCRprincipal component regression
PETpolyethylene terephthalate
PFCplasma-facing components
PGAAprompt gamma neutron activation analysis
PGMplatinum group metals
PIGEparticle-induced gamma ray emission
PIXEparticle-induced X-ray emission
PLALpulsed laser ablation in liquids
PLSpartial least-squares
PLS-DApartial least-squares discriminant analysis
PLSRpartial least-squares regression
ppbparts per billion
ppmparts per million
RAFMreduced activation ferritic/martensitic
RBSRutherford backscattering spectrometry
RDAregularised discriminant analysis
REErare earth elements
rfradiofrequency
RIMSresonance ionisation mass spectrometry
RMSECVroot mean square error of cross validation
RMSEProot mean squared error of prediction
RoHSrestriction of hazardous substances
RSDrelative standard deviation
RSFrelative sensitivity factor
SDS-PAGEsodium dodecyl sulfate–polyacrylamide gel electrophoresis
SECsize exclusion chromatography
SEMscanning electron microscopy
SEM-EDSscanning electron microscopy-energy dispersive spectrometry
SFsector field
SIAsequential injection analysis
SIBSspark-induced breakdown spectrometry
SIMCAsoft independent modelling of class analogy
SIMSsecondary ion mass spectrometry
SNRsignal-to-noise ratio
spsingle particle
SPIONsuperparamagnetic iron oxide nanoparticles
SRsynchrotron radiation
SRMstandard reference material
SXRFsynchrotron X-ray fluorescence
SVRsupport vector regression
STXMscanning transmission X-ray microscopy
TEMtransmission electron microscopy
TDSthermal desorption mass spectrometry
TGAthermogravimetric analysis
TIMSthermal ionisation mass spectrometry
TOFtime of flight
T-PGAAtime-resolved prompt gamma activation analysis
TPRtemperature programmed reduction
TXRFtotal reflection X-ray fluorescence
UAEultra-sound-assisted extraction
UV-visultra-violet-visible
WDXRFwavelength dispersive X-ray fluorescence
WEEEwaste electrical and electronic equipment
XAFSX-ray absorption fine structure spectrometry
XANESX-ray absorption near-edge structure
XASX-ray absorption spectroscopy
XPSX-ray photoelectron spectroscopy
XRDX-ray diffraction
XRFX-ray fluorescence
XRPDX-ray powder diffraction
XRRX-ray reflectometry

Conflicts of interest

There are no conflicts to declare.

References

  1. R. Clough, A. Fisher, B. Gibson and B. Russell, J. Anal. At. Spectrom., 2023, 38, 2215–2279 RSC.
  2. 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.
  3. C. Vanhoof, J. R. Bacon, U. E. A. Fittschen and L. Vincze, J. Anal. At. Spectrom., 2023, 38(9), 1730–1743 RSC.
  4. E. H. Evans, J. Pisonero, C. M. M. Smith and R. N. Taylor, J. Anal. At. Spectrom., 2024, 39, 1188–1211 RSC.
  5. M. Patriarca, N. Barlow, A. Cross, S. Hill, A. Robson, A. Taylor and J. Tyson, J. Anal. At. Spectrom., 2024, 39, 624–698 RSC.
  6. R. Clough, C. F. Harrington, S. J. Hill, Y. Madrid and J. F. Tyson, J. Anal. At. Spectrom., 2023, 38, 1339–1371 RSC.
  7. M. Iro, D. Ingerle, S. Hampel, U. Fittschen, V. Dhamgaye, O. Fox and C. Streli, Softwarex, 2023, 23, 101481,  DOI:10.1016/j.softx.2023.101481.
  8. L. B. Guo, X. C. Niu, M. S. Zhang, J. Y. Zhou, D. Zhang and B. Liu, Acta Photonica Sin., 2023, 52(3), 0352104,  DOI:10.3788/gzxb20235203.0352104.
  9. A. Krolicka, A. Maj and G. Loj, Materials, 2023, 16(20), 6641,  DOI:10.3390/ma16206641.
  10. X. Jin, T. Gu, Y. Zhan, D. Qu, G. Chen, G. Yang, D. Tian, C. Li and L. Yao, Appl. Spectrosc. Rev., 2024, 59, 1278–1321,  DOI:10.1080/05704928.2024.2320146.
  11. M. Zhang, H. Song, J. Shang, X. Liu, S. Qi and H. Li, Spectrochim. Acta, Part B, 2023, 207, 106740,  DOI:10.1016/j.sab.2023.106740.
  12. M. F. Gonzalez, N. Saadatkhah and G. S. Patience, Can. J. Chem. Eng., 2024, 102, 2004–2018,  DOI:10.1002/cjce.25218.
  13. W. Kutschera, A. J. T. Jull, M. Paul and A. Wallner, Rev. Mod. Phys., 2023, 95(3), 035006,  DOI:10.1103/RevModPhys.95.035006.
  14. E. Kepes, J. Vrabel, P. Siozos, V. Pinon, P. Pavlidis, D. Anglos, T. Chen, L. Sun, G. Lu, D. J. Diaz-Romero, S. Van den Eynde, I. Zaplana, J. Peeters, V. Kana, A. Zadera, V. Palleschi, A. De Giacomo, P. Porizka and J. Kaiser, Spectrochim. Acta, Part B, 2023, 206, 106710,  DOI:10.1016/j.sab.2023.106710.
  15. Y. Gu, Z. Chen, H. Chen and F. Nian, Electronics, 2023, 12(12), 2566,  DOI:10.3390/electronics12122566.
  16. X. M. Lin, Z. J. Zhang, J. J. Lin, X. Y. Yue, Y. T. Huang, J. F. Yang, Y. J. Dong and X. Y. Duan, Anal. Lett., 2024, 57, 1829–1840,  DOI:10.1080/00032719.2023.2277404.
  17. N. Zhao, Z. Luo, J. Li, Q. Ma, L. Guo, C. Shan and Q. Zhang, At. Spectrosc., 2023, 44(4), 219–226 CrossRef CAS.
  18. M. V. Belkov, D. A. Borisevich, K. Y. Katsalap and M. A. Khodasevich, Opt. Spectrosc., 2022, 130(8), 488–494 CrossRef CAS.
  19. M. V. Belkov, K. Y. Catsalap, D. A. Korolko and M. A. Khodasevich, J. Appl. Spectrosc., 2023, 90(2), 274–278 CrossRef CAS.
  20. Q. Ma, Z. Liu, T. Zhang, S. Zhao, X. Gao, T. Sun and Y. Dai, Talanta, 2024, 272, 125745,  DOI:10.1016/j.talanta.2024.125745.
  21. J. Peterson, M. Gilbert-Gatty, K. Ekstrom, L. Hagesjo and A. Bengtson, Appl. Spectrosc., 2023, 77(8), 907–914 CrossRef CAS PubMed.
  22. X. Bai, R. Hai, Z. He, H. Wu, X. Wang, J. Mu, C. Li, D. Wu, G. Xu, Z. Hu, F. Ding and H. Ding, Spectrochim. Acta, Part B, 2023, 206, 106724,  DOI:10.1016/j.sab.2023.106724.
  23. Y. Lu, L. Liu, Z. Wu, Z. Xu, Z. Zhao, Z. Hao, J. Shi and X. He, J. Anal. At. Spectrom., 2023, 38(12), 2619–2624 RSC.
  24. W. Bai, W. Chen, C. Yang, Y. Hong and R. Wang, Laser Optoelectron. Prog., 2022, 59(23), 2330001,  DOI:10.3788/lop202259.2330001.
  25. G. Jeon, H. Keum, H. Lee, K. Oh and J. Choi, Spectrochim. Acta, Part B, 2024, 212, 106857,  DOI:10.1016/j.sab.2024.106857.
  26. Q. Wang, G. Li, Y. Hang, S. Chen, Y. Qiu and W. Zhao, Materials, 2023, 16(16), 5599,  DOI:10.3390/ma16165599.
  27. Z. Shang, M. Dong, W. Lu, J. Cai, K. Bai, F. Tang, Z. Li, S. Yao and J. Lua, At. Spectrosc., 2024, 45(1), 15–25,  DOI:10.46770/as.2023.297.
  28. J. Yang, L. Kong and H. Ye, Appl. Opt., 2024, 63(10), 2509–2517 CrossRef CAS PubMed.
  29. V. Gupta, A. K. Rai, T. Kumar, A. Tarai, G. M. K. Gundawar and A. K. Rai, Anal. Sci., 2024, 40, 53–65,  DOI:10.1007/s44211-023-00429-y.
  30. A. D. Elias, M. S. Luz, I. Gaubeur, J. Naozuka, F. R. P. Rocha and C. S. Nomura, J. Anal. At. Spectrom., 2023, 38(10), 2105–2112 RSC.
  31. L. Quackatz, A. Griesche, J. Nietzke and T. Kannengiesser, Weld. World, 2024, 68, 915–923,  DOI:10.1007/s40194-023-01677-2.
  32. T. I. Dorofeeva, M. V. Fedorischeva, T. A. Gubaidulina, O. V. Sergeev, A. R. Sungatulin and V. P. Sergeev, Metals, 2023, 13(8), 1468,  DOI:10.3390/met13081468.
  33. P.-C. Huang, P.-C. Chu, T.-J. Yang, J.-Z. Wang and W.-C. Lin, Appl. Surf. Sci., 2024, 648, 158942,  DOI:10.1016/j.apsusc.2023.158942.
  34. M. Buoso, B. Campanella, M. Onor, E. Pitzalis and A. D'Ulivo, J. Anal. At. Spectrom., 2024, 39(1), 66–70 RSC.
  35. Y. Shen, L. Howard and X.-Y. Yu, Materials, 2024, 17(2), 528,  DOI:10.3390/ma17020528.
  36. A. Matsumoto, Y. Shimazu, S. Yae and T. Sakka, J. Anal. At. Spectrom., 2023, 38(12), 2590–2597 RSC.
  37. L. Bernat, T. Jurtsch, G. Moryson, J. Moryson and G. Wiczynski, Recycling, 2024, 9(1), 14,  DOI:10.3390/recycling9010014.
  38. S. Soumyashree and P. Kumar, Spectrochim. Acta, Part B, 2023, 208, 106761,  DOI:10.1016/j.sab.2023.106761.
  39. Z. Zhang, W. Jia, Q. Shan, J. Zhang, D. Hei, A. Sun, M. Liao and Y. Ling, Spectrochim. Acta, Part B, 2023, 210, 106820,  DOI:10.1016/j.sab.2023.106820.
  40. S. L. Chen, R. H. Li and Y. Q. Chen, Appl. Opt., 2023, 62(17), 4512–4517 CrossRef CAS PubMed.
  41. M. Ivkovic, J. Savovic, B. D. Stankov, M. Kuzmanovic and I. Traparic, Spectrochim. Acta, Part B, 2024, 213, 106874,  DOI:10.1016/j.sab.2024.106874.
  42. N. Lv, Z. Bao, K. Chen, Y. Zhang and H. Yuan, At. Spectrosc., 2023, 44(6), 418–426 CrossRef CAS.
  43. D. Zhang, X. Niu, J. Nie, S. Shi, H. Ma and L. Guo, Opt. Express, 2024, 32(7), 10851–10861 CrossRef CAS PubMed.
  44. J. F. Nie, Y. Zeng, X. C. Niu, D. Zhang and L. B. Guo, J. Anal. At. Spectrom., 2023, 38(11), 2387–2395 RSC.
  45. R. Hedwig, I. Karnadi, M. Pardede, I. Tanra, A. M. Marpaung, M. M. Suliyanti, E. M. Huang, T. J. Lie, K. H. Kurniawan and K. Kagawa, Talanta Open, 2023, 8, 100251,  DOI:10.1016/j.talo.2023.100251.
  46. H. Chen, X. Shi, Y. He and W. Zhang, Spectrochim. Acta, Part B, 2023, 208, 106763,  DOI:10.1016/j.sab.2023.106763.
  47. J. Chen, Y. Ding, A. Hu, W. Chen, Y. Wang, M. Zhao and Y. Shu, Opt. Express, 2023, 31(25), 41129–41148 CrossRef CAS PubMed.
  48. M. Lanzinger, D. Huber, V. Merk, S. Kaufmann, M. Schuster and N. Ivleva, Spectrochim. Acta, Part B, 2023, 205, 106691,  DOI:10.1016/j.sab.2023.106691.
  49. D. Qu, G. Yang, X. Jin, G. Chen, Z. Bai, C. Li and D. Tian, Spectrochim. Acta, Part B, 2023, 209, 106794,  DOI:10.1016/j.sab.2023.106794.
  50. D. Diaz-Romero, S. Van den Eynde, I. Zaplana, C. C. Zhou, W. Sterkens, T. Goedeme and J. Peeters, Resour., Conserv. Recycl., 2023, 190, 106865,  DOI:10.1016/j.resconrec.2023.106865.
  51. Y. Ding, L. Yang, J. Chen, X. Wang, X. Guo, X. Xu, X. Zhao, Y. Luo and W. Chen, Laser Optoelectron. Prog., 2022, 59(13), 1314006,  DOI:10.3788/lop202259.1314006.
  52. Q. Ma, Z. Liu, T. Sun, X. Gao and Y. Dai, Opt. Express, 2023, 31(17), 27633–27653 CrossRef CAS PubMed.
  53. G. A. Wubetu, J. T. Costello, T. J. Kelly, P. Wachulak, A. Bartnik, W. Skrzeczanowski and H. Fiedorowicz, J. Appl. Spectrosc., 2023, 90, 116–125,  DOI:10.1007/s10812-023-01512-y.
  54. M. R. Khan, S. U. Haq, Q. Abbas and A. Nadeem, Spectrochim. Acta, Part B, 2023, 200, 106612,  DOI:10.1016/j.sab.2022.106612.
  55. K. A. Thorarinsdottir, M. Maghsoudi and K. Leosson, Appl. Spectrosc., 2024, 78(1), 67–75 CrossRef CAS PubMed.
  56. L. Pinjun, X. Zhengye, M. Zidong, G. Jingyuan and W. Chunxi, Open Phys., 2023, 21(1), 20230113,  DOI:10.1515/phys-2023-0113.
  57. G. Mukhametzianova, S. Wagner, M. Eskinja, M. Moshtaghi, G. Mori and T. Prohaska, Anal. Bioanal. Chem., 2024, 416, 3373–3388,  DOI:10.1007/s00216-024-05288-8.
  58. Z. Yang, L. Yang, L. Zhao, D. Huang, J. Li, X. Chen, X. Shen and H. Wang, Talanta, 2024, 269, 125407,  DOI:10.1016/j.talanta.2023.125407.
  59. Z. Weiss, J. Capek, Z. Kacenka, O. Ekrt, J. Kopecek, M. Losertova and D. Vojtech, J. Anal. At. Spectrom., 2024, 39(4), 996–1003 RSC.
  60. T. Zhao, Q. Zhong, S. Nie, X. Liu, H. Xiao, C. Yin, F. Zhong, Y. Ke and F. Li, Adv. Photonics Res., 2024, 5(7), 2300293,  DOI:10.1002/adpr.202300293.
  61. H. Choi, H. Kim, S.-H. Han, S. Kim, S. Jung, S.-H. Nam and Y. Lee, Appl. Spectrosc., 2023, 77(4), 371–381 CrossRef CAS PubMed.
  62. Y. Ikeda, J. K. Soriano and I. Wakaida, Talanta Open, 2023, 7, 100182,  DOI:10.1016/j.talo.2023.100182.
  63. Y. Ikeda, J. K. Soriano and I. Wakaida, J. Anal. At. Spectrom., 2023, 38, 1275–1284,  10.1039/d3ja00031a.
  64. Y. Li, Y. Chen, Y. Jiang, Z. Wang and X. Huang, Spectrochim. Acta, Part B, 2024, 213, 106862,  DOI:10.1016/j.sab.2024.106862.
  65. N. Ahmed, M. Saddique, M. Jelani, R. Ahmed, M. A. ul Haq, F. W. Khan, J. Iqbal, M. A. Baig, R. Hedwig and K. H. Kurniawan, Opt. Mater., 2023, 143, 114309,  DOI:10.1016/j.optmat.2023.114309.
  66. A. J. McCoy-West, A. M. Davis, A. N. Wainwright and A. G. Tomkins, J. Anal. At. Spectrom., 2024, 39, 780–790,  10.1039/d3ja00374d.
  67. W. Matsuda, A. Morikawa, A. Ohbuchi, T. Moriyama and T. Nakamura, Spectrochim. Acta, Part B, 2024, 212, 106856,  DOI:10.1016/j.sab.2023.106856.
  68. A. A. Arkhipenko, G. E. Marina, M. S. Doronina, N. A. Korotkova and V. B. Baranovskaya, Recycling, 2023, 8(6), 84,  DOI:10.3390/recycling8060084.
  69. A. J. Aller and F. J. Pereira, J. Anal. At. Spectrom., 2024, 39(2), 330–355 RSC.
  70. L. P. Santos, J. F. Q. Pereira and I. M. Raimundo Jr, J. Anal. At. Spectrom., 2024, 39(4), 1024–1029,  10.1039/d3ja00442b.
  71. S. L. Li, S. H. Gao, Z. R. Qian, Z. X. Wang, Y. Cao, D. H. Zhu, X. W. Zhu and W. F. Yang, Laser Phys., 2023, 33(10), 105701,  DOI:10.1088/1555-6611/acf042.
  72. S. Li, Y. Yang, S. Gao, D. Lin, G. Li, Y. Hu and W. Yang, Opt. Express, 2024, 32(3), 4122–4136 CrossRef CAS PubMed.
  73. A. Turner and M. Filella, Environ. Pollut., 2023, 316, 120492,  DOI:10.1016/j.envpol.2022.120492.
  74. X. Zhang, A. Li, Y. Zhang, Y. Yin, X. Wang, Y. He, J. Lyv, Y. Shan, X. Liu, W. Yi, L. Zhong, Y. Ren, M. Xia and R. Liu, J. Anal. At. Spectrom., 2024, 39(3), 974–981 RSC.
  75. X. S. Wang, Y. G. He, Y. Zhang, A. Li, X. Y. Zhang, X. Y. Guo, T. L. Zhang, W. Guo, R. B. Liu and Y. G. Yao, J. Anal. At. Spectrom., 2023, 38(11), 2405–2413 RSC.
  76. X. Wang, Y. He, X. Zhang, M. Hu, W. Zhao, H. Sun, X. Yang, X. Liu and R. Liu, Mater. Adv., 2024, 5(9), 3921–3928 RSC.
  77. Z. Chen, R. Zhai, Y. Cai, Y. Ye, Z. Sun and Y. Liu, J. Laser Appl., 2023, 35(4), 042031,  DOI:10.2351/7.0001226.
  78. E. Evgenakis, K. Giouri, A. Bourliva, C. Christophoridis, C. K. Zacharis and K. Fytianos, Anal. Lett., 2024, 57(7), 1038–1053 CrossRef CAS.
  79. A. C. M. Aleluia, M. d. S. Nascimento, A. M. P. dos Santos, W. N. L. dos Santos, A. d. F. Santos Jr and S. L. C. Ferreira, Spectrochim. Acta, Part B, 2023, 205, 106689,  DOI:10.1016/j.sab.2023.106689.
  80. D. V. Danilov, P. Y. Sharanov and N. V. Alov, Pharm. Chem. J., 2023, 57, 306–310,  DOI:10.1007/s11094-023-02881-6.
  81. N. V. Alov, D. V. Danilov, P. Y. Sharanov, V. G. Semenov, V. V. Panchuk, S. S. Savinov and M. M. Khaydukova, X-Ray Spectrom., 2024, 53, 181–189,  DOI:10.1002/xrs.3402.
  82. Y. Hao, G. Yin, X. Wang, J. Fu, Q. Cao, Q. Sun, G. Zhang and Z. Feng, J. Anal. Methods Chem., 2024, 2024, 9952318,  DOI:10.1155/2024/9952318.
  83. T. V. Shah, S. R. Sagar, S. Shah, A. R. Chowdhury, T. Mehta and D. G. Maheshwari, Chem. Pap., 2024, 78, 4661–4669,  DOI:10.1007/s11696-024-03385-7.
  84. C. M. Senger, I. S. Dornelles, J. M. Queiroz, P. A. Mello, E. I. Muller and A. L. H. Muller, J. Pharm. Biomed. Anal., 2024, 238, 115802,  DOI:10.1016/j.jpba.2023.115802.
  85. S. J. Abellan-Martin, J. Perez, F. C. Pinheiro, J. A. Nobrega, M. A. Aguirre, L. Vidal and A. Canals, Adv. Sample Prep., 2023, 7, 100084,  DOI:10.1016/j.sampre.2023.100084.
  86. I. F. P. C. Noronha, F. C. Pinheiro and C. C. Nascentes, Talanta, 2024, 271, 125604,  DOI:10.1016/j.talanta.2023.125604.
  87. M. Horstmann, M. Austrup, F. Busch, A. Faust, M. Sperling, U. Karst and D. Clases, J. Anal. At. Spectrom., 2023, 38(10), 2038–2045 RSC.
  88. M. B. De Maria, M. Matczuk, D. Tesauro, M. Saviano, J. Sikorski, G. Chiappetta, S. Godin, J. Szpunar, R. Lobinski and L. Ronga, Anal. Bioanal. Chem., 2024, 416, 2819–2833,  DOI:10.1007/s00216-024-05140-z.
  89. K. Degardin, A. Guillemain, A. Feng, C. Saladin, S. Paratte, R. Zurbach and T. Mohn, Talanta Open, 2024, 9, 100295,  DOI:10.1016/j.talo.2024.100295.
  90. L. Yao, J. Li, Y. Liu, L. Shen and C. Wang, J. Anal. At. Spectrom., 2024, 39, 735–742,  10.1039/d3ja00333g.
  91. O. Gyamfi, J. Aboko, E. Ankapong, J. T. Marfo, N. Y. Awuah-Boateng, K. Sarpong and E. Dartey, Cutan. Ocul. Toxicol., 2024, 43, 5–12,  DOI:10.1080/15569527.2023.2268197.
  92. J. C. Garcia-Mesa, I. Morales-Benitez, P. Montoro-Leal, M. M. L. Guerrero and E. I. V. Alonso, Talanta, 2024, 268, 125360,  DOI:10.1016/j.talanta.2023.125360.
  93. D. Hebert, J. Nelson, B. N. Diehl and P. Zito, Nanomaterials, 2023, 13(19), 2681,  DOI:10.3390/nano13192681.
  94. B. Liu, L.-z. Zhou, G.-w. He and C. Wang, Anal. Sci., 2024, 40, 115–121,  DOI:10.1007/s44211-023-00437-y.
  95. C. Guerranti, C. Andreuccetti, I. Cenni, S. Santini, T. Martellini, A. Cincinelli and A. Perico, Anal. Lett., 2023, 56(7), 1115–1127 CrossRef CAS.
  96. K. R. N. da Silva, H. S. Pereira, V. A. do Nascimento and J. L. Raposo Jr, Fuel, 2024, 359, 130336,  DOI:10.1016/j.fuel.2023.130336.
  97. S. Zhu, G. Song, W. Zhang, Y. Zhang, Y. Wei, Q. Zhang, D. Chen, J. Li and T. Sun, J. Anal. At. Spectrom., 2024, 39, 841–853,  10.1039/d3ja00324h.
  98. W. Lu, Q. Yang, Z. Yu, X. Zou, H. Qin, G. Chen, W. Ma, J. Zhuo, C. Chen, X. Chen, S. Yao and J. Lu, Microw. Opt. Technol. Lett., 2024, 66(1), e33991,  DOI:10.1002/mop.33991.
  99. M. Dong, J. Cai, H. Liu, J. Xiong, G. Rao, S. Yao and J. Lu, Spectrochim. Acta, Part B, 2023, 210, 106807,  DOI:10.1016/j.sab.2023.106807.
  100. Z. Damdinsuren, P. Zuzaan and B. Damdinsuren, X-Ray Spectrom., 2023, 53, 153–158,  DOI:10.1002/xrs.3399.
  101. S. J. Abellan-Martin, M. A. Aguirre and A. Canals, J. Anal. At. Spectrom., 2023, 38, 1379–1386,  10.1039/d3ja00093a.
  102. J. Castillo, G. Gonzalez, M. Menta, J. Jimenez-Lamana and B. Bouyssiere, Energy Fuels, 2023, 37(17), 12696–12703 CrossRef CAS.
  103. J. S. Almeida, S. G. Domingos, L. A. Meira, J. S. R. Teixeira and L. S. G. Teixeira, Quim. Nova, 2024, 47, e-20240018,  DOI:10.21577/0100-4042.20240018.
  104. F. Endriss, D. Kuptz, D. Wissmann, H. Hartmann, E. Dietz, A. Kappler and H. Thorwarth, Renewable Energy, 2024, 222, 119923,  DOI:10.1016/j.renene.2023.119923.
  105. F. Endriss, D. Kuptz, H. Hartmann, S. Brauer, R. Kirchhof, A. Kappler and H. Thorwarth, Chem. Ing. Tech., 2023, 95(10), 1503–1525 CrossRef CAS.
  106. J. Yang, Y.-P. Xu, P. Chen, J.-Y. Li, D. Liu and X.-L. Chu, J. Clean. Prod., 2023, 431, 139771,  DOI:10.1016/j.jclepro.2023.139771.
  107. X. Wu, Z. Wu, M. Amde, W. Zhao, G. Li, S. Gao, M. Song, Y. Ma, Z. Tan, Y. Yin and Q. Zhang, At. Spectrosc., 2023, 44(5), 282–297 CrossRef CAS.
  108. D. Holub, J. Buday, P. Porizka and J. Kaiser, Spectrochim. Acta, Part B, 2023, 207, 106752,  DOI:10.1016/j.sab.2023.106752.
  109. A. Turner and M. Filella, Sci. Total Environ., 2023, 903, 166083,  DOI:10.1016/j.scitotenv.2023.166083.
  110. C. Chaine, A. S. Hursthouse, I. McLellan, E. Viza and J. Miller, Recycling, 2023, 8(5), 71,  DOI:10.3390/recycling8050071.
  111. A. A. Cuthbertson, C. Lincoln, J. Miscall, L. M. Stanley, A. K. Maurya, A. S. Asundi, C. J. Tassone, N. A. Rorrer and G. T. Beckham, Green Chem., 2024, 26, 7067–7090,  10.1039/d4gc00659c.
  112. A. Sakanupongkul, K. Sirisinha, R. Saenmuangchin and A. Siripinyanond, Microchem. J., 2024, 199, 110016,  DOI:10.1016/j.microc.2024.110016.
  113. S. Harycki and A. Gundlach-Graham, Anal. Chem., 2023, 95(41), 15318–15324 CrossRef CAS PubMed.
  114. T. Vonderach, A. Gundlach-Graham and D. Guenther, Anal. Bioanal. Chem., 2024, 416(11), 2773–2781 CrossRef CAS PubMed.
  115. T. Van Acker, A. Rua-Ibarz, F. Vanhaecke and E. Bolea-Fernandez, Anal. Chem., 2023, 95(50), 18579–18586 CrossRef CAS PubMed.
  116. L. Hendriks and D. M. Mitrano, Environ. Sci. Technol., 2023, 57(18), 7263–7272 CrossRef CAS PubMed.
  117. L. Hendriks, V. M. Kissling, T. Buerki-Thurnherr and D. M. Mitrano, Environ. Sci.: Nano, 2023, 10, 3439–3449,  10.1039/d3en00681f.
  118. P. Li, C. J. He and D. H. Lin, Environ. Pollut., 2022, 314, 120267,  DOI:10.1016/j.envpol.2022.120267.
  119. C. Smith, S. Brown, N. Malone, S. Bevers, J. Ranville and D. H. Fairbrother, Environ. Sci.: Nano, 2024, 11(3), 911–923 RSC.
  120. M. Baalousha, J. Wang, M. M. Nabi, M. Alam, M. Erfani, J. Gigault, F. Blancho, M. Davranche, P. M. Potter and S. R. Al-Abed, Environ. Sci.: Nano, 2024, 11(1), 373–388 RSC.
  121. P. S. Vaisakh, U. K. Adarsh, K. Amrutha, A. K. Warrier, V. B. Kartha and V. K. Unnikrishnan, Environ. Res., 2023, 231, 116198,  DOI:10.1016/j.envres.2023.116198.
  122. S. Carneado, J. F. Lopez-Sanchez and A. Sahuquillo, Molecules, 2023, 28(20), 7166,  DOI:10.3390/molecules28207166.
  123. M. de Bruin-Hoegée, J. Schoorl, P. Zoon, M. J. V. Schans, D. Noort and A. C. V. Asten, Forensic Chem., 2023, 35, 100515,  DOI:10.1016/j.forc.2023.100515.
  124. H. Komatsu, T. Nakanishi, Y. Seto and Y. Nishiwaki, Spectrochim. Acta, Part B, 2023, 209, 106785,  DOI:10.1016/j.sab.2023.106785.
  125. J. Willner, L. Brunnbauer, C. D. Quarles Jr, M. Nelhiebel, S. Larisegger and A. Limbeck, J. Anal. At. Spectrom., 2023, 38(10), 2028–2037 RSC.
  126. M. Gozzi, P. Gaines, T. Kozikowski and B. Alexander, Anal. Chem., 2023, 95, 8578–8585,  DOI:10.1021/acs.analchem.3c00728.
  127. J. Zhang, Y. Cui, T. Zhou, J. Wang, Z. Zhu, F. Liu, G. Xiao, J. Yu, X. Wang, B. Huang, C. Li and X. Huang, Accred Qual. Assur., 2023, 28, 299–310,  DOI:10.1007/s00769-023-01557-8.
  128. J. Schroder, J. A. Zamora Zeledon, G. A. Kamat, M. E. Kreider, L. Wei, A. S. Mule, A. Torres, K. Yap, D. Sokaras, A. Gallo, M. B. Stevens and T. F. Jaramillo, ACS Energy Lett., 2023, 8(7), 2962–2969 CrossRef CAS.
  129. L. Lahn, A. M. Mingers, A. Savan, A. Ludwig and O. Kasian, Chemelectrochem, 2024, 11, e202300399,  DOI:10.1002/celc.202300399.
  130. J. M. Linge, V. Briega-Martos, A. Hutzler, B. Fritsch, H. Erikson, K. Tammeveski and S. Cherevko, ACS Appl. Energy Mater., 2023, 6(22), 11497–11509 CrossRef CAS PubMed.
  131. A. Kormanyos, P. Buettner, M. Bosch, M. Minichova, A. Koerner, K. J. Jenewein, A. Hutzler, K. J. J. Mayrhofer, J. Bachmann and S. Cherevko, ACS Mater. Au, 2024, 4, 286–299,  DOI:10.1021/acsmaterialsau.3c00092.
  132. H. Li, X. Weng, Y. Kang, H. Lei, Y. Li, C. Zhou, R. Huang, Y. Kong, T. Liu, W. Wei, Z. Gong, D. Ding, Z. Chen and Y. Cui, Appl. Surf. Sci., 2024, 644, 158734,  DOI:10.1016/j.apsusc.2023.158734.
  133. O. Martinez-Mora, L. F. Leon-Fernandez, M. Velimirovic, F. Vanhaecke, K. Tirez, J. Fransaer and X. Dominguez-Benetton, Mater. Adv., 2023, 4(23), 6183–6191 RSC.
  134. A. Rua-Ibarz, T. Van Acker, E. Bolea-Fernandez, M. Boccongelli and F. Vanhaecke, J. Anal. At. Spectrom., 2024, 39, 888–899,  10.1039/d3ja00271c.
  135. S. Strekopytov, J. Entwisle, S. Hill and H. Goenaga-Infante, J. Anal. At. Spectrom., 2024, 39, 868–878,  10.1039/d3ja00416c.
  136. E. Zambrzycka-Szelewa and B. Godlewska-Zylkiewicz, Spectrochim. Acta, Part B, 2024, 213, 106859,  DOI:10.1016/j.sab.2024.106859.
  137. A. Wach, X. Zou, K. Wojtaszek, Y. Kayser, C. Garlisi, G. Palmisano, J. Sa and J. Szlachetko, X-Ray Spectrom., 2023, 52, 261–268,  DOI:10.1002/xrs.3363.
  138. A. Kazlagic, M. Rosner, A. Cipriani, D. A. Frick, J. Glodny, E. J. Hoffmann, J. M. Hora, J. Irrgeher, F. Lugli, T. Magna, T. C. Meisel, A. Meixner, A. Possolo, A. Pramann, M. J. Pribil, T. Prohaska, A. Retzmann, O. Rienitz, D. Rutherford, G. M. Paula-Santos, M. Tatzel, S. Widhalm, M. Willbold, T. Zuliani and J. Vogl, Geostand. Geoanal. Res., 2023, 47(4), 821–840 CrossRef CAS.
  139. L. Javier Fernandez-Menendez, C. Mendez-Lopez, C. Gonzalez-Gago, J. Pisonero and N. Bordel, J. Anal. At. Spectrom., 2023, 38(2), 325–332 RSC.
  140. Y. Qiu, X. Guo, M. Shi, Y. Zhou, J. Wu, J. Li, H. Sun, Z. Zhang, Y. Hang, X. Li and Y. Li, Spectrochim. Acta, Part B, 2023, 209, 106799,  DOI:10.1016/j.sab.2023.106799.
  141. Y. Cai, X. Ma, B. Huang, R. Zhang and X. Wang, Appl. Opt., 2024, 63(6), A24–A31 CrossRef CAS PubMed.
  142. C. Zhang, W. Song, Z. Hou and Z. Wang, J. Anal. At. Spectrom., 2023, 38(12), 2554–2561 RSC.
  143. N. Mijatovic, M. Vasic, A. Terzic, M. Radomirovic and L. Milic, Spectrochim. Acta, Part B, 2023, 209, 106796,  DOI:10.1016/j.sab.2023.106796.
  144. A. Dey, T. L. Cavalline, M. Mamirov and J. Hu, ACI Mater. J., 2023, 120(5), 53–66,  DOI:10.14359/51738890.
  145. I. Ben Amar, A. Thomas, C. Bachmann, A. Hafnoaui, H. Griguer, A. Miled and Y. Messaddeq, Sci. Rep., 2023, 13(1), 17925 CrossRef CAS PubMed.
  146. J. Park, S.-H. Han, S.-H. Nam and Y. Lee, Appl. Spectrosc., 2023, 77(12), 1351–1361 CrossRef CAS PubMed.
  147. H.-B. Choi, S.-H. Moon, H. Kim, N. Guthikonda, K.-S. Ham, S.-H. Han, S.-H. Nam and Y.-H. Lee, Sensors, 2023, 23(22), 9280,  DOI:10.3390/s23229280.
  148. S. Nisar, A. Alsalme, R. Zulfiqar, M. Rizwan, D.-K. Kim, G. Dastgeer and Z. M. Shazad, Mod. Phys. Lett. B, 2024, 38(15), 2450124,  DOI:10.1142/s0217984924501240.
  149. F. Degryse, H. Harris, R. Baird, I. Andelkovic, R. C. da Silva, S. Kabiri, A. Yazici, I. Cakmak and M. J. McLaughlin, Plant Soil, 2024, 495(1–2), 27–41 CrossRef CAS.
  150. Q. Shan, Y. Luo, X. Zhang, Z. Zhang, W. Jia, Y. Ling, D. Hei, J. Zhang, C. Shi and J. Bao, Anal. Lett., 2024, 57, 2694–2703,  DOI:10.1080/00032719.2023.2301045.
  151. G. M. Faria, T. D. Galva, P. S. Parreira and F. L. Melquiades, Appl. Radiat. Isot., 2023, 200, 110964,  DOI:10.1016/j.apradiso.2023.110964.
  152. A. Pacella, P. Ballirano, M. C. Di Carlo, M. Fantauzzi, A. Rossi, E. Nardi, C. Viti, L. Arrizza, A. Campopiano, A. Cannizzaro, A. Bloise and M. R. Montereali, Nanomaterials, 2023, 13(22), 2933,  DOI:10.3390/nano13222933.
  153. K. Mori, T. Hourai, T. Matsuyama, S. Zhuo and K. Tsuji, Anal. Sci., 2024, 40, 367–373,  DOI:10.1007/s44211-023-00464-9.
  154. M. F. Gazulla, M. Orduna and M. Rodrigo, X-Ray Spectrom., 2024 DOI:10.1002/xrs.3418.
  155. M. Kobayashi, H. Namiki and H. Hayashi, Anal. Sci., 2023, 39, 1757–1762,  DOI:10.1007/s44211-023-00390-w.
  156. I. V. Zagaynov and A. K. Buryak, Ceram. Int., 2023, 49(23), 39729–39735 CrossRef CAS.
  157. J. Hao, J. Chen, S. Wang and T. Liang, Mater. Chem. Phys., 2023, 309, 128370,  DOI:10.1016/j.matchemphys.2023.128370.
  158. Y. He, Q. Wen, C. Ke, Z. Qian, D. Zhao, W. Li, M. Xu, T. Xu, Y. Zhang and Y. Zhao, Ceram. Int., 2023, 49(22), 35802–35811 CrossRef CAS.
  159. Q.-T. Tran, M.-D. Ho, T.-N. Nguyen, T.-A. Tran, V.-D. Ho, T.-S. Truong and P. Lathdavong, J. Radioanal. Nucl. Chem., 2023, 332(8), 3493–3498 CrossRef CAS.
  160. C. Le Losq, M. Tarrago, W. Blanc, P. Georges, L. Hennet, D. Zanghi, G. Lelong, S. Pannefieu and D. R. Neuville, Mater. Tech., 2022, 110(4), 403,  DOI:10.1051/mattech/2022041.
  161. G. Pontes, R. Schneider, G. S. Senesi and G. Nicolodelli, J. Non-Cryst. Solids, 2024, 627, 122814,  DOI:10.1016/j.jnoncrysol.2023.122814.
  162. P. Gadoros, I. Peczeli, I. Kocsanyi and P. Richter, J. Laser Micro/Nanoeng., 2023, 18(1), 51–57 CAS.
  163. Z. Cernosek, M. Chladkova and J. Holubova, J. Solid State Chem., 2022, 314, 123366,  DOI:10.1016/j.jssc.2022.123366.
  164. Z. Cernosek and J. Holubova, J. Non-Cryst. Solids, 2024, 623, 122694,  DOI:10.1016/j.jnoncrysol.2023.122694.
  165. J. Sulowska, M. Szumera, A. Berezicka, Z. Olejniczak, J. Stepien, S. Pollastri and L. Olivi, J. Alloys Compd., 2024, 976, 172995,  DOI:10.1016/j.jallcom.2023.172995.
  166. K. Mervic, J. T. van Elteren, M. Bele and M. Sala, Talanta, 2024, 269, 125379,  DOI:10.1016/j.talanta.2023.125379.
  167. R.-Y. Li, J. Hao, W. Yang, H.-C. Tian, S. Hu and Y. Lin, J. Anal. At. Spectrom., 2023, 38(10), 1962–1972,  10.1039/d3ja00100h.
  168. S.-S. Shin, G.-Y. Kim, B. Lee, J.-H. Yang, Y.-S. Son, J.-H. Choi, J.-Y. Pyo, K. R. Lee, H.-S. Park and H. W. Kang, J. Non-Cryst. Solids, 2024, 624, 122728,  DOI:10.1016/j.jnoncrysol.2023.122728.
  169. I. I. Evdokimov, A. E. Kurganova and A. P. Velmuzhov, J. Anal. Chem., 2023, 78(5), 644–651 CrossRef CAS.
  170. G. D. Soria, M. Gonzalez, M. L. Crespillo, J. S. Prieto and G. Garcia, Surf. Interface Anal., 2024, 56, 372–381,  DOI:10.1002/sia.7298.
  171. S. J. Shetty, M. Veis, S. Atikukke, P. Barton, M. Jerab, R. Dejarnac and P. Veis, Nucl. Mater. Energy, 2023, 37, 101547,  DOI:10.1016/j.nme.2023.101547.
  172. Y. X. He, Q. F. Wen, C. Ke, Z. R. Qian, D. Y. Zhao, W. X. Li, M. Xu, T. Xu, Y. Zhang and Y. Zhao, Ceram. Int., 2023, 49(22), 35802–35811 CrossRef CAS.
  173. H. Schamis, C. Jaramillo-Correa, M. Parsons, T. Marchhart, J. P. Allain, C. Hargrove, A. Judge, K. Kawashimo, E. Kunz, C. Lopez Perez, M. Nieto-Perez and H. Shober, Rev. Sci. Instrum., 2024, 95(4), 043508,  DOI:10.1063/5.0165857.
  174. D. Nishijima, M. Tokitani, D. Nagata, T. Schwarz-Selinger, A. Zaloznik, F. Chang, R. P. Doerner, M. I. Patino, M. J. Simmonds, M. J. Baldwin and G. R. Tynan, Nucl. Fusion, 2023, 63(12), 126003,  DOI:10.1088/1741-4326/acf509.
  175. E. D. Marenkov, I. P. Tsygvintsev, D. A. Kim, A. S. Grushin, N. E. Efimov, D. N. Sinelnikov and Y. M. Gasparyan, Spectrochim. Acta, Part B, 2023, 210, 106822,  DOI:10.1016/j.sab.2023.106822.
  176. J. Liu, D. Wu, D. Zhao, S. Liu, X. Hu, C. Li, L. Cai and H. Ding, Fusion Eng. Des., 2023, 195, 113930,  DOI:10.1016/j.fusengdes.2023.113930.
  177. Z. Hu, H. Wu, X. Bai, F. Ding, M. Imran, G.-N. Luo, C. Li, R. Hai and H. Ding, Fusion Eng. Des., 2023, 197, 114078,  DOI:10.1016/j.fusengdes.2023.114078.
  178. Y. Chen, J. Gao, T. Long, L. Nie, J. Gao, Y. Ma, Y. Huang, W. Tian, Y. Liu, X. Zhu, G. Zhuang, W. Zhong and M. Xu, Plasma Sci. Technol., 2024, 26(3), 034004,  DOI:10.1088/2058-6272/ad0c95.
  179. M. Imran, Z. Hu, F. Ding, H. Sattar and G.-N. Luo, Appl. Phys. B: Lasers Opt., 2024, 130(2), 27,  DOI:10.1007/s00340-023-08165-x.
  180. Z. He, Y. Lyu, D. Wu, R. Hai, X. Bai, H. Wu and H. Ding, Nucl. Mater. Energy, 2023, 36, 101493,  DOI:10.1016/j.nme.2023.101493.
  181. P. G. Bhat, P. Veis, A. M. Roldan, J. Karhunen, P. Paris, I. Jogi, A. Hakola, S. Almaviva, W. Gromelski, M. Ladygina, P. Gasior, J. Ristkok, I. B. Radovic, Z. Siketic, O. Romanenko, C. Porosnicu, J. Likonen and C. Lungu, Nucl. Mater. Energy, 2023, 37, 101549,  DOI:10.1016/j.nme.2023.101549.
  182. I. Jogi, J. Ristkok, J. Butikova, J. Raud and P. Paris, Nucl. Mater. Energy, 2023, 37, 101543,  DOI:10.1016/j.nme.2023.101543.
  183. Z. X. Cui, H. Wu, D. Wu, J. M. Liu, X. L. Hu, R. Hai, C. Li and H. B. Ding, Acta Photonica Sin., 2023, 52(9), 0930002,  DOI:10.3788/gzxb20235209.0930002.
  184. H. Wu, D. Wu, C. Li, S. Liu, Z. Hu, Y. Li, X. Wang, X. Bai, R. Hai and H. Ding, Spectrochim. Acta, Part B, 2023, 207, 106755,  DOI:10.1016/j.sab.2023.106755.
  185. C. Li, Q. Li, L. Li, B. Men, H. Wu, D. Wu, R. Hai, X. Wu and H. Ding, Spectrochim. Acta, Part B, 2023, 206, 106735,  DOI:10.1016/j.sab.2023.106735.
  186. Z. H. Hu, F. Ding, M. Imran, G. N. Luo, C. Li, R. Hai and H. B. Ding, Nucl. Mater. Energy, 2023, 37, 101542,  DOI:10.1016/j.nme.2023.101542.
  187. S. Atikukke, M. Veis, W. Khan, E. Grigore, F. Baiasu, P. Durina, T. Roch, P. Dvorak and P. Veis, Nucl. Mater. Energy, 2024, 38, 101558,  DOI:10.1016/j.nme.2023.101558.
  188. M. Trtica, J. Savovic, M. Kuzmanovic, D. Rankovic and J. Stasic, J. Appl. Spectrosc., 2024, 90(6), 1318–1324 CrossRef CAS.
  189. M. L. Sankhe, A. Favre, J.-B. Sirven, A. Bultel, D. L'Hermite, A. Semerok, S. Vartanian and C. Grisolia, Fusion Eng. Des., 2023, 197, 114077,  DOI:10.1016/j.fusengdes.2023.114077.
  190. N. T. Djigailo, S. I. Kravchuk, V. B. Lazarev and N. P. Petrova, Phys. At. Nucl., 2023, 86(7), 1702–1711 CrossRef CAS.
  191. G. Zhao, Y. Shao, M. Luo, D. Xu, D. Li, Z. Liu and L. Ma, J. Environ. Radioact., 2023, 270, 107313,  DOI:10.1016/j.jenvrad.2023.107313.
  192. E. I. Holland, Y. Verbelen, D. T. Connor, T. Martin, M. Higginson and T. B. Scott, Sustainability, 2023, 15(7), 6178,  DOI:10.3390/su15076178.
  193. J. Lorincik, K. Sihelska, D. Vesela, I. Elantyev, K. Rezankova, C. Aparicio, T. M. Do, J. Kucera, J. John, M. Nemec, M. Mindova, V. Miler, D. Gotz, J. Janda and A. Kelnarova, J. Radioanal. Nucl. Chem., 2024, 333, 3675–3684,  DOI:10.1007/s10967-023-09336-y.
  194. A. Serban, F. Albota, E. Ionuz, C. S. Tuta and M. Virgolici, J. Radioanal. Nucl. Chem., 2024, 333, 165–173,  DOI:10.1007/s10967-023-09189-5.
  195. K. L. LeBlanc, K. Nadeau, J. Meija, L. Yang, P. A. Babay, M. A. Bavio, C. Boome, D. Chipley, R. S. C. Leguizamon, J. Denton, D. L. Drew, M. A. Fernandez, V. Fugaru, V. D. Genetti, F. Gonzalez, J. D. Inglis, S. Jovanovic, E. Keegan, T. Kell, Y. Kimura, W. Kinman, S. Kiser, R. E. Lindvall, E. Loi, K. Mayer, J.-F. Mercier, R. Millar, A. Nicholl, L. Orlovskaya, J. L. Ramella, A. Serban, M. A. Sharp, Y. Shi, C. Tobi, L. Valenzuela, Z. Varga, A. Vesterlund, M. Virgolici, H. Yamazaki, E. N. Zubillaga, A. El-Jaby and Z. Mester, J. Radioanal. Nucl. Chem., 2022, 331(9), 4031–4045 CrossRef CAS.
  196. L. Qiu and R. G. Cooks, J. Am. Soc. Mass Spectrom., 2024, 35(3), 642–645 CrossRef CAS PubMed.
  197. J. D. Inglis, K. M. Samperton, A. A. Reinhard, T. Tenner, R. E. Steiner and S. P. Lamont, J. Radioanal. Nucl. Chem., 2023, 332(8), 3371–3381 CrossRef CAS.
  198. R. Park, C.-G. Lee, K. H. Chung and J. Park, J. Radioanal. Nucl. Chem., 2023, 332(7), 2833–2840 CrossRef CAS.
  199. K. Fenclova, M. Nemec and T. Prasek, Chem. Listy, 2023, 117(2), 128–133 CrossRef CAS.
  200. E. H. Kwapis, J. Borrero, K. S. Latty, H. B. Andrews, S. S. Phongikaroon and K. C. Hartig, Appl. Spectrosc., 2024, 78(1), 9–55 CrossRef CAS PubMed.
  201. Z. Varga, A. Nicholl, M. Wallenius and K. Mayer, J. Radioanal. Nucl. Chem., 2024, 333, 3553–3558,  DOI:10.1007/s10967-023-09229-0.
  202. M. Susset, A.-C. Humbert, V. Granger, F. Pointurier and C. Pecheyran, J. Radioanal. Nucl. Chem., 2024, 333, 877–888,  DOI:10.1007/s10967-023-09322-4.
  203. N. A. Zirakparvar, B. Manard, C. Hexel, D. Dunlap, S. Metzger, D. Bostick, V. Bradley and B. Ticknor, Int. J. Mass Spectrom., 2023, 492, 117114,  DOI:10.1016/j.ijms.2023.117114.
  204. V. C. Bradley, J. Burleson, H. B. Andrews, C. V. Thompson, T. L. Spano, D. R. Dunlap, N. A. Zirakparvar, B. W. Ticknor, C. R. Hexel and B. T. Manard, Analyst, 2024, 149, 2244–2251,  10.1039/d3an02101g.
  205. H. Jaegler and A. Gourgiotis, J. Anal. At. Spectrom., 2023, 38(10), 1914–1919 RSC.
  206. Q. R. Shollenberger, J. M. Rolison, M. A. Sharp and N. E. Marks, J. Nucl. Mater., 2023, 584, 154583,  DOI:10.1016/j.jnucmat.2023.154583.
  207. W. Wang, J. Xu, R. Xi, S. Guo, Y. Su, S. Fang, H. Zhang, Y. Wang, J. Fan, L. Feng, Y. Wang and Z. Li, J. Anal. At. Spectrom., 2024, 39(1), 178–189 RSC.
  208. K. Sanyal and S. Dhara, Crit. Rev. Anal. Chem., 2024, 1–16,  DOI:10.1080/10408347.2024.2316234.
  209. M. A. Boggs, A. A. Reinhard, J. Inglis, J. Tutt, G. Lee and C. Margiotta, J. Radioanal. Nucl. Chem., 2023, 332(10), 4157–4165 CrossRef CAS.
  210. P. Goswami, K. S. Bhushan and P. G. Jaison, J. Am. Soc. Mass Spectrom., 2024, 35, 935–942,  DOI:10.1021/jasms.4c00004.
  211. J. V. Goodwin, B. T. Manard, B. W. Ticknor, P. Cable-Dunlap and R. K. Marcus, Microchem. J., 2024, 196, 109645,  DOI:10.1016/j.microc.2023.109645.
  212. C. Nandi, S. Kesari, R. Phatak, K. Sanyal, R. Rao and A. Prakash, J. Nucl. Mater., 2024, 590, 154892,  DOI:10.1016/j.jnucmat.2023.154892.
  213. A. C. S. de Souza and L. R. P. de Andrade Lima, Nucl. Sci. Eng., 2024, 198(5), 1051–1061 CrossRef.
  214. Y. Suzuki, T. Watanabe and T. Goto, Nucl. Eng. Des., 2023, 415, 112699,  DOI:10.1016/j.nucengdes.2023.112699.
  215. P. Wu, L. Wang, J. Wang, J. Luo, Y. Lu, X. Song, J. Liu, Y. Qin, L. Hou and J. Ma, RSC Adv., 2024, 14(10), 7031–7039 RSC.
  216. D. Diaz and D. W. Hahn, Appl. Spectrosc., 2023, 77(9), 1033–1043 CrossRef CAS PubMed.
  217. H. B. Andrews and J. McFarlane, Sensors, 2023, 23(24), 9797,  DOI:10.3390/s23249797.
  218. D. Yan, X. Cao and K. Shen, J. Nucl. Mater., 2024, 588, 154814,  DOI:10.1016/j.jnucmat.2023.154814.
  219. S. Shalnev, S. Krzack, F. Kuester, H.-J. Friedrich, B. Meyer and M. Grabner, Fuel, 2024, 357, 129756 CrossRef CAS.
  220. I. Llopart-Babot, M. Vasile, A. Dobney, B. Russell, S. Kolmogorova, S. Boden, M. Bruggeman, M. Leermakers, J. Qiao, V. De Souza, A. Tarancon, H. Bagan and P. Warwick, Appl. Radiat. Isot., 2023, 202, 111046,  DOI:10.1016/j.apradiso.2023.111046.
  221. S. Bartalini, D. Mazzotti, G. Insero, M. Inguscio and P. De Natale, Nuovo Cimento C, 2023, 46(2), 41,  DOI:10.1393/ncc/i2023-23041-x.
  222. L. Li, Z. Qian, H. Zhang, X. Duan, X. Liu, Q. Zhang and Y. Qiao, J. Radioanal. Nucl. Chem., 2023, 332, 5035–5043,  DOI:10.1007/s10967-023-09204-9.
  223. E. V. Chukhlantseva, O. V. Stepanova, O. Y. Dichenko, T. A. Zaitseva, K. A. Dzhevello, E. L. Murashova, Y. M. Tatarnikova, E. A. Belanova and P. A. Blokhin, J. Anal. Chem., 2023, 78(12), 1701–1711 CrossRef CAS.
  224. M. Dano, M. Nemec, M. Mindova, J. John, E. Viglasova and B. Cziferi, Chem. Listy, 2023, 117(2), 114–121 CrossRef CAS.
  225. K. P. Hobbs, A. D. French, K. M. Melby, E. J. Bylaska, K. Harouaka, R. M. Cox, I. J. Arnquist and C. L. Beck, Anal. Chem., 2024, 96(15), 5807–5814 CrossRef CAS PubMed.
  226. H. Kazama, K. Konashi, T. Suzuki, S. Koyama, K. Maeda, Y. Sekio, T. Onishi, C. Abe, Y. Shikamori and Y. Nagai, J. Anal. At. Spectrom., 2023, 38(8), 1676–1681 RSC.
  227. I. Papp, N. Vajda, E. Bokori and Z. Molnar, J. Radioanal. Nucl. Chem., 2024, 333, 3639–3653,  DOI:10.1007/s10967-023-09350-0.
  228. S. Morii, T. Yomogida, S. Asai, K. Ouchi, T. Oka and Y. Kitatsuji, Bunseki Kagaku, 2023, 72(10–11), 441–448 CrossRef CAS.
  229. S. Tanaka, N. Tezuka, T. Taniuchi, H. Ueda and T. Sakuragi, MRS Adv., 2024, 9, 479–482,  DOI:10.1557/s43580-024-00788-2.
  230. V.-K. Do, T. Furuse, Y. Ohta, H. Iwahashi, T. Hirosawa, M. Watanabe and S. Sato, J. Radioanal. Nucl. Chem., 2022, 331(12), 5631–5640 CrossRef CAS.
  231. A. G. Shard, D. R. Baer and C. A. Clifford, Surf. Interface Anal., 2024, 56, 305–307,  DOI:10.1002/sia.7284.
  232. X. Li, R. Zhang, D. H. Tian, L. Cui, Y. Zhang, X. Li, Y. M. Lai and Y. L. Wang, Coatings, 2023, 13(10), 1773,  DOI:10.3390/coatings13101773.
  233. P. S. Boone, S. Harris, C. Ramon, J. Shusterman, A. J. Pascall and T. Parsons-Davis, Mater. Lett., 2024, 361, 136089,  DOI:10.1016/j.matlet.2024.136089.
  234. A. Krolicka, A. Maj and G. Loj, Materials, 2023, 16(20), 6641,  DOI:10.3390/ma16206641.
  235. S. Melhem, Y. Menesguen, E. Nolot and M.-C. Lepy, X-Ray Spectrom., 2023, 52(6), 412–422 CrossRef CAS.
  236. P. Jonnard, K. Le Guen, R. Delaunay, Y. Menesguen, M-C. Lepy, E. Briand, D. Schmaus and I. Vickridge, X-Ray Spectrom., 2023, 52(11), 437–446 CrossRef CAS.
  237. S. Aoyagi, J. Vac. Sci. Technol., A, 2023, 41(6), 063202,  DOI:10.1116/6.0002858.
  238. Y. H. Zhao, S. K. Otto, T. Lombardo, A. Henss, A. Koeppe, M. Selzer, J. Janek and B. Nestler, ACS Appl. Mater. Interfaces, 2023, 15(43), 50469–50478 CrossRef CAS PubMed.
  239. S. E. Bamford, D. Yalcin, W. Gardner, D. A. Winkler, T. M. Kohl, B. W. Muir, S. Howard, E. A. Bruton and P. J. Pigram, Anal. Chem., 2023, 95(20), 7968–7976 CrossRef CAS PubMed.
  240. S. E. Bamford, R. T. Jones, W. Gardner, B. W. Muir, D. A. Winkler and P. J. Pigram, Adv. Mater. Interfaces, 2024, 11, 2300645,  DOI:10.1002/admi.202300645.
  241. K. L. Bassett, T. Watkins, J. Coleman, N. Bianco, L. S. Bailey, J. Pillars, S. G. Williams, T. F. Babuska, J. Curry, F. W. DelRio, A. A. Henriksen, A. Garland, J. Hall, B. A. Krick and B. L. Boyce, Integr. Mater. Manuf. Innov., 2023, 12, 430–440,  DOI:10.1007/s40192-023-00315-5.
  242. S. Muramoto, D. J. Graham and D. G. Castner, J. Vac. Sci. Technol., A, 2024, 42(2), 023416,  DOI:10.1116/6.0003249.
  243. W. Filali, M. Boubaaya, E. Garoudja, F. Lekoui, I. Abdellaoui, R. Amrani, S. Oussalah and N. Sengouga, Z. Naturforsch., 2023, 78, 1143–1151,  DOI:10.1515/zna-2023-0200.
  244. T. Miyamoto, S. Numao, J. Sameshima and M. Yoshikawa, Surf. Interface Anal., 2023, 55(12), 880–885 CrossRef CAS.
  245. A. H. Alsaedi, A. S. Walton and N. P. Lockyer, J. Vac. Sci. Technol. B, 2023, 41(4), 044007,  DOI:10.1116/6.0002591.
  246. S. M. Liu, C. Li, H. C. Wu, L. F. Li, J. M. Liu, D. Wu, R. Hai and H. B. Ding, Spectrochim. Acta, Part B, 2023, 209, 106783,  DOI:10.1016/j.sab.2023.106783.
  247. X. Y. Xia, J. S. Wu, C. Q. Wu, Z. Gao, Z. Li, J. S. Xiu and H. Q. Liu, Thin Solid Films, 2023, 784, 140088,  DOI:10.1016/j.tsf.2023.140088.
  248. N. Giannakaris, P. Siozos, V. Pinon, S. P. Banerjee, M. Sentis, J. D. Pedarnig and D. Anglos, Appl. Surf. Sci., 2023, 640, 158354,  DOI:10.1016/j.apsusc.2023.158354.
  249. Y. Geng, S. Ma, H. Yang, J. Wang and C. Xu, Spectrochim. Acta, Part B, 2023, 206, 106730,  DOI:10.1016/j.sab.2023.106730.
  250. L. Huang, W. Zeng, R. Qian, S. J. Zhuo, Z. M. Gu, A. Q. Liu, J. Gao and Q. Chen, Microchem. J., 2023, 192, 108904,  DOI:10.1016/j.microc.2023.108904.
  251. S. Imashuku, Spectrochim. Acta, Part B, 2024, 212, 106842,  DOI:10.1016/j.sab.2023.106842.
  252. J. Lüchtefeld, H. Hemmelmann, S. Wachs, C. Behling, K. J. J. Mayrhofer, M. T. Elm and B. Berkes, J. Phys. Chem. C, 2023, 127(43), 21211–21219 CrossRef.
  253. K. V. Nikolaev, A. I. Safonov, O. A. Kondratev, G. V. Prutskov, I. A. Likhachev, I. A. Subbotin, M. M. Borisov, S. A. Tikhomirov, E. M. Pashaev and S. N. Yakunin, J. Appl. Crystallogr., 2023, 56, 1435–1445 CrossRef CAS.
  254. D. Bleiner, J. Anal. At. Spectrom., 2024, 39(4), 1057–1069,  10.1039/d3ja00399j.
  255. T. Yuan, L. Zhang, L. Hua, H. Li, Y. Zhong, B. Liao, C. Li, T. Sun and X. Sun, Nucl. Instrum. Methods Phys. Res., Sect. A, 2024, 1058, 168923,  DOI:10.1016/j.nima.2023.168923.
  256. J. Kang and D.-K. Ko, Appl. Phys. B: Lasers Opt., 2023, 129(4), 61,  DOI:10.1007/s00340-023-08011-0.
  257. J. Fujihara and N. Nishimoto, J. Iran. Chem. Soc., 2023, 20, 2555–2560,  DOI:10.1007/s13738-023-02853-3.
  258. T. Lombardo, F. Walther, C. Kern, Y. Moryson, T. Weintraut, A. Henss and M. Rohnke, J. Vac. Sci. Technol., A, 2023, 41(5), 053207,  DOI:10.1116/6.0002850.
  259. M. Vahnstiege, M. Winter, S. Nowak and S. Wiemers-Meyer, Phys. Chem. Chem. Phys., 2023, 25(36), 24278–24286 RSC.
  260. B. B. M. Sultan, O. Gharbi, K. Ogle and J. Han, Curr. Opin. Electrochem., 2023, 41, 101350,  DOI:10.1016/j.coelec.2023.101350.
  261. X. Wang, Y. Liu, H. Jia, S. Li, X. Zhang, X. Xing, Z. Wu, Z. Wu and W. Cheng, Russ. J. Electrochem., 2023, 59(12), 1206–1220 CrossRef.
  262. F. Langer, A. Utami, J. Kirres, E. Kramer and J. Bachmann, ACS Appl. Mater. Interfaces, 2023, 15(28), 33571–33580 CrossRef CAS PubMed.
  263. D. Kohmoto, K. Miyazaki, M. Morita, K. Fukuda and T. Abe, X-Ray Spectrom., 2024, 53(3), 166–180 CrossRef CAS.
  264. L. Cressa, Y. Y. Sun, D. Andersen, M. Gerard, O. De Castro, D. Kopljar, M. Nojabaee, K. A. Friedrich, G. Schmitz, T. Wirtz and S. Eswara, Anal. Chem., 2023, 95(26), 9932–9939 CrossRef CAS PubMed.
  265. D. S. Ferreira, F. M. V. Pereira, A. C. Olivieri and E. R. Pereira-Filho, Anal. Chim. Acta, 2024, 1303, 342522 CrossRef CAS PubMed.
  266. S. Ichikawa, Y. Hirokawa, T. Kurisaki and T. Nakamura, Spectrochim. Acta, Part B, 2023, 210, 106819,  DOI:10.1016/j.sab.2023.106819.
  267. H. Lee, S. Joo and D. Suh, J. Anal. At. Spectrom., 2023, 38, 1808–1815,  10.1039/d3ja00110e.
  268. Z. Khosravizadeh, P. Dziawa, S. Dad and R. Jakiela, Thin Solid Films, 2023, 781, 139974,  DOI:10.1016/j.tsf.2023.139974.
  269. J. Song, X. Qin, Q. Lyu, W. Fan, Q. Wang, X. Wang and Z. Jia, Electr. Eng., 2023, 105(3), 1775–1782 CrossRef.
  270. M. A. Alam, M. K. Tiwari, D. Devi, S. Tripathi, A. Trivedi, S. Ojha, R. Singh and M. Gupta, Spectrochim. Acta, Part B, 2023, 206, 106707,  DOI:10.1016/j.sab.2023.106707.
  271. D. Gallot-Duval, T. Meyer, C. Quere, T. Gutel, E. De Vito and J.-B. Sirven, Spectrochim. Acta, Part B, 2023, 206, 106731,  DOI:10.1016/j.sab.2023.106731.
  272. N. S. Medvedev, V. D. Kurbatova, T. Y. Guselnikova and A. I. Saprykin, At. Spectrosc., 2024, 45(1), 26–32,  DOI:10.46770/AS.2023.256.
  273. K. Cruz, G. L. Donati, F. R. P. Rocha and M. C. Hespanhol, Anal. Methods, 2023, 15(30), 3675–3682 RSC.
  274. D. Muss and R. Koch, Microelectron. Reliab., 2023, 148, 115177,  DOI:10.1016/j.microrel.2023.115177.
  275. M. Maisuradze, I. Carlomagno, A. Mullaliu, M. Li, G. Aquilanti and M. Giorgetti, J. Phys. Chem. C, 2023, 127(44), 21498–21503 CrossRef CAS.
  276. F. Zhang, H. Yuan, A. Yang, X. Wang, J. Chu, D. Liu and M. Rong, J. Anal. At. Spectrom., 2024, 39(1), 281–292 RSC.
  277. J. Labuda, J. Barek, Z. Gajdosechova, S. Jacob, L. Johnston, P. Krystek, Z. Mester, J. Moreira, V. Svitkova and K. J. Wilkinson, Pure Appl. Chem., 2023, 95(11), 1159–1196 CrossRef CAS.
  278. Z. Zhou and D. Beauchemin, Chem. Commun., 2024, 60(14), 1826–1839 RSC.
  279. C. Davison, D. Beste, M. Bailey and M. Felipe-Sotelo, Anal. Bioanal. Chem., 2023, 415(28), 6931–6950 CrossRef CAS PubMed.
  280. S. Fernandez-Trujillo, M. Jimenez-Moreno, N. Rodriguez-Farinas and R. C. Rodriguez Martin-Doimeadios, Anal. Bioanal. Chem., 2024, 416(11), 2657–2676 CrossRef CAS PubMed.
  281. P. Jagadeesh, S. M. Rangappa and S. Siengchin, Adv. Ind. Eng. Polym. Res., 2024, 7(1), 122–143 CAS.
  282. K. Loeschner, M. E. Johnson and A. R. Montoro Bustos, Nanomaterials, 2023, 13(18), 2547,  DOI:10.3390/nano13182547.
  283. M. Borowska and K. Jankowski, Anal. Bioanal. Chem., 2023, 415(18), 4023–4038 CrossRef CAS PubMed.
  284. Z. W. Meng, L. N. Zheng, H. Fang, P. Yang, B. Wang, L. Li, M. Wang and W. Y. Feng, Processes, 2023, 11(4), 1237,  DOI:10.3390/pr11041237.
  285. S. Yamashita, S. I. Miyashita and T. Hirata, Nanomaterials, 2023, 13(13), 1958,  DOI:10.3390/nano13131958.
  286. O. Geiss, I. Bianchi, G. Bucher, E. Verleysen, F. Brassinne, J. Mast, K. Loeschner, L. Givelet, F. Cubadda, F. Ferraris, A. Raggi, F. Iacoponi, R. Peters, A. Undas, A. Mueller, A.-K. Meinhardt, B. Hetzer, V. Graef, A. R. Montoro Bustos and J. Barrero-Moreno, Nanomaterials, 2022, 12(4), 725,  DOI:10.3390/nano12040725.
  287. L. Hendriks, R. Brunjes, S. Taskula, J. Kocic, B. Hattendorf, G. Bland, G. Lowry, E. Bolea-Fernandez, F. Vanhaecke, J. J. Wang, M. Baalousha, M. von der Au, B. Meermann, T. R. Holbrook, S. Wagner, S. Harycki, A. Gundlach-Graham and F. von der Kammer, Nanoscale, 2023, 15(26), 11268–11279 RSC.
  288. A. H. Henke, K. Flores, A. J. Goodman, K. Magurany, K. LeVanseler, J. Ranville, J. L. Gardea-Torresdey and P. K. Westerhoff, Sci. Total Environ., 2024, 912, 168686 CrossRef CAS PubMed.
  289. F. Bennet, R. Opitz, N. Ghoreishi, K. Plate, J.-P. Barnes, A. Bellew, A. Belu, G. Ceccone, E. de Vito, A. Delcorte, A. Franquet, F. Fumagalli, D. Gilliland, H. Jungnickel, T. G. Lee, C. Poleunis, D. Rading, H. K. Shon, V. Spampinato, J. G. Son, F. Wang, Y.-C. A. Wang, Y. Zhao, A. Roloff, J. Tentschert and J. Radnik, J. Vac. Sci. Technol., A, 2023, 41(5), 053210,  DOI:10.1116/6.0002814.
  290. B. Moreira-Alvarez, L. Cid-Barrio, F. Calderon-Celis, J. M. Costa-Fernandez and J. R. Encinar, Anal. Chem., 2023, 95(27), 10430–10437 CrossRef CAS PubMed.
  291. K. E. Murphy, A. R. Montoro Bustos, L. L. Yu, M. E. Johnson and M. R. Winchester, Spectrochim. Acta, Part B, 2024, 212, 106841 CrossRef CAS.
  292. H. G. Infante, Spectrochim. Acta, Part B, 2024, 216, 106876 CrossRef CAS.
  293. S. Cuello-Nuñez, I. Abad-Álvaro, D. Bartczak, M. E. del Castillo Busto, D. A. Ramsay, F. Pellegrino and H. Goenaga-Infante, J. Anal. At. Spectrom., 2020, 35(9), 1832–1839 RSC.
  294. K. E. Murphy, A. R. Montoro Bustos, L. L. Yu, M. E. Johnson and M. R. Winchester, Spectrochim. Acta, Part B, 2024, 216, 106902 CrossRef CAS.
  295. D. Torregrosa, G. Grindlay, L. Gras and J. Mora, J. Anal. At. Spectrom., 2023, 38, 1874–1884,  10.1039/d3ja00134b.
  296. M. Elinkmann, S. Reuter, M. Holtkamp, S. Heuckeroth, A. Koehrer, K. Kronenberg, M. Sperling, O. Rubner, C. D. Quarles, M. Hippler and U. Karst, J. Anal. At. Spectrom., 2023, 38(12), 2607–2618 RSC.
  297. Y. Suzuki, M. Kondo, M. Harimoto, Y. Okamoto, Y. K. Tanaka, Y. Ogra and H. Akiyama, J. Anal. At. Spectrom., 2024, 39(1), 190–203 RSC.
  298. S. G. Bevers, C. Smith, S. Brown, N. Malone, D. H. Fairbrother, A. J. Goodman and J. F. Ranville, Environ. Sci.: Nano, 2023, 10, 3136–3148,  10.1039/d3en00425b.
  299. A. Schardt, J. Schmitt and C. Engelhard, J. Anal. At. Spectrom., 2024, 39(2), 389–400,  10.1039/d3ja00373f.
  300. A. Bazo, M. Aramendia, F. V. Nakadi and M. Resano, Nanomaterials, 2023, 13(12), 1838,  DOI:10.3390/nano13121838.
  301. G. Kajner, A. Belteki, M. Cseh, Z. Geretovszky, T. Ajtai, L. Barna, M. A. Deli, B. Pap, G. Maroti and G. Galbacs, Nanomaterials, 2023, 13(23), 3018,  DOI:10.3390/nano13233018.
  302. M. Aramendia, D. Leite, J. Resano, M. Resano, K. Billimoria and H. Goenaga-Infante, Nanomaterials, 2023, 13, 2392,  DOI:10.3390/nano13172392.
  303. L. Hendriks and D. M. Mitrano, Environ. Sci. Technol., 2023, 57(18), 7263–7272 CrossRef CAS PubMed.
  304. L. Hendriks, V. M. Kissling, T. Buerki-Thurnherr and D. M. Mitrano, Environ. Sci.: Nano, 2023, 10(12), 3439–3449 RSC.
  305. T. E. Lockwood, R. G. de Vega, Z. Du, L. Schlatt, X. Xu and D. Clases, J. Anal. At. Spectrom., 2024, 39(1), 227–234 RSC.
  306. B. T. Manard, V. C. Bradley, C. D. Quarles Jr, L. Hendriks, D. R. Dunlap, C. R. Hexel, P. Sullivan and H. B. Andrews, Nanomaterials, 2023, 13(8), 1322,  DOI:10.3390/nano13081322.
  307. R. L. Buckman and A. Gundlach-Graham, J. Anal. At. Spectrom., 2023, 38(6), 1244–1252 RSC.
  308. A. Gundlach-Graham, S. Harycki, S. E. Szakas, T. L. Taylor, H. Karkee, R. L. Buckman, S. Mukta, R. Hu and W. Lee, J. Anal. At. Spectrom., 2024, 39, 704–711,  10.1039/d3ja00421j.
  309. M. Erfani, M. Baalousha and E. Goharian, Sci. Total Environ., 2023, 905, 167176 CrossRef CAS PubMed.
  310. A. Sajnóg, K. Bierla, J. Szpunar and J. Jiménez-Lamana, J. Anal. At. Spectrom., 2023, 38(11), 2448–2457 RSC.
  311. M. S. Jimenez, M. Bakir, K. Ben-Jeddou, E. Bolea, J. Perez-Arantegui and F. Laborda, Mikrochim. Acta, 2023, 190(6), 204,  DOI:10.1007/s00604-023-05777-0.
  312. A. Justo-Vega, R. Dominguez-Gonzalez, P. Bermejo-Barrera and A. Moreda-Pineiro, J. Anal. At. Spectrom., 2024, 39, 573–581,  10.1039/d3ja00227f.
  313. V. Kinnunen, J. Frimodig, S. Peramaki and R. Matilainen, Spectrochim. Acta, Part B, 2023, 203, 106662,  DOI:10.1016/j.sab.2023.106662.
  314. T. Cen, Y.-B. Zhao, A. Testino, J. Wang, L. Torrent and C. Ludwig, J. Aerosol Sci., 2024, 177, 106317,  DOI:10.1016/j.jaerosci.2023.106317.
  315. L. Paton, T. T. Moro, T. Lockwood, T. de Andrade Maranhão, G. Gössler, D. Clases and J. Feldmann, Environ. Sci.: Nano, 2024, 11(5), 1883–1890 RSC.
  316. D. Baron, T. Pluhacek and J. Petr, Anal. Chem., 2024, 96(14), 5658–5663 CrossRef CAS PubMed.
  317. S. Yamashita and T. Hirata, Mass Spectrom., 2023, 12(1), A0116 CrossRef CAS PubMed.
  318. S. B. Seiffert, M. Elinkmann, E. Niehaves, A. Vennemann, D. Mozhayeva, S. Kroger, M. Wiemann and U. Karst, Anal. Chem., 2023, 95(15), 6383–6390 CrossRef CAS PubMed.
  319. W. Luo, F. Dong, M. Wang, T. Li, Y. Wang, W. Dai, J. Zhang, C. Jiao, Z. Song, J. Shen, Y. Ma, Y. Ding, F. Yang, Z. Zhang and X. He, Anal. Chem., 2023, 95(15), 6391–6398 CrossRef CAS PubMed.
  320. J. Cárdenas-Escudero, V. Gardette, A. Villalonga, A. Sanchez, R. Villalonga, V. Motto-Ros, D. Galan-Madruga and J. O. Caceres, J. Anal. At. Spectrom., 2024, 39(3), 962–973 RSC.
  321. C. Burgos-Palop, P. Purohit, F. J. Fortes and J. Laserna, Anal. Chem., 2023, 95(39), 14541–14550 CrossRef CAS PubMed.
  322. T. Liu, X. Cheng, Q. Zhang, T. Zhang and J. Bai, Talanta, 2024, 268, 125326,  DOI:10.1016/j.talanta.2023.125326.
  323. G. Abbondanza, A. Grespi, A. Larsson, D. Dzhigaev, L. Glatthaar, T. Weber, M. Blankenburg, Z. Hegedus, U. Lienert, H. Over, G. S. Harlow and E. Lundgren, Nanotechnology, 2023, 34(50), 505605,  DOI:10.1088/1361-6528/acf66e.
  324. E. Abdel-Fattah, D. Ogawa and K. Nakamura, Vacuum, 2024, 222, 113002,  DOI:10.1016/j.vacuum.2024.113002.
  325. Q. Bai, Q. Li, Z. Tan and J. Liu, Talanta, 2024, 272, 125738 CrossRef CAS PubMed.
  326. M. Ding, B. Moreira-Alvarez, F. C. Celis, J. M. Costa-Fernandez, J. R. Encinar and R. Gref, RSC Adv., 2024, 14(3), 1676–1685 RSC.
  327. S. Fernandez-Castillo Suarez, C. Courreges, J. Jimenez-Lamana, S. Godin, S. Nolivos, R. Grimaud, L. Ronga, J. Szpunar and J. Allouche, Environ. Sci.: Nano, 2024, 11(4), 1429–1441,  10.1039/d3en00523b.
  328. A. J. Goodman, S. Kanapilly, A. J. Bednar and J. F. Ranville, J. Anal. At. Spectrom., 2024, 39(2), 364–369 RSC.
  329. J. Han, X. Wu, J. X. Zhao and D. T. Pierce, Nanomaterials, 2024, 14(7), 637,  DOI:10.3390/nano14070637.
  330. D. Itabashi and K. Mizukami, Tetsu to Hagane, 2023, 109(7), 573–580 CrossRef.
  331. A. Krot, S. Savilov, E. Arkhipova and S. Kalmykov, J. Anal. At. Spectrom., 2024, 39(1), 71–75 RSC.
  332. O. Martinez-Mora, K. Tirez, F. Beutels, W. Brusten, L. F. Leon-Fernandez, J. Fransaer, X. Dominguez-Benetton and M. Velimirovic, Nanomaterials, 2023, 13(18), 2610,  DOI:10.3390/nano13182610.
  333. P. P. Michalowski, J. Muller, C. Rossi, A. Burenkov, E. Bar, G. Larrieu and P. Pichler, Measurement, 2023, 211, 112630,  DOI:10.1016/j.measurement.2023.112630.
  334. D. J. Palasti, L. Villy, B. Leits, A. Keri, A. Kohut, A. Belteki, G. Kajner, F. A. C. Plaza, E. Kovacs-Szeles, T. Ajtai, M. Veres, Z. Geretovszky and G. Galbacs, Spectrochim. Acta, Part B, 2023, 209, 106804,  DOI:10.1016/j.sab.2023.106804.
  335. L. Phuc, V. H. Thien and N. Van Dong, Vietnam J. Chem., 2023, 61, 109–117,  DOI:10.1002/vjch.202300086.
  336. P. Purohit, F. J. Fortes and J. Lasema, Nano Res., 2023, 16(5), 7470–7480 CrossRef CAS.
  337. R. Taitt, M. Urbain, K. Bredillet, Z. Behel, G. Ceccone, J. Bañuls-Ciscar, S. Beauquis, Y. Mugnier, P. F. Brevet, R. Le Dantec, Y. Chevolot and V. Monnier, Part. Part. Syst. Charact., 2022, 39(11), 220093,  DOI:10.1002/ppsc.202200093.
  338. T. Techarang and A. Siripinyanond, Microchem. J., 2023, 193, 109089,  DOI:10.1016/j.microc.2023.109089.
  339. A. C. Zhang, S. M. Maguire, J. T. Ford and R. J. Composto, Microsc. Microanal., 2023, 29(5), 1557–1565 CrossRef PubMed.
  340. M. Zhang, H. Wang, Y. Wu and X. Yu, Anal. Chim. Acta, 2023, 1279, 341846 CrossRef CAS PubMed.
  341. D. Kobylarz, A. Michalska and K. Jurowski, TrAC, Trends Anal. Chem., 2023, 169, 117355,  DOI:10.1016/j.trac.2023.117355.
  342. G. P. Jackson and M. A. Barkett, J. Am. Soc. Mass Spectrom., 2023, 34(7), 1210–1224 CrossRef CAS PubMed.
  343. X. Wang, Y. Li, C. Wang, S. Huang, F. Zhou and X. Lin, X-Ray Spectrom., 2024 DOI:10.1002/xrs.3423.
  344. S. Tambuzzi, L. Bonizzoni, D. P. Francesco, M. Debora, C. Giulia and C. Cristina, X-Ray Spectrom., 2024, 53(2), 112–120 CrossRef CAS.
  345. D. Charlton, C. Costa, S. J. Hinder, J. F. Watts and M. J. Bailey, Molecules, 2023, 28(15), 5687,  DOI:10.3390/molecules28155687.
  346. Z. L. Liu, W. J. Li, Y. Wu, H. Z. Man, Y. B. Zhao and Z. P. Li, Anal. Methods, 2023, 15(39), 5225–5232 RSC.
  347. M. Furukawa, Y. Niida, K. Kobayashi, M. Furuishi, R. Umezawa, O. Shikino and T. Suzuki, Anal. Sci., 2023, 39, 1957–1966,  DOI:10.1007/s44211-023-00403-8.
  348. S. Krishna and P. Ahuja, Egypt. J. Forensic Sci., 2023, 13(1), 51,  DOI:10.1186/s41935-023-00369-8.
  349. S. Szakas and A. Gundlach-Graham, Talanta, 2024, 268, 125368,  DOI:10.1016/j.talanta.2023.125368.
  350. S. E. Szakas, K. Menking-Hoggatt, T. Trejos and A. Gundlach-Graham, Appl. Spectrosc., 2023, 77(8), 873–884 CrossRef CAS PubMed.
  351. S. R. Khandasammy, L. Halamkova, M. Baudelet and I. K. Lednev, Spectrochim. Acta Mol. Biomol. Spectrosc., 2023, 291, 122316,  DOI:10.1016/j.saa.2023.122316.
  352. H. Guo, C. Hu, P. Wang, H. Mei, Y. Li and J. Zhu, Forensic Sci. Int., 2023, 351, 111812,  DOI:10.1016/j.forsciint.2023.111812.
  353. A. Doña-Fernández, J. A. Rodriguez-Pascual, I. de Andres-Gimeno, E. Gutiérrez-Redomero, E. Valtuille-Fernández and F. J. Gomez-Laina, Forensic Sci. Res., 2023, 8, 256–264,  DOI:10.1093/fsr/owad022.
  354. J. A. Rodriguez-Pascual, A. Doña-Fernández, Y. Loarce-Tejada, I. de Andres-Gimeno, E. Valtuille-Fernandez, E. Gutierrez-Redomero and F. Javier Gomez-Laina, Forensic Sci. Int., 2023, 353, 111886,  DOI:10.1016/j.forsciint.2023.111886.
  355. C. Vander Pyl, K. Menking-Hoggatt, L. Arroyo, J. Gonzalez, C. Liu, J. Yoo, R. E. Russo and T. Trejos, Spectrochim. Acta, Part B, 2023, 207, 106741,  DOI:10.1016/j.sab.2023.106741.
  356. M. Donghi, S. Orsenigo, L. Manna, A. Profumo, A. Mattino and D. Merli, J. Forensic Sci., 2024, 69, 1035–1044,  DOI:10.1111/1556-4029.15484.
  357. J. Ding, T. Zhang and H. Li, TrAC, Trends Anal. Chem., 2023, 166, 117197,  DOI:10.1016/j.trac.2023.117197.
  358. M. Manhas, A. Tomar, M. Tiwari and S. Sharma, X-Ray Spectrom., 2024 DOI:10.1002/xrs.3421.
  359. K. Heimler, C. Gottschalk and C. Vogt, Anal. Bioanal. Chem., 2023, 415(21), 5083–5100 CrossRef CAS PubMed.
  360. K. Johnson, C. P. Quinn, N. Goodale and R. Conrey, Adv. Archaeol. Pract., 2024, 12, 156–162,  DOI:10.1017/aap.2024.6.
  361. M. Chiti, D. Chiti, F. Chiarelli, R. Donghia, A. Esposito, M. Ferretti and A. Gorghinian, Condens. Matter, 2024, 9(1), 1,  DOI:10.3390/condmat9010001.
  362. C. Moreau, J.-P. Dumoulin, M. Jaber, I. Caffy, E. Delque-Kolic, C. Goulas, S. Hain, M. Perron, V. Setti, M. Sieudat, B. Thellier and L. Beck, Radiocarbon, 2024, 1–12,  DOI:10.1017/RDC.2023.118.
  363. S. Porcinai, A. Cagnini, M. Galeotti and M. Ferretti, Spectrochim. Acta, Part B, 2023, 210, 106808,  DOI:10.1016/j.sab.2023.106808.
  364. A. Bliujiene, S. Jatautis, S. Suzdalev and G. Petrauskas, Archaeol. Balt., 2023, 30, 6–40 CrossRef.
  365. E. Konstantakopoulou, A. C. Municchia, L. Luvidi and M. Ferretti, Condens. Matter, 2024, 9(1), 5,  DOI:10.3390/condmat9010005.
  366. J. Brocchieri, R. Vitale and C. Sabbarese, X-Ray Spectrom., 2023, 53(6), 452,  DOI:10.1002/xrs.3395.
  367. J. Brocchieri, R. Vitale and C. Sabbarese, Quantum Beam Sci., 2024, 8(1), 2,  DOI:10.3390/qubs8010002.
  368. E. Berard, F. Caro and P. Dillmann, J. Archaeol. Sci., 2024, 54, 104465,  DOI:10.1016/j.jasrep.2024.104465.
  369. P. Penkova, G. Malcheva, M. Grozeva, T. Hristova, G. Ivanov, S. Alexandrov, K. Blagoev, V. Tankova and V. Mihailov, J. Archaeol. Sci., 2023, 7(3), 22,  DOI:10.3390/qubs7030022.
  370. M. Porcaro, R. Cesareo, A. Bustamante and A. Brunetti, Materials, 2023, 16(22), 7211,  DOI:10.3390/ma16227211.
  371. N. K. Kladouri, S. Skaltsa, T. Gerodimos, K. Pezouvani and A. G. Karydas, Archaeol. Anthropol. Sci., 2023, 15(9), 141,  DOI:10.1007/s12520-023-01834-0.
  372. J. M. del Hoyo-Melendez, M. Matosz, A. Walanus and P. Krupska-Wolas, J. Archaeol. Sci., 2023, 51, 104156,  DOI:10.1016/j.jasrep.2023.104156.
  373. V. Gupta, A. K. Rai, T. Kumar, R. Junjuri, G. M. Kumar and A. K. Rai, Plasma Sci. Technol., 2023, 25(9), 095501,  DOI:10.1088/2058-6272/acccd9.
  374. F. Abate, M. De Bernardin, M. Stratigaki, G. Franceschin, F. Albertin, M. Bettuzzi, R. Brancaccio, A. Bressan, M. P. Morigi, S. Daniele and A. Traviglia, J. Cult. Herit., 2024, 66, 436–443 CrossRef.
  375. I. Carlomagno, P. Zeller, M. Amati, G. Aquilanti, E. Prenesti, G. Marussi, M. Crosera and G. Adami, Sci. Rep., 2022, 12(1), 15919,  DOI:10.1038/s41598-022-19682-8.
  376. F. Albarède, C. Malod-Dognin and P. Télouk, J. Anal. At. Spectrom., 2023, 38(10), 2159–2166 RSC.
  377. L. de Palaminy, F. Poitrasson, S. Baron, M. Blet-Lemarquand and L. Perriere, J. Anal. At. Spectrom., 2024, 39, 1302–1321,  10.1039/d3ja00312d.
  378. S. S. Gomes, P. Valério, V. T. de Freitas, C. Fabiäo, A. M. Soares and M. F. Araújo, Archaeol. Anthropol. Sci., 2023, 15(11), 171,  DOI:10.1007/s12520-023-01878-2.
  379. M. Perino, L. Pronti, C. Moffa, M. Rosellini and A. C. Felici, Heritage, 2024, 7(2), 683–696 CrossRef.
  380. I. M. Cortea, A. Chirosca, L. M. Angheluta and G. Seritan, ACM J. Comput. Cult. Herit., 2023, 16(2), 40,  DOI:10.1145/3593427.
  381. A. Bombini, F. G.-A. Bofias, C. Ruberto and F. Taccetti, Rendiconti Lincei. Sci. Fis. Nat., 2023, 34(3), 867–887 CrossRef.
  382. M. Occhipinti, R. Alberti, T. Parsani, C. Dicorato, P. Tirelli, M. Gironda, A. Tocchio and T. Frizzi, X-Ray Spectrom., 2024, 53(6), 520,  DOI:10.1002/xrs.3405.
  383. E. Kechaoglou, K. A. Agrafioti, G. P. Mastrotheodoros, D. F. Anagnostopoulos and C. Kosmidis, J. Anal. At. Spectrom., 2024, 39, 854–867,  10.1039/d3ja00199g.
  384. A. Swiecicka, A. Towarek, M. Strawski and B. Wagner, Spectrochim. Acta, Part B, 2023, 206, 106717,  DOI:10.1016/j.sab.2023.106717.
  385. V. Andric, G. Kvascev, M. Cvetanovic, S. Stojanovic, N. Bacanin and M. Gajic-Kvascev, Sci. Rep., 2024, 14(1), 3666,  DOI:10.1038/s41598-024-53988-z.
  386. X. Bai, T. Calligaro, B. Moignard, S. Richiero and V. Detalle, Appl. Phys. Lett., 2024, 124(6), 061107,  DOI:10.1063/5.0179543.
  387. F. T. H. Broers, K. Janssens, J. N. Weker, S. M. Webb, A. Mehta, F. Meirer and K. Keune, J. Am. Chem. Soc., 2023, 145(16), 8847–8859 CrossRef CAS PubMed.
  388. G. Pastorelli, A. S. O. Miranda, E. A. Clerici, P. d'Imporzano, B. V. Hansen, K. Janssens, G. R. Davies and N. Borring, Microchem. J., 2024, 199, 109912,  DOI:10.1016/j.microc.2024.109912.
  389. N. Thaumaturgo, C. R. F. Souza, T. J. N. A. Fialho, R. S. Liarth, A. P. Oliveira, D. Guimaraes, V. S. Felix, A. R. Pimenta, A. L. C. Oliveira, M. B. Oliveira, M. A. Oliveira, D. S. R. Ferreira and R. P. Freitas, Spectrochim. Acta Mol. Biomol. Spectrosc., 2024, 304, 123431,  DOI:10.1016/j.saa.2023.123431.
  390. T. Ghigo and A. N. Dominguez, Archaeol. Anthropol. Sci., 2023, 15(9), 132,  DOI:10.1007/s12520-023-01839-9.
  391. V. Biolcati, J. Woolley, E. Leveque, A. Rossi, A. G. Hoffmann, A. Visentin, P. O. Machain and D. Iacopino, Heritage Sci., 2023, 11(1), 265,  DOI:10.1186/s40494-023-01107-y.
  392. M. Albrecht, S. Meloni, C. Pottasch, A. Vandivere and A. van Loon, X-Ray Spectrom., 2024, 53, 249–262,  DOI:10.1002/xrs.3348.
  393. L. Zhang, Z. Y. Song, S. D. Zuo, F. Hou and S. Q. Chen, Heritage Sci., 2023, 11(1), 230,  DOI:10.1186/s40494-023-01074-4.
  394. F. T. H. Broers, I. Verslype, K. W. Bossers, F. Vanmeert, V. Gonzalez, J. Garrevoet, A. van Loon, E. van Duijn, A. Krekeler, N. De Keyser, I. Steeman, P. Noble, K. Janssens, F. Meirer and K. Keune, Sci. Adv., 2023, 9(50), eadj9394,  DOI:10.1126/sciadv.adj9394.
  395. V. M. Chubarov, G. V. Pashkova, A. S. Maltsev, M. M. Mukhamedova, M. A. Statkus and A. G. Revenko, J. Anal. Chem., 2024, 79(3), 262–272 CrossRef CAS.
  396. C. Sirisathitkul, Makara J. Sci., 2024, 28(1), 5,  DOI:10.7454/mss.v28i1.2127.
  397. A. Horvath, B. Bajnoczi, D. Gyorkos and L. Palcsu, Archaeometry, 2023, 65, 1004–1017,  DOI:10.1111/arcm.12872.
  398. D. Paghi, R. Manca, M. Casalini, L. Chiarantini, A. Bragagni, S. Tommasini and M. Benvenuti, Eur. Phys. J. Plus, 2024, 139(3), 260,  DOI:10.1140/epjp/s13360-024-05024-2.
  399. J. Li, Y. Yang, D. Zou, L. Wu, Q. Wu, Z. Lin and Q. Li, Heritage Sci., 2024, 12(1), 79,  DOI:10.1186/s40494-024-01193-6.
  400. L. M. He, S. Yao, Z. Y. Sun, J. Shao, N. Di and T. Li, J. Archaeol. Sci., 2023, 48, 103844,  DOI:10.1016/j.jasrep.2023.103844.
  401. L. Zhang, Y. Sun, H. Xu and F. Hou, Int. J. Architect. Herit., 2024, 18(1), 150–163 CrossRef.
  402. V. Atanassova, V. Tankova, V. Mihailov and A. Pirovska, Minerals, 2024, 14(2), 152,  DOI:10.3390/min14020152.
  403. B. Massimo, R. Simona, Q. Costanza, M. Jose and C. Ginevra, Open Archaeol., 2023, 9(1), 20220322,  DOI:10.1515/opar-2022-0322.
  404. D. Zampierin, P. Moita, S. Lischi, M. van Aerde, P. Barrulas and J. Mirao, Archaeometry, 2024, 66, 967–1015,  DOI:10.1111/arcm.12949.
  405. I. Silachyov and Y. Akymbek, J. Radioanal. Nucl. Chem., 2023, 332(9), 3799–3811 CrossRef CAS.
  406. M. Bayazit, A. A. Akyol and M. Ariyurek, X-Ray Spectrom., 2024, 53, 38–59,  DOI:10.1002/xrs.3355.
  407. P. Li, Y. Hu, X. Zhang, C. Zhang, X. Chu and T. Li, Archaeol. Res. Asia, 2024, 38, 100504,  DOI:10.1016/j.ara.2024.100504.
  408. F. Gherardi, C. Hole, E. Campbell, M. Cotte, R. Tyson and S. Paynter, J. Phys., 2024, 6(2), 025001,  DOI:10.1088/2515-7647/ad2259.
  409. J. Then-Obluska and L. Dussubieux, Archaeometry, 2023, 65(5), 1018–1031 CrossRef CAS.
  410. A. K. Hodgkinson, Q. Lemasson, M. Maeder, F. Munnik, L. Pichon, S. Roehrs and I. Reiche, J. Archaeol. Sci., 2024, 54, 104412,  DOI:10.1016/j.jasrep.2024.104412.
  411. K.-W. Wang, L. Dussubieux, Y. Iizuka, K.-t. Li and C.-h. Tsang, Heritage Sci., 2023, 11(1), 255,  DOI:10.1186/s40494-023-01093-1.
  412. E. Gliozzo, M. Ferri, E. Braschi, S. Cadamuro and A. Cianciosi, Microchem. J., 2023, 189, 108511,  DOI:10.1016/j.microc.2023.108511.

This journal is © The Royal Society of Chemistry 2024
Click here to see how this site uses Cookies. View our privacy policy here.