JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL. 8 767 Developments and Trends in Plasma Spectrochemistry-A View* Plenary Lecture Paul Boumans c/o Philips Research Laboratories WB7 P. 0. Box 80.000 5600 JA Eindhoven The Netherlands A review of the history of spectrochemistry over the past 40 years is taken as a starting-point for presenting a personal view on recent developments and current trends in plasma spectrochemistry . Booms and trends featured in graphical form are introductorily discussed. In that context an iconographic approach for the visualization of ratings of analysis methods is proposed. The main body of the discussion deals with assessments of the present position and future perspectives of furnace atomization plasma emission spectrometry plasma source mass spectrometry plasma source emission spectrometry and as a perhaps curious intruder and possibly serious competitor total-reflection X-ray fluorescence spectrometry.The potentials of recently introduced echelle array spectrometers for plasma source emission spectrometry are considered in the light of new or improved insights into the spectroscopic methodology and in the scope of input from developments in chemometrics. The paper is concluded with a brief review of the plasma spectrochemistry conferences in the past 18 years and their role for communication. Keywords Plasma spectrochemistry; graphite furnace plasma emission spectrometry; plasma source atomic emission spectrometry; plasma source mass spectrometry; X-ray fluorescence spectrometry Probably the term ‘Plasma Spectrochemistry’ appeared for the first time as the title of the Proceedings of the 1982 Winter Conference in 0rlando.l It followed the designation ‘Atomic Plasma Spectrochemical Analysis’ used in connec- tion with the first international Winter Conference held in San Juan de Puerto Rico in 1980.* The keyword in these expressions is ‘plasma’ an abbreviation of ‘plasma source’ a term which was introduced in the 1960s and has remained in use as a magic winged word up to the present day.The introduction of plasma sources in the 1960s marked the renaissance of atomic emission spectroscopy (AES).3 The term ‘plasma sources’ must then have been coined to show the analytical world that AES had rejuvenated itself by the invention and exploration of novel excitation sources whose capabilities were argued to extend far beyond the domains of arcs and sparks i.e.sources that the ‘parvenu’ of the 196Os atomic absorption spectroscopy (AAS) had marked out as old-fashioned and incapable. It therefore seemed psychologically better not to tell the AAS commu- nity that arcs and sparks were also ‘plasmas’ fundamentally not different from the new ‘plasma sources’ but to indoctrinate them with panegyrics singing the unique features of these ‘plasma sources’ a high temperature and an ability to accept aerosols of solutions in much the same way as combustion flames did. What is in a name if it can adequately serve as a battle-cry on the banners of crusaders who believed in their cause and made efforts to convince the community that ‘Plasma Spectrochemistry’ had to be identified with innovation and renovation? The belief and conviction of the crusaders and all their followers have in fact reshaped analysis in many respects including the extension of the umbrella ‘Plasma Spectrochemistry’ to cover also plasma source mass spectrometry (MS) glow discharge (GD) optical and mass spectrometry as well as particular areas of laser spectroscopy.It may be not only tempting but also inspiring to look at the past and to analyse and assess some developments in order to under- stand the present better and to extrapolate to the future. I will attempt to do this more or less under the motto ‘Plasma Spectrochemistry in search of innovation or ~~~ ~ *Presented in part as a Plenary Lecture at the opening of the 1993 European Winter Conference on Plasma Spectrochemistry Granada Spain January 10- 15 1993.confirmation?’ In that context I shall also pay attention to the communicative role of the Plasma Spectrochemistry Conferences in the United States and Europe including the European Plasma Spectrochemistry Conferences avant la lettre known as the ‘Noordwijk ICP Conferences’ (1 976 1 978).4*5 My intention is to present a view based on the insights I have gained during the nearly 40 years that I have been closely connected with analytical atomic spectrometry. The reader should thus not expect a review in which every statement is documented by references. Their number will be limited partly because an overdose of them would 1950 1960 1970 1980 1990 Year Fig.1 Characterization of the history of spectrochemistry in terms of ‘booms and trends’. The ordinate of each ‘whale’ depicts the ‘attention’ the relevant method received in the course of time whereby ‘attention’ should be understood to comprise research development and application. The scale is relative ‘within’ each method but does not include relative weights for the various methods. Acronyms are explained in Table 1768 Analytical and economic criteria chec klis f 0 Precision Accuracy Detection power Traces Minor/major 0 Solids metals 0 Solids non-conducting Liquids 0 Multi-element Selectivity Cornplexity/cost 0 Automation /cost JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL. 8 Table 1 Glossary of acronyms Table 2 Possible motives behind developments AAS AES A1 CCD CCP CID CMP DCP DSI ED ETV FANES FAPES FIA GC GD GF GIXA HA HC ICP LA LC MIP MS R.f.RSD RSDB RSDN SBR SNR TXRF WD XRFS Atomic absorption spectrometry Atomic emission spectrometry Artificial intelligence Charge coupled device Capacitively coupled plasma Charge injection device Capacitively coupled microwave plasma Direct current plasma Direct sample insertion Energy dispersive Electrothermal vaporization Furnace atomization non-thermal emission Furnace atomization plasma emission Flow injection analysis Gas chromatography Glow discharge Graphite furnace Glancing incidence X-ray analysis Hollow anode Hollow cathode Inductively coupled plasma Laser ablation Liquid chromatography Microwave-induced plasma Mass spectrometry Radiofrequency Relative standard deviation Relative standard deviation of the background Relative standard deviation of the net line signal Signal-to-background ratio Signal-to-noise ratio Total-reflection X-ray fluorescence spectrometry Wavelength dispersive X-ray fluorescence spectrometry spectrometry spectrometry detract from the view character of this paper but chiefly because it is impossible to retrace the roads that led to particular insights.Booms and Trends in Spectrochemistry a Brief Characterization Fig. 1 characterizes the history of spectrochemistry over the past 40 years. The acronyms used in this figure or elsewhere in the text are explained in Table 1. The picture shows a series of ‘whales’ representing the booms and trends on a relative scale for each method separately but does not cover the relative weights of the various methods.The diagram is approximate only and represents simply what I have in mind. The most important features are in the upper two-thirds of the diagram from XRFS-TXRF upwards. The trends of GFAAS-GFAES XRFS-TXRF plasma MS and plasma AES will be further discussed below. It may be interesting not only to consider the develop- ments themselves (Fig. l) but to look also briefly at the motives that have possibly driven these developments (Table 2). I leave it to analysts of human behaviour to make a balanced assessment but believe that the first four points stated in the table are the truly driving forces behind innovation. Why were scientists such as Greenfield Fassel Robin Mermet Ohls and I for example pioneers in the early years of the ICP? Because we believed in multi- element emission spectroscopy as a viable and indispen- sable analytical tool and also because we wished to avenge what we thought to be the initial cheap successes of the poor man’s spectroscopy of the 1960s atomic absorption! The reasons for these successes are clear (Table 3) which is why it was difficult to convince the man in the street that Urge to creativity Scientific ambition/competition Belief and conviction Challenges and dreams Social andor industrial needs Instrument markets Entrepreneurial making money Supply and demand Table 3 Reasons for initial success of flame AAS Direct anaiysis of liquids Analysis of dissolved solids (= isoformation) Many elements but typically one at a time Acceptable precision and accuracy Tailored to the minds and hands of wet chemists (sample treatment no problem!) Cheap and simple Spectroscopy of the ‘minute man’! Fig.2 Checklist with the most important features to be consi- dered in assessments of analysis methods there were better things to come! Admittedly the field of AAS has also seen radiant developments and has its enthusiastic and prominent pioneers such as Walsh Willis L’vov Slavin and others who believed in their causes and justly so! Much of the beliefs of both groups of missionaries the ‘emissionaries’ and the ‘absorptionists’ have been translated into hardware and software that now populate numerous laboratories. The first commercial ICP emission spectrometers became available in the mid-l970s resulting in the gradual accep- tance of the ICP as an analytical tool for the reasons shown in Table 4.Note in particular the last point which indicates that the emancipation of the wet analytical chemist has been an important factor! Assessing the Merits and Capabilities of Analytical Methods Generally speaking we may assess analytical methods with the aid of checklists of factors of the type shown in Fig. 2. Essentially these factors can be classified into two groups depending on whether they are connected with the sample or the method (Table 5). However this issue may also be approached in a modern graphical way using multivariate,JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL. 8 769 Table 4 Reasons for gradual acceptance of ICP-AES Analysis of liquids and dissolved solids Acceptable precision Simultaneous multi-element analysis Accuracy low multiplicative matrix effects Economy sample treatment for ICP far easier than for AAS Cost and complexity of AAS apparatus rose Emancipation of the wet ‘minute man’! Table 5 Sample-related and method-related criteria for assessing analytical methods Sample- Solid versus liquid Conducting versus non-conducting Trace versus minor versus major Detection power-range-selectivity Accuracy-precision Complexity-required expertise-cost Method- Maximum Minimum (a ) B Cost effectiveness I I Analytical figures of merit r\ Simplicity 0 0 00 A 0 Universality (sample types) Overall (zJ@ maximum Overall minimum Fig.3 Multivariate iconographic rating of analysis methods. (a) Icons representing four essential criteria to which the most important features of analysis methods can be reduced.The picture shows the maximum and minimum ‘values’ of the icons whereby ‘maximum’ refers to ‘best case’ and ‘minimum’ to ‘worst case’. (b) Assemblage of the icons into faces depicting the overall maximum and minimum ratings of an analysis method iconographic in which all factors are condensed to four essential features each represented by an icon with a minimum and a maximum [Fig. 3(a)]. A sophisticated computer program enables the user to assemble the icons into a four-dimensional picture as illustrated in Fig. 3 (b) which depicts the two extremes only. The power of the program is that it not only fits the correct face to the input data but can also provide a picture of the fate of a new analysis method as shown in Fig.4 whereby further icons are automatically added as new dimensions if the built-in artificial intelligence feels that the overall assessment should be enhanced. The program includes an automatic ‘bias protection feature’ in that the one-sided champion’s view is always supplemented by that of the adversaries. This approach will thus put a new analysis method always in the right perspective. Alluding to the above multivariate iconographic rating of analysis methods was not just for fun and entertainment only. In fact I did so as an introduction to an intriguing question about the reasons behind research projects. Fate of an analysis mefbod faced in phases and faces Actual finally lnit ia I Actual R expectation upon release Champions’ view @ 0 W Adversaries‘ view @ c~ 8 Fig.4 Fate of an analysis method faced in phases and faces Fig. 5 Causal relationships and chronology of the booms and trends in spectrochemistry. Arrows and connecting lines give a qualitative indication of the relationships and simultaneously the chronology. The diagram includes some diversifications within the scope of ‘methods’. Acronyms are explained in Table I Why are solutions often being sought to problems that others claimed to have solved 10 or 20 years ago using the same method or appear to have (better?) solved by other methods? Possible reasons are the following (1) Unawareness of existing solutions (2) No concrete evidence claimed solutions being (3) Context limitation claimed solutions being satis- (4) Redefined requirements extension of analytical (5) Technological innovation prompts. valid only ‘in principle’.factory only in a particular environment. capabilities or cost reduction needed. Apart from the trivial yet persistent unawareness of existing solutions ( l ) there are valid practical reasons (2-4) to embark upon research that confirms or extends the outcomes of earlier work while instigation by technological innovation (5) is doubtless a most laudable incentive. Below I shall point out a few developments in which innovation prompts have played a major role.770 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL. 8 l:i I 0 HC-FANES FAPES 0 Grimm 0 Boosted 0 Jet- enhanced Fig. 6 Block diagram of Fig. 5 stripped of items that have at present lost much of their attraction and topicality.The stripping covers both entire methods (‘defunct whales’ in Fig. 1) and diversifications (Fig. 5) that have flopped or turned out to be less world shocking than originally anticipated Booms and Trends in Blocks and Boxes It may be instructive to discuss current developments against the background of a self-explanatory block diagram (Fig. 5) that reveals simultaneously the causal relationships in and the chronology of the booms and trends in spectrochemistry. Obviously the parallelism between chro- nology and causal relationships is not at all surprising. In contrast what is surprising is what remains of the block diagram if it is stripped of the items that have lost all or most of their attraction and novelty as shown in Fig. 6.Only a few methods are seen to be really defunct most of them being still wholly or partly alive or even in full blossom in agreement with the tenacity of the correspond- ing ‘whales’ in Fig. I. This must be due to innovation within the scope of the methods leading to an extension of their capabilities and fields of application. Innovations Extensions of GFAAS Among the extensions of GFAAS (Fig. 5 ) several versions have ‘ES’ as part of their acronyms which reflects a trend from GFAAS to GFAES. The original GFAES as well as HA-FANES have more or less flopped while HC-FANES as a method for special purposes copes in particular with the problems of molecular background and adequate correction for this.* The most viable and interesting version appears to be FAPES (including GF-CCP) which uses a graphite furnace additionally heated by a radiofrequency dis- ~ h a r g e .~ - l ~ This development will enhance the possibility of extending graphite furnace techniques into the field of simultaneous multi-element analysis. It is interesting to see that the research on the GF-CCP started by Liang and Blades9J3 some 6 years ago has resulted in the development and commercialization of equipment for both emission and absorption.14 A report on the detection limits attained with this device shows striking improvements for a number of elements whose detection limits form a ‘weak‘ spot of ICP performan~e.~~ These improvements by one or two orders of magnitude with respect to the values reported by Winge et a1.,l6 concern in particular elements such as Ag As Cd Pb Sb Se Te T1 and Zn.In summary this work marks an innovation that supplements and extends (a) the GF technique with the possibility of multi-element analysis and (b) the ICP method by providing essentially improved detection limits for a particular set of elements without requiring operation in a vacuum although adaptations for use at various pressures can be made. Overall FAPES is a refreshing development not only in the sense of technological innova- tion but also politically as its acceptance in analysis might help to bridge the gap between the AES and AAS factions in the spectrochemistry community! Total-reflection X-ray Fluorescence Spectrometry A totally different aspect of ‘booms and trends’ is included in the ‘X-ray tree’ at the far right of Figs. 5 and 6 the birth and development of total-reflection X-ray fluorescence spectrometry for which the acronym TXRF has been generally adopted.X-ray fluorescence spectrometry (XRFS) was started in the mid 1950s as wavelength dispersive (WD) XRFS and later extended by energy dispersive (ED) XRFS. We optical spectroscopists know that XRFS is being used worldwide as a powerful tool in both research and industry. We also ‘know that XRFS leads a life of its own usually far away from our optical spectroscopic laboratories in particular research laboratories. The situation is certainly different in industries and government or private organizations where laboratory managers have to make the optimum choice of analysis methods to run their analyses efficiently and cost effectively. It would be useful however if also in optical spectroscopic research laboratories more attention would be paid to what happens in the X-ray field.Perhaps many optical spectroscopists are prejudiced about XRFS they think that the method deals with metals or geological samples requires large samples cannot handle solutions is precise but inaccurate has poor detection limits and is extravagantly expensive. That picture may have applied around 1960 but is no longer valid. An extremely important innovation is TXRF. Recent reviews may serve as key referencesI7-l9 and introductions to the principles of this method. Essentially in TXRF the sample is applied for example as a thin layer of a dried solution or slurry on a carrier of quartz glass or glassy carbon.X-ray radiation from a W or Mo tube is directed to a reflecting mirror or monochromator for selective filtration and subsequently to the sample at a glancing angle of about 4 min. As a result of this grazing incidence the beam is totally reflected at the carrier surface. The thin sample layer on the carrier is passed through by the primary and the totally-reflected beam whereby it is excited to X-ray fluorescence. The emitted radiation is detected by an Si(Li) detector mounted directly above the sample and is recorded as an energy dispersive spectrum. The essential benefits and limitations of this approach are summarized in Table 6 while conspicuous features are shown in Table 7. Note here that matrix separation is required not to avoid matrix effects but to achieve good detection limits. The amount of scattered radiation and hence the background increases proportionally to the matrix concentration.Also note the hidden or perhaps denied (Fig. 4!) question about the possible problems inherent in the drying step. Optical spectroscopists may consider all this as interest- ing but still take it for granted! However why is bringing up TRXF important in the context of plasma spectrochemis- try? Table 8 provides an answer. The crucial point is that TXRF brings XRFS into the hands of the wet chemist who in spite of all instrumental developments in the past 40 years has remained a key figure in the analytical chemistry picture. The question at the bottom of Table 8 has thus not been added without reason (cJ Tables 3 and 4).AnotherJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL. 8 77 1 I 1 \ I ~ ~~ Table 6 Essential benefits and limitations of TXRF Substantial reduction of the background intensity and duplication of the line intensity leads to improved SNR and detection limits (three orders of magnitude compared with classical EDXRFS) Use of thin specimens provides for simple and reliable quantifica- tion No matrix effects Fixed element-specific sensitivities Single internal standard for universal calibration curve requirement of separations Fig. 4) However a matrix worsens the detection limits entailing the Hidden or denied question is the drying step free of problems? (cJ ~ ~ Table 7 Conspicuous features of TXRF Microanalysis < 1 pg solid < 10 pl solution Trace analysis picogram level detection limits Direct Aqueous solution 1 pg ml-’ Acids 10 pg ml-1 Biological materials 10 ng ml-1 Ultra-pure metals 10 ng g-’ After matrix separation Multi-element capability 80 elements above Z = 11 (Na) Fields of application environmental mineralogical medical biological ultrapure reagents surfaces thin layers,.. . . . . . . . . . . Table 8 General reasons for bringing TXRF into the context of plasma spectrochemistry TXRF versus optical methods including MS TXRF is competitive or even more capable TXRF can handle solutions slurries powders tissues thin TXRF uses similar sample preparation methods TXRF fits the minds and hands of wet chemists TXRF is not fabulously expensive! Will history repeat itselj? layers.. . . . . Actual finally expectation upon release Initial A I Champions‘ view @ view Adversaries’ @ r7 8 Fig. 7 Fate of an analysis method faced in phases and faces with changed caps alluding to the competitive position of TXRF and illustrating that it is not the cap but the face that makes up the substance in the proposed iconographic rating of analysis methods IT= I ____________________.------ I I I I - 1950 1960 1970 1980 1990 GF-CCP question raised above in the section Assessing the Merits and Capabilities of Analytical Methods might be repeated here in a somewhat modified form Why are solutions often being sought to problems that others claim to have (better?) solved by other methods? Obviously the reasons are not different from those brought up above. Possibly TXRF may become a lever for bridging the gap between X-ray and optical spectroscopy and more impor- tantly between the two factions X-ray and optical spectros- copists.As an attempt to make the factions approach each other I mention the publication of the proceedings of a biennial workshop on TXRF in Spectrochirnica Acta Part B.20-22 To conclude I believe that the efficiency of research as well as instrument and methods development in atomic spectroscopy could substantially benefit if both groups would have a better understanding of each other’s fields. Whether familiarization with the principles and applica- tions of TXRF will induce plasma spectroscopists to change caps or not it is not the cap but the face that makes up the substance as illustrated in Fig. 7.Plasma Source AES and Plasma Source MS Fig. 8 brings us back to plasma spectrochemistry with the projection of a part of the ‘booms and trends’ picture (Fig. Fig. 8 Projection of a part of the ‘booms and trends’ picture of Fig. 1 on the block diagram of Fig. 6 with plasma source AES and plasma source MS highlighted 1) on a simplified version of the block diagram of Fig. 5 in which a blank oval puts plasma source AES and plasma source MS into the limelight. The oval might tempt the adversaries of these techniques to raise the ‘hopeful’ question symbolized in Fig. 9 but that is not what the iconographic rating program provides. On the contrary the program automatically tilts the oval and shows a family portrait with two friendly happy snowmen shaking hands (Fig.10). Their faces are not entirely identical but I had to accept what the computer dictated and the computer did so because I had furnished the input impressions derived from some recent conferences verbally summarized in Table 9. The last point stated in this table is emphasized and expanded in the block diagram ‘under construction’ depicted in Fig. 11 with in the right-hand upper corner a rising sun called ‘arc’. One reason for this is the present trend to explore and exploit solid sampling using such methods as direct sample insertion slurries and laser ablation. These approaches provide us of course with some of the possibilities of the arc avoiding however its instability but also introducing again many problems connected with solid sampling such as incomplete evapora-772 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL.8 1950 imo 1970 1980 iwo Fig. 9 Part 1 of a strip story Fig. 8 with a ‘face fit’ symbolizing a possible ‘adversaries’ view’ I BOOMS s. TRENDS^ GFMS - GFAES Fig. 10 Part 2 of the strip story Fig. 8 with a ‘face fi approximately symbolizing current reality tion selective distillation and thermochemical reactions. On the other hand selective distillation and thermochemi- cal reactions may be exploited to obtain emission or mass spectra of relatively volatile traces stripped of a possibly cumbersome spectrum of a refractive matrix in much the same way as the old ‘carrier distillation’ method23 or its numerous variants (cf ref. 24). That rising sun however has more to tell as preluded by the computer output faced in Fig.12 showing one of our two snowmen in a medallion happy and chuckling rubbing his hands! The message of the rising sun in Fig.13 is labelled ‘photographic plate’ but not really that plate itself but the arc methodology connected with it multi-line multi-element analysis and general survey analysis. Actu- ally the reason for the existence of this sun lies in another sun called ‘array’ (Fig. 14) which term refers to some major recent advances in the development of echelle spectrometers with several types of array detectors. Plasma Source AES Array Spectrometers This development reminded me of a perhaps long-forgotten article by Margoshes ‘Data acquisition and computation in Table 9 Brief assessment of the impact of plasma source MS on plasma source AES and the current positions of these two methods Initially strong ‘drain’from AES to MS- MS promised very low detection limits MS promised ‘no’ matrix effects MS promised much higher selectivity Both stabilizing in mutual interaction AES and MS mutually supplementary-combined systems? Severest handicap of MS matrix effects; ‘no’ has evaporated into Plasma source chiefly argon ICP MIP-AES for special purposes (e.g.in GC or LC) He-MIP-MS as supplement (Fe As Se Br) Chiefly quadrupole mass spectrometers Double focusing high-resolution mass spectrometers entering Ion trap and time-of-flight MS cradled MS appears ‘stagnant’ in details of applications AES and MS are both ‘fighting’ with solids At present- the spectrometer! practice spectrochemical analysis’,25 from which I might cite two statements ‘.. . . . .A spectrometer employing a television camera tube could have many advantages.. .However many problems must be solved before such an instrument could be made practical.. .’ It appears that this 1970 Midsummer- Night’s Dream has finally become a reality! Denton and co-workers in particular must be credited with having performed extensive ~ t u d i e s ~ ~ - ~ ~ aimed at the implementation of two-dimensional (2-D) array detectors in ICP-AES. Various efforts have resulted in the develop- ment of commercial instruments using different types of echelle spectrometers with 2-D dispersion and different types of 2-D array detectors briefly characterized by the data given in Table 1 033,34 and Table 1 1 ,35,36 respectively.The future will have to show how these systems perform in practice and how the constraints inherent in array detectors have been compromised to meet the requirements of wide spectral range small spectral bandwidth high detector resolution (=number of pixels per unit of bandwidth) signal-to-noise ratio ruggedness and cost effectiveness. Improved Insights and Chemometrics Fig. 14 contains an overlaid block with the headings ‘Improved insights into the spectroscopic methodology’ and ‘Chemometrics’ suggested as an input item for ‘Plasma source AES’ on the same level as input item ‘Echelle spectrometer-array detector’. The suggestion implies that the advent of echelle spectrometers with array detectors opens perspectives for advanced applications of recently gained or improved insights into the spectroscopic metho- dology in general as well as applications of results of new developments in chemometrics.In the following sections I shall recapitulate a few issues as a basis for pointing out some directions from which further developments can be expected. Application of Improved Insights in Plasma Source AES Array Spectrometers and SBR-RSDB Approach It is no secret that in recent years I have crusaded to convince the world that the use of signal-to-background ratio (SBR) in combination with the relative standard deviation of the background (RSDB) has a number of advantages over the approach using signal noise and signal-773 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL. 8 ICP MIP Fig.11 Phase 1 of the construction of a block diagram emphasiz- ing the exploration of techniques for solid sampling in plasma source AES and MS. The rising sun symbolizes these developments as a reincarnation of the ‘good old arc’. Acronyms are exdained in Table 1 in spectrochemistry 1950 1960 1970 1980 1990 W ) W Fig. 12 Part 3 of the strip story Fig. 8 with a ‘face fit’ symbolizing and emphasizing another current reality and its perspectives to-noise ratio (SNR). To convince a person or a group of people of the appropriateness of a thesis a proposal or an idea it does not suffice to show ‘that something works’ or even ‘that it works better than something else’ one should also show why it works better and what is gained. This requires one to penetrate profoundly not only into the approach one wishes to recommend but also into the approach one wishes to banish.In the case of the SBR-RSDB and SNR approaches there was fortunately nothing to banish but a lot to compare and above all to connect. I have tried to do that in a series of lectures the preparation of which as well as the discussions following them have gradually enhanced my own insight into this delicate matter with the result that I could eventually let this insight culminate in a paper ‘Atomic emission detec- tion limits more than incidental analytical figures of merit!-a tutorial discussion of the differences and links between two complementary appro ache^'.^' This paper along with three papers dealing with the theory and measurement of detection limits in ICP-AES using the SBR-RSDB a p p r ~ a c h ~ * - ~ ~ and the preceding outpost skir- mishes in the lectures have probably prompted the defini- tive infection of the North American spectroscopy commu- nity with the SBR-RSDB virus.This happened some 45 years after Kaiser published the fundamental work ‘Die Berechnung der Nachwei~empfindlichkeit’,~~~~~ the nursery from which the virus spread over Europe. At present the SBR-RSDB virus appears to have settled well in North America even to the extent that Hieftje’s group has launched within the incubation period a program for data collection and calculation of figures of merit using the SBR-RSDB as an electronic publication in Spectrochimica Acta Ele~tronica,~~ while the approach has been recently applied in the scope of the development assessment and performance testing of a new type of Cchelle array s p e c t r ~ m e t e r .~ ~ - ~ ~ It may be interesting to look in that context again at the SBR-RSDB approach and to show in which respect this approach also provides a convenient framework for understanding why and how array detectors may bring us an eagerly sought after crucial improvement of the precision in ICP-AES and a concomitant improve- ment in the detection limits. As an aid for appreciating the arguments the following tutorial introduction may be useful. Basically the SBR-RSDB approach involves three vital equations from which other equations may be developed for specific situation^.^^-^^ First we have the equation for the detection limit (cL) (1) CO c,= k x 0.0 1 x RSDB= - SBR where c is the analyte concentration to which the SBR applies and k is a statistical factor chosen to be 2 2 2 or 3.Note that in this equation RSDB is expressed in %. The same will apply to the related quantities RSDN aA and aB introduced in the equations that follow. The second equation specifies RSDB:37-39 Y RSDB= in terms of coefficients for source flicker noise (aB) shot noise (D) and detector noise ( y ) and the background signal (X,). Eqn. (2) has been further developed into a form convenient for dealing with the detector noise,39 but it would only unnecessarily complicate the present discussion to consider that equation here. Let it suffice to say that detector noise may play a predominant role in commercial instruments using photomultiplier tubes (PMT) in particu- lar in the low UV spectral region,4o and also in instruments with array detectors unless they satisfy particular require- ments as indicated below.In the present tutorial we may confine discussion to situations with a negligible detector noise contribution (7x0 and/or XB is large) r--- The third equation formulates the relative standard deviation (RSDN) of the net line signal (X,) as a function of SBR X and the various noise coefficient^:^^-^^ whereby an additional coefficient a covers the flicker774 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL. 8 Arc methodology Photographic plate 0 Multiline multi- element analysis 7-1 0 General survey I analysis j U Plasma source AES ICP 0 MIP Fig. 13 Phase 2 of the construction of the block diagram (started in Fig.1 1 ) emphasizing the predictable rebirth of the essential arc methodology prompted by the reshaping of the ‘photographic plate’ in the form of array detectors (Fig. 14) 0 General survey 0 Spectrum resolution ICP 0 W I ? Spectrum simulation True detection limit 0 tine selection and Al Chemometrics Multivariate analysis 0 Multicomponent L analysis Kalman filters Fig. 14 Phase 3 of the construction of the block diagram (Figs. I 1 and 13) emphasizing not only the foreseeable renaissance of the essential arc methodology induced by the recent maturing of echelle spectrometers with array detectors but also expansions of this methodology as a result of the exploitation implementation and further exploration of recently gained insights into the spectroscopic methodology and advancements in the field of chemometrics noise in the net line signal.Coefficient aA may be of the same magnitude as a but this is not necessarily so. If we assume that a A = a B then eqn. (3) may be conveniently written as I This result is obtained by combining eqns. (I) (2) and (3). Finally if y=O and SBR is large then eqn. (3) may be converted into I Table 10 Brief characterization of Thermo Jarrell Ash IRIS-CID ichelle spe~trometer~~,~~ Conventional echelle arrangement with toroidal camera mirror CIDI 7BAS array (CIDTEC Liverpool NY USA) pixel 23 pm x 27 pm 388 pixels x 244 pixels 8.7 mm x 6.6 mm !Special phosphor to increase quantum efficiency in UV Cryogenic cooling to 135 K !Spectral bandwidth 20 pm !Spectral range 170-800 nm which is the net line equivalent of eqn.(2a) for the background. The similarity of the two equations is immedi- ately understood if one considers that a large SBR in fact means that the background is negligible. With the present state of the art we must be generally satisfied with values of the flicker noise coefficients of about 1% or slightly lower. Therefore the minimum value of RSDB will also be about 1 O/o and this condition is met if the background signal X (as the resultant of the background radiant flux and the detector sensitivity) is high enough to make the shot noise and detector noise terms in eqn. (2) negligible. As noted above with PMTs and most array detectors this cannot be achieved in the low UV. However whenever it can be realized we have a situation in which the flicker noise term dictates the lower limit of RSDB.A property of array detectors is that virtually simulta- neous measurements can be made at adjacent wavelengths in the spectrum. Since it is likely that the intensity signals measured at such wavelengths are correlated one may exploit the correlations involved by instantaneous ratioing of the intensities. It is irrelevant here how this is precisely implemented in the hardware and software. It is essential that one can define a ‘measure’ in terms of a ratio either a classical quotient of just two signals or in a multivariate form and that the flicker noise of this ratio will approach zero if the correlations are intrinsically perfect and are fully exploited. Consequently if we also represent in this case the flicker noise coefficient by aB then the RSDB will be limited by shot noise and detector noise rather than flicker noise.Such a situation is represented in Fig. 15 not as an innovation but as a revitalization of an ‘old truth’ reproduced from a paper dealing with PMTs published in 198 1 .46 Actually the figure published was not for RSDB but for RSDN. However it referred to conditions where eqns. (2a) and (3b) are valid whence it may be taken to represent both RSDB as a function of X and RSDN as a function of X,. What is crucial is that the diagram shows the experi- mental curve of the RSDB or RSDN to reach a constant level of 1% when the intensity exceeds a particular value. This constant level is the flicker noise limit. In the original publication the curve for the shot noise component was included to show that the RSDB or RSDN is entirely dictated by shot noise at the lower end of the intensity range.In contrast in the present context it is the upper end of the intensity range which deserves our attention since array detectors can permit a switch to the curve for the shot noise component as the ‘experimental’ curve. This has been recently demonstrated by Barnard et af.36 and Ivaldi and B a ~ n a r d ~ ~ to be a reality for the array spectrometer they explored. This point is documented here by merely reproducing a figure from the latter paper as Fig. 16 whereby the reader is referred to that paper for explanations. A vital requirement for obtaining results of the type shown in Fig. 16 is that a sufficiently large background signal be obtained.Otherwise the working point will lie at the lower end of the range where shot noise is predominant making any attempt to exploit the correlations in the flickerJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL. 8 775 ~~ Table 11 Perkin-Elmer segmented CCD array (SCD) echelle spectrometeP~~~ Non-conventional echelle spectrometer- Separate output regions for UV and VIS Each optimized for resolution and throughput Range 167-782 nm Bandwidth 6 pm at 220 nm High throughput luminosity per unit A = 72 x luminosity of Thermo Jarrell Ash IRIS system Segmented CCD array detector (SCD)- Custom integrated circuit 224 linear array segments totalling Covering 5.7% of the spectrum (1 67-782 nm) Potentially 5000 lines 3-4 primary lines of each of 72 elements Pixel width 12.5 pm Pixel height 80-1 70 pm Detector cooled to -40 “C Virtual gate buried-channel structures Front side illumination No overlaying poly-silicon gate structure-maximum UV quantum efficiency David Sarnoff Research Center Pinceton NJ 16 photolithography mask layers 300 semiconductor processing steps 6336 pixels 0.1 I 1 x lo2 1 x 10’ 1 x lo‘ 1 x 10’ 1 x lo‘ I 1 I J Background or net line signal (abitrary units) Fig.15 Relative standard deviation of background (RSDB) or net line signal (RSDN) in dependence on the corresponding back- ground (A’,) or net line signal (X,) respectively. The diagram shows the experimental curve for an EM1 9789 QA PMT reflecting the contributions from source flicker noise and shot noise and the curve for the shot noise component alone as derived from the experimental curve by a theoretical analysis.Reproduced with permission from Boumans P.W.J.M. McKenna R. J. and Bosveld M. Spectrochim. Acta Part B 198 1 36 103 1 noise ineffective. A high optical throughput of the spectro- meter and a high quantum efficiency of the detector in turn are the prerequisites for obtaining a high background signal since the ‘intensity resources’ at the side of the ICP are virtually exhausted. This has been achieved in the pertinent spectrometer (cJ Table 11). What consequences has the reduction of the flicker noise on analytical figures of merit? First if we are able to reduce RSDB from 1% to for example O.l% this implies according to eqn.(1) an improvement of the detection limits by a factor of 10. On the other hand the precision of the net line signal at the detection limit will fundamentally not improve since this precision is entirely dictated by the second term under the square root sign in eqn. (3a). This I 0.1 1 9 -<. 0 %s 10 100 1000 10000 Background sig na I/cou n ts s-’ Fig. 16 Theoretical and experimental plots of RSDB versus background signal as obtained with a segmented CCD array (‘SCD’) Cchelle spectrometer (Table 1 1). All curves are theoretically derived approximations representing total noise shot noise and flicker noise respectively whereas the points are experimental data. In contrast to Fig. 15 the present figure includes experimental points for the shot noise component alone as obtained through ‘multicomponent spectral fitting’.The diagram demonstrates that the array spectrometer permits the elimination of the source flicker noise component from the experimental results. Reproduced with permission from Ivaldi J. C. and Barnard T. W. Spectrochim. Acta Part B 1993 (ref. 45). precision in terms of RSDN would thus remain 50% if k = 2 G and about 50% if k=3. However this precision is reached at a 10 times lower detection limit while the limit of determination (RSDN= 10%) now coincides with the ‘former’ detection limit. If correlations in net line signals can be exploited in a similar way an essential gain in the overall precision of the method comes within sight. This is illustrated with the aid of a classical picture which I must have shown for the first time in 197847 and has since appeared in various publica- tions or tutorials e.g.refs. 39 and 48. It is reproduced here in a slightly modified form as Fig. 17. The diagram is essentially a representation of eqn. (3a). The curves are for various values of a namely 5 2 1 and 0.5% respectively. For obvious reasons the curves have a shape similar to that of the experimental curve in Fig. 15. Also for obvious reasons each curve reaches a plateau at a relatively high analyte concentration corresponding to a value of for example c/c,= 500 which in turn corresponds to SBRs of 70 28 14 and 7 for the respective curves from top to bottom; in the plateaus RSDNSCY,. Fig. 17 can be extended with curves for still smaller values of CY repre- senting situations that may be realized with array detectors if the correlations (or anticorrelations?) in the flicker noise of the line signals at different wavelengths are well exploited.The set of curves might eventually include the shot noise limit shown as the broken line in Fig. 17. If this could be actually achieved in ICP-AES the precision of ICP-AES would become competitive with the precision of XRFS which is high (RSDNzO. 1%) because XRFS does not cope with the critical feature of optical methods flicker noise arising from fluctuations in sample introduction and its repercussions on the source conditions. Line Selection in ICP-AES The overlaid block in Fig. 14 includes ‘line selection and AI’ (artificial intelligence) as a separate item. However ‘line selection’ does not stand on its own but is firmly linked with all the other topics listed in the block.Two questions are of vital interest here. Firstly what progress has been776 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL. 8 50 10 A - $ 5 z P v) a 1 0.5 - Limit of \ a A 5 2 1 I 0.5 determination \ \ 1 1 1 \ I 1 50 100 5001000 0.1 1 5 10 C/CL Fig. 17 Plots of the relative standard deviation of the net line signal (RSDN) as a function of the ratio of the concentration present to the detection limit (c/cL) according to eqn. (3a) for various values of the source flicker noise coefficient (a,) 5 2 1 0.5% and 0 respectively. Adapted and reproduced with permission from Boumans P. W. J. M. Fresenius’ Z. Anal. Chem. 1979,299 337 made in recent years? Secondly what will be the impact of array spectrometers and comparable multichannel instru- ments? The former question is considered in this section the latter in the next.The introduction and the gradual understanding and penetration of the concept of ‘true detection limit’ has induced researchers to devise and implement chemometric procedures not only to facilitate line selection but also to correct for spectral interferences due to line overlap. The concept introduced and elaborated in the second half of the 1 980s,49-54 sprouted from our studies on high-resolution spectrometry and our initial disappointment about the fact that high resolution brought only rather marginal profits for the detection limits if the contribution from an interfering line to a gross analyte line signal was treated in the same way as continuous backgro~nd.~~ Accepting this hard ‘truth’ was difficult.The problem was not a failure of high- resolution spectroscopy which could not be reproached for not behaving as it should physically. The case in point was that the contribution from an interfering line should be brought into the context of ‘selectivity’ and not be consi- dered as if it were merely a simple addition to the background under the analysis 1ine.52~53 What this means is elucidated in Fig. 18 which shows a line that is thought partly to overlap an analysis line (not shown) thus contributing an interfering signal XI at the peak wavelength Aa of the analysis line. The picture further shows the contributions from the original background (XB) and back- ground due to a continuum or line wings (X,) produced by the interferent. The basic idea was that continuous back- ground (even if slightly sloping) can be accurately deter- mined by one- or two-point measurements.However this is impossible with the interfering line signal since an accurate measurement would make demands on the accuracy of the wavelength positioning device of the spectrometer that are incompatible with the actual state of the art. This problem manifests itself not only with the measurement of XI in the situation shown in Fig. 18 i.e. for the spectrum of the pure interferent but even more seriously in the sample spec- trum where we can only observe the resultant profile of analysis line and interfering line (Fig. 19). In conclusion an interfering line signal had to be considered as a contribution t o the background that can be measured only with a far larger uncertainty than continuous background if tradi- tional peak height measurements are applied.It was this insight that prompted the introduction of the concept of k Wavelength - Fig. 18 Schematic representation of the situation that results if an interfering line partly overlaps an analysis line having its peak at wavelength la. The interfering signal is made up from three distinct contributions X,=interfering line signal; Xw=interfering continuum signal; X,= original background (pure water) I I 1 Wavelength - Fig. 19 Line profiles observed for a blank solution of an interferent and the same solution spiked with analyte. The peak of the analysis line is located at wavelength la.To determine the net line signal (X,) accurately it is required to make an accurate measurement of the gross signal (X,+ X,+X,) the interfering line signal (XI) and the background signal (XB). The peak height measuring mode precludes accurate measurement of the former two owing to wavelength positioning errors in a spectrometer. This contrasts with the scanning mode in combination with a multivari- ate technique. Reproduced with permission from Boumans P. W. J. M. and Vrakking J. J. A. M. Spectrochim. Acta Part B 1987 42 819 ‘true detection limit’ (cL to be distinguished from the ‘conventional detection limit’ (cL ,,,,) and the ‘common detection limit’ (cL) for pure water as follows. The detection limit for pure water (or diluted acid) is defined by eqn.(I) which may also be written in terms of the background equivalent concentration cBEQ instead of the quotient cdSBR (4) C = k x 0.0 1 x RSDB x cBECJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL. 8 C may be ') (iterative calculation) (iterative operation) (' must be '1 (' may be '1 777 With reference to Fig. 18 the conventional detection limit is then defined analogously where cWEQ and cIEq are the equivalent concentrations associated with the signals Xw and XI the sting being in obtaining a realistic value of RSDB in order to let c ~ ~ ~ ~ ~ reflect a realistic situation! Various considerations have prompted the introduction of the true detection limit according to this definition V CL true= - 5 CIEC + CL conv (6) where v is a parameter whose value is likely to lie between 0 and 2.In eqns. (5) and (6) we may now use a value for RSDB of the order of 1% such as it would be derived from a series of intensity integrations at La with a fixed wave- length setting of the spectrometer while a solution of the pure interferent is nebulized. This is so because we have made a distinction between two detection limits an unrealistic one cL conv and a realistic one cL The latter not only provides a fair estimate of the detection limit that can actually be achieved under real analysis conditions but is also by excellence the criterion for line selection in trace analysis of real The implementation of the cL true concept in ICP-AES has been illustrated in various p ~ b l i c a t i o n s ~ ~ ~ ~ ~ ~ ~ ~ while an electronic publication on spectrum simulation has made the tools available for the 'readers' to perform case studies on their own computers.57 However all that work emphasized the diagnostic aspects of the problem but did not directly contribute to remedying the disease i.e.providing means for nullifying or at least reducing the disastrous effect of line interference on the true detection limit. It is the merit of van Veen and c o - w o r k e r ~ ~ ~ - ~ ~ that they were the first to meet the challenge and to take up the gauntlet. Their work involves the application of a multicomponent or multivari- ate technique known as 'Kalman filtering'. This approach has brought the desired 'release' as is well explained in the pubications of van Veen and co-workers.The essence is the judicious collection of such additional spectral information of analytes and interferents that the magnitude of interfer- ing line signals (XI) can be accurately 'predicted' which in terms of eqn. (6) means that the value of parameter v is reduced to zero. This additional information is obtained from scans instead of simple peak height measurements. If the value of v is reduced to zero the true detection limit becomes equal to the conventional detection limit. The risk of line interference is thus diminished to that produced by a simple background enhancement equalling X I . A further consequence is that the conventional detection limit may now be used as the criterion for line selection which avoids some complications and is more straightforward.The approach also entails an appreciable relief of the require- ments upon the dynamic range of databases for the initial tentative line selection preceding the definitive measure- ments involving Kalman filtering.62 Yang and c o - ~ o r k e r s ~ ~ - ~ ~ have made an in-depth analysis of the factors implied in parameter v in eqn. (6) and also established under which conditions Kalman filtering tends to bring the sketched profits and when not. Ivaldi et al.67 have tested an alternative multivariate method for increas- ing the quality of the information extracted from ICP spectra also involving spectral scans instead of peak height measurements. Their publication provides keys to addi- tional literature. I off-line (available data base) on-line (ad hoc data collection) 1 a prior/ ( single element determinations (' may be ' I 2 multielement analysis off -I ine on-line multielement analysis a posterlor/ I multiline analysis I I Fig.20 Schematic representation of possible protocols for line selection in AES trace analysis as explained in the text Pure substance simple spectrum Pure substance complex spectrum ' Known ' simple mlxture I \ Istmpie/complex spectrum I I I Requirements I 0 Criterion 0 Data base 0 Knowledge of sample composition slmplelcomplex spectrum U Fig. 21 Schematic representation of situations permitting in principle a priorz line selection in AES trace analysis. The conditions grow steadily more difficult in the direction of the arrow making increasingly more severe demands upon data base knowledge of the sample composition and spectrometer flexibility although strictly speaking a posteriori line selection becomes mandatory only if all knowledge on the sample composition is lacking (Fig.22) Line Selection and Multichannel Spectrometers Generally we may classify approaches to line selection in AES trace analysis on the basis of the schedule presented in Fig. 20. Accordingly the first possibility is a priori line selection which may be done either off-line or on-line. In the former case we must have a database available in the latter we collect dedicated data ad hoc. This approach may be applied to single-element determinations or to multi- element analysis. More powerful is a posteriori line selec- tion. The off-line mode implies iterative calculations using the data measured in a complete spectrum such as happened in general survey analysis using a d.c.arc and photographic detection with an advanced automated den- sitometer.'j* The on-line mode implies not only iterative calculations but also iterative operation to collect data. It is feasible only with multichannel detection. Application of either off-line or on-line a posteriori line selection requires the analysis to be a multi-element analysis since the results for the one element are needed for decision making with respect to the other. It may be a multi-line or multiple line analysis i.e. a multi-element analysis using more than one analysis line per analyte. This is the ideal approach but its implementation requires the availability of hardware and software that provide rapid access to several hundred spectral lines.This was actually so with the now defunct d.c.778 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL. 8 ~~ Table 12 Plasma Spectrochemistry Conferences Prehistorical Time the Stone Age- ICP Information Newsletter founded ICP Round Table Discussion XVIlI CSI First European ICP Symposium Amherst MA USA Grenoble France Munich Germany Transient Era the Golden Age- Noordwijk aan Zee The Netherlands First ICP Conference Noordwijk aan Zee The Netherlands Second ICP Conference San Juan Puerto Rico International Winter Conference on Atomic Spectrochemical Analysis with ICPs MIPS and DCPs Historical Time the Silicon Age- Orlando FL USA San Diego CA USA Leysin Switzerland Kailua-Kona HI USA Lyon France San Diego CA USA Reutte Austria St.Petersburg FL USA Dortmund Germany San Diego CA USA Granada Spain San Diego CA USA June 1976 September 1975 June 1976 June 1976 April 1978 January 1980 1982 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 Barnes editor Robin chairman Kontron sponsor de Galan de Galan Barnes Barnes Barnes Mermet Barnes Mermet Barnes Wegscheider Barnes Broekaert Barnes Sanz-Medel Barnes Es=l\1. 0 Data base Unknown ',comple mlxture vrrl8ble 0 Hardware I Software 0 The ' old ' knowledge and experience of Fig. 22 Schematic representation of the situation that makes a posteriori line selection indispensable. Actually this is the situation to be covered by general survey analysis formerly entrusted to the d.c.arc in combination with photographic recording of the spectra. The advent of array spectrometers opens perspectives for a renaissance of the old arc methodology involving multi-line multi- element analysis arc system for general survey analysis used for many years in Philips Research Laboratories6* Which condition essentially dictates whether a priori or a posteriori line selection can or has to be used? Fig. 2 1 covers a series of situations that in principle permit a priori line selection. However all situations have as a common denominator the condition that the sample composition should be a priori 'known' the only difference between the situations being that the complexity of the sample increases from top to bottom making thus ever higher demands upon both the database and the flexibility of the spectrometer.The criterion is always the same namely the true detection limit or the conventional detection limit depending on whether a multivariate approach such as Kalman filtering is used or not. NOordwljk 197U - 1978 Purrlo Rko 1980 0 Orlando 1002 ff h n Ol.00 1984 ff Lapin 1985 Hiwall 1986 Lym W? &n Mwo lOl8 * Routla 1969 St htrraburp 1)90 Oomnund loo1 * &n Dbgo 1992 Graruda 1903 'LASMA SPElXROCHEMISTRY CONFERENCES # ,rd B I 6 P e c 4 i 7 Fig. 23 Erecting a triumphal arch for the 'good old arc' might still seem premature. Erecting it in honour of the Plasma Spectroche- mistry Conferences their Venues and Organizers (see attic at other side) is entirely appropriate along with a special tribute to Ramon M.Barnes and his ICP Information Newsletter which lent a charm to our 'Life with for and through the ICP' and made it more comfortable In contrast as Fig. 22 shows an 'unknown' complex mixture with variable composition precludes a priori line selection making a posteriori line selection mandatory. In fact a system using a posteriori line selection should be analysis proceed concomitantly and the consistency 01 the concentration values found with different lines of each analyte is used as the main criterion. This was the way in which the d.c. arc system referred to above was operated. It appears that the advent of powerful spectrometers with array detectors and perhaps also other types of multichan- nel instruments will produce a revitalization of 'spectrogra- phic general survey analysis' in the sense of casting the old arc methodology into the mould of modern technologies,JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL.8 779 fully profiting from the abundant possibilities of modern hardware and software. The greatest challenge will then be to achieve a balanced integration of old practical experi- ence knowledge available in the arsenal labelled ‘spectros- copic methodology’ and ‘abstract’ chemometrics. One might thus predict that the year 2000 will have seen the comeback of the arc in the jacket of an arch of the type shown in Fig. 23. Plasma Spectrochemistry Conferences To prevent premature celebrations the arch in Fig. 23 has been cautiously erected not for the arc but in honour of both the ICP and the Plasma Spectrochemistry Confer- ences whose venues are carved in the attic of the arch like battlefields in ancient times.The side of the arch shows the Imperator’s name Barnes along with his attribute the ICP Information Newsletter the sponsor and promoter of many conferences. The names of the Chairmen of the Confer- ences carved at the back of the arch are listed in Table 12 under three sub-headings characterizing the eras. Although it is not a Spectrochemistry Conference I found it appropriate to open the listing by making reference to the foundation of the ICP Information New~letter,~~ shortly before a Round Table Discussion in Grenoble acted as the germ of the Plasma Conferences. This discussion immediately following a cocktail party has been one of the best open discussions in the history of the ICP.The report in the Newsletter is still topical.70 The Stone Age was followed by the turbulent Golden Age also marked by open discussions unspoilt by commerce and trade fully recorded transcribed and afterwards published in edited form in the New~letter.~~~ In that period in particular the Newsletter disseminated a wealth of information including ‘Questions’ from readers and ‘Answers’ from pioneers two of which are listed in Table 1 3 for both instruction and entertainment! The last stage the Silicon Age was less turbulent than the previous era yet most interesting and challenging. Table 12 lists a continuous stream of streamlined Winter Confer- ences with sun or snow and skis or snorkels. These conferences have contributed greatly to communication among spectroscopists primarily Europeans and North Americans.The latest conference in Granada might prelude an extension of the communication lines in such a way as to embrace also Latin America in this network. Intensifying our contracts with Latin America is important not only to disseminate spectroscopic knowledge and promote the proliferation of modern spectroscopic methods and ideas in that part of the world but also to bring together people whose cultures have been continuously connected in the past five centuries. It will be a challenge to discover these connections personally on the track of what the Mexican Table 13 Samples from ‘Questions’ including the ‘Kinnunen Questions’ (1 975)71*72 Class of spectral lines best used with ICP arc or spark? Any special Answer- criteria for selecting lines? Arc lines No special criteria except perhaps interference by matrix lines How much wet chemistry has been replaced by ICP? Answer- Too early to tell how much has been replaced Predictable that the ICP will replace wet chemical analysis (including AAS) where a multi-element technique is more economical than single-element methods writer Carlos Fuentes has pictured in his fascinating book ‘The Buried Mirror’.73 Let that mirror be dug up in search of innovation or confirmation! Conclusions If a spontaneous discussion on the ICP at the XVIIIth CSI in Grenoble in 1975 is marked as the start of the ‘Plasma Spectrochemistry Conferences’ this institution will cele- brate this year its 18th birthday the age at which in many parts of the world youths reach maturity.Plasma spectro- chemistry itself is much older being cradled in the early 1960s. In spite of many explosive developments it is rejoicing and encouraging to see that the field still shows a remarkable vitality which prompts innovations such as furnace atomization plasma emission spectrometry and practically usable Cchelle array spectrometers. The latter development opens the door for an effective implementa- tion of many insights into the spectroscopic methodology that have been obtained in the past decade. This in turn may entail the reincarnation of the multi-line multi- element methodology associated with the now defunct arc. Also the present state of the art of array detectors solicits the exploitation of correlations in the spectra.A judicious application of these correlations could provide a lever for nullifying the adverse effects of source flicker noise on the precision of optical spectroscopic methods. If in this way the effect of flicker noise could be banished then the precision of optical methods might become competitive with that of X-ray fluorescence spectrometry. On the other hand plasma spectrochemists should be well aware of developments in the X-ray field such as the maturing of total-reflection X-ray fluorescence spectrometry which is capable of handling samples as solutions and slurries although often requiring chemical separation or preconcen- tration. However it appears that also in plasma spectroche- mistry ever more chemistry is involved but chemistry with a better status than formerly having gained appreciably in social prestige and standing from its happy liaison with the plasmas and the spectra of plasma spectrochemistry.1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 References Plasma Spectrochemistry guest ed. Barnes R. M. Spectro- chim. Acta Part B 1983 38 pp. 1-445. Developments in Atomic Plasma Spectrochemical Analysis ed. Barnes R. M. Heyden London 1981. Boumans P. W. J. M. in Inductively Coupled Plasma Emission Spectroscopy Part 1 Methodology Instrumentation and Per- formance ed. Boumans P. W. J. M. Wiley New York 1987 p. 93. ICP In$ Newsl. 1976 2 67 and 101. ICP In$ Newsl. 1978,4 89 147 and 199. Chernoff H. J. Am. Stat. Ass. 1973 68 361. Schubert A. and Braun T. Scientometrics 1992 23 3.Baxter D. C. Nichol R. Littlejohn D. Ludke C. Skole J. and Hoffmann E. J. Anal. At. Spectrom. 1992 7 727. Liang D. C. and Blades M . W. Spectrochim. Acta Part B 1989,44 1059. Sturgeon R. E. Willie S. N. Luong V. T. Berman S. S. and Dunn J. G. J. Anal. At. Spectrom. 1989 4 669. Hettipathirana T. D. and Blades M. W. Spectrochim. Acta Part B 1992 47 493. Sturgeon R. E. Luong V. T. Willie S. N. and Marcus R. K. Spectrochim. Acta Part B 1993 48 893. Liang D. C. and Blades M. W. Spectrochim. Acta Part B 1989 44 1059. The GF-CCP 100 Catalogue Aurora Instruments Vancouver B.C. Canada 1992. Application Data Aurora Instruments Vancouver B.C. Canada 1992. Winge R. K. Peterson V. G. and Fassel V. A. Appl. Spectrosc.. 1979 33 206.7 80 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1993 VOL. 8 17 Klockenkamper R.and von Bohlen A. J. Anal. At. Spec- 44 Boumans P. W. J. M. Spectrochim. Acta Part B 1991 46 trom. 1992,-7 273.- 18 Prange A. and Schwenke H. Adv. X-ray Anal. 1992 35,899. 19 Klockenkamper R. and Tolg G. Spectrochim. Acta Part B 1993,48 111. 20 Total-Reflection X-Ray Fluorescence Spectrometry guest ed. Klockenkamper R. and ed. Boumans P. W. J. M. Spectro- chim. Acta Part B 1989,44 pp. 433-549. 2 1 Total-Reflection X-Ray Fluorescence Spectrometry guest ed. Wobrauschek P. and ed. Boumans P. W. J. M. Spectrochim. Acta Part B 1991 46 pp. 1313-1436. 22 Total-Reflection X-Ray Fluorescence Spectrometry guest ed. Prange A. and ed. Boumans P. W. J. M. Spectrochim. Acta Part €3 1993 48 pp. 107-300. 23 Scribner B. F. and Mullin H.R. J. Res. Natl. Bur. Std. 1946 37 379. 24 Boumans P. W. J. M. in Inductively Coupled Plasma Emission Spectroscopy Part I Methodology Instrumentation and Per- formance ed. Bourmans P. W. J. M. Wiley New York 1987 p. 25. 25 Margoshes M. Spectrochim. Acta Part B 1970 25 2 13. 26 Bilhorn R. B. Sweedler J. V. Epperson P. M. and Denton M. B. Appl. Spectrosc. 1987 41 1 114. 27 Bilhorn R. B. Epperson P. M. Sweedler J. V. and Denton M. B. Appl. Spectrosc. 1987 41 1125. 28 Sweedler J. V. Bilhorn R. B. Epperson P. M. Sims G. R. and Denton M. B. Anal. Chem. 1988,60 282A. 29 Epperson P. M. Sweedler J. V. Bilhorn R. B. Sims G. R. and Denton M. B. Anal. Chem. 1988,60 327A. 30 Bilhorn R. B. and Denton M. B. Appl. Spectrosc. 1989,43 1. 3 1 Sweedler J. V. Jalkian R. D. Pomeroy R.S. and Denton M. B. Spectrochim. Acta Part B 1989 44 683. 32 Fields R. E. Baker M. E. Radspinner D. A. Pomeroy R. S. and Denton M. B. Spectroscopy 1992 7(9) 28. 33 Pilon M. J. Denton M. B. Schleicher R. G. Moran P. M. and Smith S. B. Jr. Appl. Spectrosc. 1990 44 1613. 34 Kolczynsky J. D. Radspinner D. A. Pomeroy R. S. Baker M. E. Noms J. A. Denton M. B. Foster R. W. Schleicher R. G. Moran P. M. and Pilon M. J. Am. Lab. 1991 May. 35 Barnard T. W. Crockett M. I. Ivaldi J. C. and Lundberg P. L. Anal. Chem. 1993 65 1225. 36 Barnard T. W. Crockett M. I. Ivaldi J. C. Lundberg P. L. Yates D. A. Levine P. A. and Sauer D. J. Anal. Chem. 1993,65 1231. 37 Boumans P. W. J. M. Spectrochim. Acta Part B 1991 46 917. 38 Boumans P. W. J. M. Spectrochim. Acta Part B 1990 45 799.39 Boumans P. W. J. M. Spectrochim. Acta Part B 1991 46 431. 40 Boumans P. W. J. M. Ivaldi J. C. and Slavin W. Spectrochim. Acta Part B 199 1 46 64 1. 41 Kaiser H. Spectrochim. Acta 1947 3 40. 42 Reprint of ref. 41 in Future Trends in Spectroscopy Proceed- ings of the Symposium held at the Pontifical Academy of Sciences The Vatican 27-28 June 1989 to mark the 50th Anniversary of Spectrochimica Acta 1939- 1989 Special Sup- plement to the 1989 volumes of Spectrochimica Acta 1989 p. 177. 43 Borer M. W. Sesi N. N. Starn T. K. and Hieftje G.M. Spectrochim. Acta Electronica included in Spectrochim. Acta Part B 1992 47 El 135. 693. 45 Ivaldi J. C. and Barnard W. Spectrochim. Acta Part B 1993 in the press. 46 Boumans P. W. J. M. McKenna R. J. and Bosveld M. Spectrochim.Acta Part B 1981 36 1031. 47 Boumans P. W. J. M. Fresenius’ 2. Anal Chem. 1979 299 337. 48 Boumans P. W. J. M. in Inductively Coupled Plasma Emission Spectroscopy Part I Methodology Instrumentation and Per- formance ed. Boumans P. W. J. M. Wiley New York 1987 p. 165. 49 Boumans P. W. J. M. and Vrakking J. J. A. M. Spectrochim. Acta Part B 1985 40 1085. 50 Boumans P. W. J. M. and Vrakking J. J. A. M. Spectrochim. Acta Part B 1985 40 1107. 51 Boumans P. W. J. M. Fresenius’ Z. Anal. Chem. 1986 324 397. 52 Boumans P. W. J. M. and Vrakking J. J. A. M. Spectrochim. Acta Part B 1987 42 8 19. 53 Boumans P. W. J. M. and Vrakking J. J. A. M. J. Anal. At. Spectrom. 1987 2 513. 54 Boumans P. W. J. M. Spectrochim. Acta Part B 1989 44 1325. 55 Boumans P. W. J. M. and Vrakking J. J. A. M. Spectrochim. Acta Part B 1984 39 1261. 56 Boumans P. W. J. M. Tielrooy J. A. and Maessen F. J. M. J. Spectrochim. Acta Part B 1988 43 173. 57 Boumans P. W. J. M. and van Ham-Heijms A. H. M. Spectrochim. Acta Electronica included in Spectrochim. Acta Part B 1991 46 E1863. 58 van Veen E. H. and de Loos-Vollebregt M. T. C. Spectro- chim. Acta Part B 1990 45 313. 59 van Veen E. H. Oukes F. J. and de Loos-Vollebregt M. T. C. Spectrochim. Acta Part B 1990 45 1109. 60 van Veen E. H. and de Loos-Vollebregt M. T. C. Anal. Chem. 1991 63 1441. 61 van Veen E. H. Spectroscopy 1992 7(6) 43. 62 Boumans P. W. J. M. Spectrochim. Acta Part B 1990 45 1121. 63 Yang J.-f. Piao Z.-x. and Zeng X.-j. Spectrochim. Acta Part B 1991 46 953. 64 Yang J.-f. Piao Z.-x. Zeng X.-j. Zhang Z.-y. and Chen X.- h. Spectrochim. Acta Part B 1992 47 1055. 65 Yang J.-f. Piao Z.-x. and Zeng X.-j. Spectrochim. Acta Part B 1993 48 359. 66 Yang J.-f. Piao Z.-x. and Zeng X.-j. Spectrochim. Acta Part B 1993 48 543. 67 Ivaldi J. C. Tracy D. Barnard T. W. and Slavin W. Spectrochim. Acta Part B 1992 47 136 1. 68 Witmer A. W. Jansen J. A. J. van Gool G. H. and Brouwer G. Philips Tech. Rev. 1974 34 322. 69 ICP InJ Newsl. ed. Barnes R. M. 1975 1 1. 70 Scott R. H. ICP Inf Newsl. 1975 1 144. 71 ICP InJ Newsl. 1975 1 46. 72 ICP In$ Newsl. 1976 1 206. 73 Fuentes C. The Buried Mirror Reflections on Spain and the New World Houghton Mifflin 1992. Paper 3/01321I Received March 5 I993 Accepted April 3 1993