J. M.
Mermet
Laboratoire des Sciences Analytiques (UMR CNRS 5180), Université Claude Bernard-Lyon 1, F-69622, Villeurbanne Cedex, France
First published on 3rd December 2004
J. M. Mermet | Jean-Michel Mermet is Research Director at the National Center for Scientific Research (CNRS). He was born in Lyon, France, conducted his graduate degree work at the University of Strasbourg in 1964, received his doctor’s degree in chemistry from the University of Lyon in 1974, and worked as post-doctoral fellow at the University of Florida in 1977 with Prof. J. D. Winefordner. In 1985, he was appointed Director of the Laboratory of Analytical Sciences at the University of Lyon. His research interests have been focused on the spectrochemistry of plasmas and lasers from fundamental, analytical and instrumental aspects. He has published more than 200 papers, reviews and book chapters. |
Sir Humphry Davy
It is often said that there is no more need for research in atomic emission spectrometry (AES), and that it is a dead-end street. Therefore, a question could be raised: Is it still possible, necessary and beneficial to perform research in atomic emission spectrometry? Actually this question contains some keywords. “Possible” implies that financial support may be obtained to conduct research regardless of the origin of the support, i.e., national or international agencies, instrument companies, or industry. “Necessary” means that strong needs can be defined by users or instrument companies, or are required because of new regulations. “Beneficial” means that this type of research is recognized through conferences, scientific journals and awards, and that students preparing a PhD thesis on this subject would easily find a job corresponding to their expertise. More complex is what should be “research” in analytical chemistry, and in particular in AES. In contrast to what non-analysts think, research cannot be confined to the use of an analytical method, as sophisticated as it might be. It is similar with computers: using complex softwares does not mean that we are computerists.
The advantages of AES are well known: photons can easily travel and be collected with simple optics, they do not exhibit memory effects or degrade detectors, and various detectors are available, including array or 2D detectors. Besides, depending on the radiation source, the spectrum of each element present in the sample can be obtained, which results in an inherent multi-element technique. AES is widely used for elemental analysis in the routine laboratory. The need for elemental analysis will remain for ever as the determination of elements in various matrices will always be a request. The only change will be a trend to determine more and more elements with lower and lower concentrations, in more and more complex matrices, with a demand for more information such as speciation (for bioavailability and toxicity reasons) and structure. Because there are not so many multi-element techniques, there is still room for AES, provided that it will be the subject of further improvements.
However, ICP-AES suffers from some limitations: the LOD are excellent for light elements, but are not satisfactory enough for elements such as As, Se, Pb… Spectral interferences due to spectrum richness, line broadening and insufficient resolution can hamper the analysis with matrices such as Co, Fe, Mo, Nb, Ta, U, V, W and some rare earth elements. Non-spectral matrix effects are usually low, but present, which makes them dangerous as the effects are not always obvious. This last limitation is probably the most crucial one.
On the whole, the use of ICP-AES in routine and research laboratories is perfectly justified, which is illustrated by the worldwide market that is about 1400 units a year, corresponding to a turnover of about 120M$.
In terms of publications, it could be thought that the introduction of ICP-MS led to a decrease in the number of ICP-AES papers. If we select a journal such as JAAS, where a large number of ICP-MS papers are published, it is interesting to see that, although the number of ICP-MS publications is still growing, the number of ICP-AES papers has not varied drastically over the past 20 years (Fig. 1).
Basically, an ICP-AES system has remained the same since its introduction: a hf generator, a pneumatic nebulizer associated with a spray chamber, a torch, a dispersive system, a detector and a computer for data acquisition and processing. The system can be partially or fully computer-driven. Major changes have centred on the detection level, with some consequences on the optical mounting of the dispersive system. Before the 1990s, dispersive systems were either monochromators (so-called sequential systems) or polychromators (so-called simultaneous systems). Both types were equipped with PMTs. Most polychromators were based on the Paschen–Runge optical mounting, with the PMTs set up on the Rowland circle. The commercial introduction of solid state detectors, first of the photodiode type, then of the CTD type, either CCD or CID, has revived echelle grating-based dispersive systems, with most of them making use of a post-cross-dispersion to obtain a 2D spectrum. Even the Rowland circle is now equipped with a linear assembly of CCD arrays. As a consequence, conventional polychromators, i.e., those equipped with PMTs, almost disappeared from the market, and sequential systems, which represented about two-third of the market at the end of the 1980s, were overpassed by the new polychromator generation in 1992–1993. Note that there is a dispersive system intermediate between a monochromator and polychromator: the system is making use of a multichannel CTD detector, and instead of a single narrow bandpass at a time, a more or less wide spectral window is selected and moves sequentially.
Search for new physical/chemical processes | |
---|---|
Understanding of the processes | Diagnostics |
Mechanisms | |
Modelling | |
Non-spectral matrix effects | |
Internal standardization | |
Development of the method | Instrumentation |
Data processing, information retrieval, chemometrics | |
Analytical results | |
Optimization | |
Applications to illustrate the potential and the current limits of the methods | |
Method validation | Analytical performance |
Uncertainty and traceability | |
Linearity of the calibration graph |
It is obvious that the vast majority of publications is related to applications. To illustrate the various fields of research, Fig. 2 gives the number of publications for some fields such as fundamentals, i.e., non-spectral matrix effects, excitation mechanisms and internal standardization, and instrumentation, i.e. axial viewing, micronebulization, laser ablation, hydride and vapour generation, and electrothermal vaporization. Also, as an example of data processing, the number of papers related to qualitative analysis is given. These figures were extracted from over 2300 ICP-AES publications stored in my own database.
Fig. 2 Number of ICP-AES publications related to matrix effects (MX), mechanisms (ME), internal standardization (IS), acid effects (AC), qualitative analysis (QA), axial viewing (AX), electrothermal atomization (ET), hydride and vapor generation (HG), laser ablation (LA), micronebulization (MN) and speciation (SP). |
Over this number of publications, a similar number of roughly 500 ICP-AES publications have been published in the Journal of Analytical and Atomic Spectrometry and in Spectrochimica Acta, Part B (Fig. 3).
Fig. 3 Number of ICP-AES publications against the scientific journal. Note that the figure for Anal. Bioanal. Chem. contains also publication from the former Fresenius’ Z. Anal. Chem. and Fresenius’ J. Anal. Chem. |
Note that the classification was not based on keywords, but on my own perception of the contents of the papers. However arbitrary it might be, it is nevertheless an indicator of the past research in ICP-AES. A hundred papers on non-spectral matrix effects means that less than 5% of those papers have been related to this field. It can be easily understood that only very few laboratories have the expertise to run diagnostics based on Thomson scattering because of the complex set-up, along with the need to adapt the instrument for this purpose. In contrast, most of the studies on matrix effects, at least in terms of results, could be performed on commercially available instruments by any analyst willing to do it. Interesting to note is that axial viewing is the default viewing mode for most systems that are currently commercialized. However, less than 60 papers dealt with the influence or the role of such a viewing mode.
A large number of papers are concerned with alternative sample introduction systems such as laser ablation, micronebulization, volatile species generation, electrothermal vaporization. First, a need exists for the replacement of the conventional pneumatic nebulizer. Then, it is probably the easiest part to modify on a commercially available system. Early research in the 1960s was based on the availability of a hf generator either designed for inductive heating (usually with a free-running technology) of for broadcasting (crystal-controlled technology), a dispersive system equipped with photomultiplier tube(s), and a simple acquisition system, for instance a picoammeter associated to a chart recorder. Because systems were not computer driven with the absence of firmwares, and consisted of separated parts, it was relatively simple to modify the instrument configuration. Currently, apart the sample introduction system, it is far more complex to modify a commercially available system.
Evidently, applications are numerous and this field is currently the most popular and justifies the success and the high citation index of scientific journals such as JAAS. Surprisingly enough, speciation has been the subject of almost 60 papers, although it could be thought that ICP-MS is the most appropriate instrument for this purpose. In contrast, the number of papers related to method validation, and in particular to the estimation of uncertainty budget, is almost below the limit of detection.
To extrapolate current research, it is convenient to define an ideal system for elemental analysis and to verify where the missing features are located. A system may be characterized by the quality of the analytical results, the analytical quality of the system, the instrument operation characteristics, and the economical aspects. Details are given in Table 2.
Quality of the analytical results | Accuracy and precision (including repetability, intermediate precision and the various reproducibilities) |
Uncertainty and traceability using the metrology approach | |
Analytical quality of the system | Number of elements which can be determined by the system, mostly related to the wavelength range |
Sensitivity and low limits of detection and quantitation | |
Long-term stability | |
Selectivity, i.e., the ability to separate the analyte line from concomitant lines, which is related to the practical resolution | |
Robustness, i.e., the absence of non-spectral matrix and interelement effects | |
High linearity and dynamic range | |
Instrument operation characteristics | Ease of operation |
Ease of maintenance | |
Intelligent and fully automated system | |
Use of any form of the sample, solid, liquid or gas | |
Low sample consumption if any | |
Low size and low floor space of the system | |
Economical aspects | Fast analysis and high sample throughput |
Reliability | |
Safety | |
Low capital investment | |
Low running cost |
Considering that the analyst is a problem solver and needs accurate and precise results, along with low limits of detection and quantitation, some challenges remain in ICP-AES and are summarised in Table 3.
Plasma generation | Mixed-gas plasma |
Sample introduction system | Total-consumption system |
Chemical vapor generation | |
Direct solid analysis | |
Photon collection and detection | Axial viewing |
Reduction in shot noise | |
Improved solid-state detection | |
Intelligent software | Automatic (multi)line selection |
Semi-quantitative analysis | |
True self-diagnostics | |
Analytical performance | Improvement in limits of detection |
Origin and minimization of matrix effects | |
Method validation | Guideline |
Uncertainty budget |
Direct injection is probably a way to pursue, if some current limitations can be overcome: too a large solvent loading, and large size and high velocity of the droplets. An alternative is to modify the role of the spray chamber, so that the chamber acts as an evaporation cavity. This is possible because of the availability of efficient micronebulizers.
Chemical vapor and volatile species generation is a growing field. There is no more need for evaporation, and it is possible to separate the analyte from the matrix. At least half of the elements of the periodic table have a volatile chemical form. Besides hydrides, it may be possible to form chelates, halides (AsCl3), oxides (OsO4), carbonyl ((Ni(CO)4), alkyls with Cd, Hg, Pb, Se, Sn, S-containing compounds (H2S). Hydride generation has been extended to unconventional elements such as Ag, Au, Cd, Co, Cu, Ni, Sn, Zn, Os, Pd,10 with Cd appearing as one of the most successful generations.11 Probably the most challenging research is to allow the simultaneous formation of several volatile species to keep the multi-element capability of the ICP-AES. Moreover, as mentioned above, study of plasma characteristics with a sample in the form of a vapor and with possible excess of foreign gas such as H2 should be conducted as water no longer acts as a load buffer.
Because the expertise in wet chemistry is disappearing, particularly when trace elements are of concern, the demand for direct solid analysis is growing. Laser ablation ICP-AES is one of the possibilities. It is evident that if a single or a few shots are performed, ICP-MS is more relevant because of its higher sensitivity. In contrast, LA-ICP-AES is highly suited to solid bulk analysis, because a large amount of ablated material can be injected into the plasma without the risk of blocking or having memory effects, as observed with an ICP-MS. However, the key point remains calibration. Even moving to the UV and to very short pulses, LA-ICP-AES (or MS) is still sensitive to matrix mismatch. As there is lack of matrix-matched and homogeneous standards, in particular for exotic samples, e.g., composite materials, polymers, glasses, ceramics, and environmental samples, calibration remains problematic for most samples when a high accuracy is requested. Considering the cost of a certified reference element (CRM), particularly if trace element and homogeneity at the few micrometres scale are a request, only a limited number of CRMs can be released. It is then necessary to develop a far more flexible calibration procedure, which means that the analytical process should not be too sensitive to the composition and the nature of the matrix. Alternative calibration methods, such as the use of wet aerosols, imply that atomization and excitation mechanisms should be similar regardless of the sample form. Note that diagnostics based on AES are more informative than those performed in ICP-MS, because of the access to different ionization states, excitation levels and optical transitions.
A PMT presents some major advantages, such as a large wavelength range, including the UV region down to 120 nm, a noise that is usually negligible at room temperature, thus not requiring any cooling device, and a high amplifier gain. However, compared with a photographic plate, the major drawback of a PMT is to be a single-channel detector. AES implies the emission of the spectrum of each element, which means that the use of a single channel detector leads to a drastic waste of information, even when several detectors are set up in a polychromator. There is then a need for a detector that associates photon-current conversion of a photoelectric detector and the richness of information of a photographic plate: this can be obtained by using a solid-state multichannel detector. This type of detector has progressively replaced the PMT since the beginning of the 1990s. In ICP-AES, multichannel detectors used with commercially available ICP-AES systems are currently based on charge transfer technology. Both charge-coupled device (CCD) and charge-injected device (CID) detectors are in use. Two dimensional CCD and CID detectors, or an association of linear CCD arrays, currently equip commercially available ICP-AES systems and permit a fast acquisition of the entire uv–visible spectrum. An alternative is the use of a solid-state detector leading to a spectral window of several nm, which is sequentially moved through the spectrum.
Benefits, or at least potential benefits, of multichannel detection may be divided into two groups. A first group is related to the richness of the acquired information, i.e., the entire uv–visible spectrum: (i) full flexibility in analytical line selection; (ii) use of several lines of the same element to extend the dynamic range; (iii) use of a large number of lines of the same element to improve accuracy and to verify possible matrix effects or spectral interferences; (iv) qualitative analysis; and (v) fast diagnostics. The second group is related to true simultaneous measurements: (i) speed of analysis; (ii) time correlation between lines of different elements to improve repeatability by internal standardization; and (iii) time correlation between line and adjacent background intensities to improve limits of detection and limits of quantitation.
The use of CTD detectors has revolutionized AES. However, these detectors still suffer from some limitations related to pixelation, UV response, dynamic range and shot noise. Because the spectral bandpass of a pixel is larger than the physical line width, the pixelation phenomenon results in the difficulty of obtaining a fair measurement of the peak intensity, and summation of pixels must be performed to the detriment of the practical resolution. When UV response is of concern, several techniques have to be used: lumogen coating, open electrode technology and backside illumination. Dynamic range, i.e., the ratio between the full well capacity and the readout noise, needs to be increased in order to facilitate the simultaneous measurement of lines with different intensities. This increase may be obtained by reducing the readout noise down to a few electrons RMS. Another significant limitation is the shot noise. For low signals such as background in the UV, the systems are usually shot noise limited, which necessitates long integration times to significantly decrease the relative standard deviation of the signal. This is particularly true when determining limits of detection. Moreover, time correlation between signals can only be observed if the non-correlated shot noise is not a limitation. In order to minimize shot noise, it would then be necessary to improve photon collection, regardless of the viewing mode, radial or axial. It may be said that there is no doubt that multichannel detection is highly beneficial to AES. At least for the time being, CCD and CID are appropriate detectors but not yet ideal.
The use of chemometrics should significantly improve data processing, and several publications have already shown the potential benefit of advanced statistics,7 based on the use of principal component analysis, multiple linear regression, fuzzy logic, artificial neural networks, and multivariate calibration. Implementation of chemometrics in software and better knowledge of matrix effects should result in the introduction of an ideal software that should ask a limited number of questions such as: (i) What is your matrix? (ii) Which elements do you want to determine? (iii) What are the expected concentrations? (iv) What precision is required? From the answers, and based on a data base, the software should suggest several sets of analytical lines and appropriate operating conditions, i.e., power, nebulizer gas flow rate and integration time, at least for the most common matrices.
There has been some work on signal modulation.16 Because of the availability of direct injection nebulizers, the damping effect of the spray chamber no longer exists as a problem, and a signal modulation would be easier to perform.
The other way is to reduce the background noise. There are three major causes of noise: flicker (multiplicative) noise, mostly originating at the sample introduction level (fluctuations of the aerosol cloud at the exit of the spray chamber/tip of the injector), shot noise due to the random emission and detection of the photons, and detector noise (dark current and readout noise in the case of solid-state detectors). Shot noise can usually be minimized by increasing the exposure/integration time. However, for limits of detection, integration time up to 1 min may be necessary as mentioned above, which, associated with a significant number of replicates, may lead to a real time penalty. Because the detector noise is negligible with PMTs and made negligible with solid-state detectors by cooling them to reduce dark current noise, and by selecting a detector with a low readout noise, the ultimate limitation should be flicker noise. It can be reduced by using a sample introduction system with less fluctuations. Several decades ago, a 1% RSD was estimated as a reasonable value for the background fluctuations, while values as low as 0.2% can be currently achieved. However, another possibility is based on truly simultaneous measurements. In this case, possible time correlation between signals can be used to reduce background noise. For that, shot noise should be negligible against flicker noise, as has been previously emphasized.
As regards non-spectral matrix interferences, I would like to quote the conclusions of a recent paper.17 “Since the beginning of the studies on ICP-AES, non-spectral matrix interferences have been the subject of numerous publications. It can be easily understood that the selection of a limited number of lines and elements can lead to various conclusions, according to the selection, the operating parameters, the viewing mode, and the ICP type. It is probable that the generator design and its tuning play a role in the magnitude of matrix effects. Use of a large number of lines per element illustrates how complex matrix effects are. Several processes may be involved, including possible different spatial distributions for these mechanisms. Knowledge of the origin of matrix effects remains probably one of the last challenges in ICP-AES in order to obtain highly accurate results. Not only should a large number of lines be used, but the same experiments should be performed on different ICP-AES instruments, probably through a collaborative study.” Because there is a large diversity of ICP-AES systems, with different generator technologies, torch designs, observation modes …, no single laboratory can cover this variety. It would be most worthwhile to define some key experiments with a well-defined protocol, including test elements, given matrices, range of operating conditions, diagnostics of the plasma …, which would be performed through several laboratories. This is certainly one of the most important remaining challenges, because the first quality that an analyst expects is accuracy, which cannot be obtained if calibration leads to a bias.
Miniaturization was not mentioned so far. Significant work has been devoted to microplasmas.18 However, miniaturization of the whole system is far more difficult. Compact dispersive systems have been described,6 but to the detriment of practical resolution, which is still a major concern in AES.
Many remaining challenges are related to instrumentation and should be, therefore, solved by R&D teams in instrument companies, with the possible collaboration of academic teams. However, there has been a drastic change in the strategy of companies where most efforts are directed to sales and not to an improvement in instrumentation and applications.19 This has been clearly reflected in recent conferences where almost no R&D papers were given by instrument companies. A decade or two ago, scientists with international recognition could work and publish while working for an instrument company, but it seems difficult to observe that nowadays. Therefore, can we perform more fundamental and applied research without the support of instrument companies, and if not, which body is able/willing to support this research?
Whatever the use of chemometrics, if matrix effects are present, accuracy will be poor. Their studies are interesting for any research group as most research topics are involved: diagnostics, mechanisms, role of the parameters, viewing mode, efficient use of internal standardization, influence of the sample introduction system, i.e., including fundamental, instrumental and application aspects. This raises another question about the relation with instrument companies: can companies take full benefit of the academic research?
As ICP-AES is still a most appropriate technique for elemental analysis, with some unique features and still subject to significant improvements, the reply to the question “is it still possible, necessary and beneficial to perform research in atomic emission spectrometry?” is clearly “yes”.
A last comment concerns the ICP-MS users. They should read results obtained in ICP-AES and not rediscover the wheel when considering plasma characteristics and capabilities.
This journal is © The Royal Society of Chemistry 2005 |