Ž .Spectrochimica Acta Part B 54 1999 2081]2094

Atomic absorption spectrometry } pregnant againafter 45 years

Bernhard WelzU

Departamento de Quımica, Uni ersidade Federal de Santa Catarina, 88040-900 Florianopolis, SC, Brazil´ ´

Received 1 June 1999; accepted 4 October 1999

Abstract

Ž .Because atomic absorption spectrometry AAS seems to be so simple at first glance, its forthcoming end andreplacement by more exciting techniques has been forecasted more than once over the past 45 years. However, AAShas received strong impetus again and again, e.g. by the introduction of the graphite furnace technique, and of flowinjection, to mention but a few. Although more and more researchers, and even more instrument manufacturers areturning their back on AAS these days, this author believes that AAS is about to give birth to new offspring in thevery near future. The most important ones are solid sampling and speciation analysis on the application side, a muchdeeper exploitation of the potential of flow injection analysis, the use of diode lasers as radiation sources, and theintroduction of continuum-source AAS on the instrumental side. The latter could replace conventional line-sourceAAS in the foreseeable future because of its obvious advantages in essentially all analytical aspects. Q 1999 ElsevierScience B.V. All rights reserved.

Keywords: Continuum source AAS; Diode laser AAS; Flow injection; Solid sampling; Speciation analysis

1. Introduction

Ž .Atomic absorption spectrometry AAS is not atechnique that one gets excited about at firstglance. One has to come a little bit closer andlook twice in order to find out about its charm.Maybe this is the reason why, after the first

U Ž .E-mail address: welz@qmc.ufsc.br B. Welz

complete description of the processes of absorp-tion and emission of radiation by atoms in flames

w xby Kirchhoff and Bunsen 1]3 in the 1860s, itwas optical emission that caught the interest ofspectroscopists, and not atomic absorption. And ithas taken almost a century until Alan Walshbegan to wonder why molecular spectra wereusually obtained in absorption and atomic spectra

w xin emission 4 . And he came to the conclusionthat there was no good reason for neglecting

0584-8547r99r$ - see front matter Q 1999 Elsevier Science B.V. All rights reserved.Ž .PII: S 0 5 8 4 - 8 5 4 7 9 9 0 0 1 5 4 - 8

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]20942082

atomic absorption spectra; on the contrary, theyappeared to offer many vital advantages overatomic emission spectra, and so he started tocarry out his first experiments. Even though Alke-

w xmade and Milatz 5,6 published two papers aboutAAS in the same year when Walsh’s first paperw x7 appeared, it was Alan Walsh who had thevision of a bright future for this technique, andwho was supporting and ‘preaching’ AAS with thededication of a missionary, until it finally foundacceptance in the mid-1960s.

However, even with the support of a visionaryscientist such as Alan Walsh, AAS is not neces-sarily exciting at first glance. When Alan, back in1952, called his colleague John Willis, and showedhim his first experiments with the words ‘look,that’s atomic absorption...’ the disappointing

w xresponse was only ‘so what?’ 4 . This lack ofinterest continued when an atomic absorptionspectrometer was publicly demonstrated for thefirst time in Melbourne in 1954. The only personwho got excited about AAS at first glance in theseearly years was Boris L’vov, who decided to checkthe validity of the author’s ideas immediatelyafter he had stumbled over Walsh’s first publica-

w xtion in 1956 8 , which qualifies him as anothervisionary scientist who could recognize the impor-tance of a discovery that was disregarded byalmost everyone else. And L’vov should becomeanother dedicated missionary of AAS, after hehad succeeded to slowly escape the ‘splendidisolation’ of the socialist system of his country.

But similar to the experience that Alan Walshhad to make, the excitement of Boris L’vov wasnot shared by his colleagues, and it is obvious thatAlan and Boris were the exception, whereas thisauthor was more the rule, if I may insert some ofmy personal history. I remember very well my firstcontact with AAS. It was when I applied for aposition as an application specialist for infraredspectroscopy at Bodenseewerk Perkin Elmer in1966, and they offered me a position as applica-tion specialist for AAS instead. ‘Atomic what...?’was my first question, and when I was digging intoAlan Walsh’s paper soon after, there was a com-plete lack of excitement on my part. You aspiratea solution into a flame and you get a resultimmediately } something was missing; there is

no spectrum that needs interpretation as inmolecular spectroscopy. This technique was toosimple for me to be exciting, and I really don’tknow why I nevertheless accepted the position asan application specialist for a technique that I didnot know anything about at that time. However,the excitement came soon, I only had to meetwith the two missionaries of AAS, Alan Walshand Boris L’vov, the latter at the First AtomicAbsorption Spectrometry Symposium in Prague,1967. It was at that conference when the spark ofexcitement ignited a flame that would continue toburn throughout my career, and that is reflectedin more than 250 publications that carry my name,including a book on AAS that just appeared in its

w xthird edition 9 .It may well be the apparent simplicity of AAS

Žthat made me wonder back in 1966 if there is.anything of interest in this technique that caused

renowned scientists again and again to forecastthe end of AAS for the very near future. Theremust have been a good reason for Alan Walsh towrite his article Atomic Absorption Spectroscopy

w x} Stagnant or Pregnant 10 back in 1974, i.e.there must have been rumors in the air about aforthcoming decay of AAS. But obviously, by thattime AAS had already given birth to a new kid,

Ž .the graphite furnace GF which, however, wasnot yet well understood, and it required the input

w xof L’vov 11 , and the introduction of the StabilizedŽ .Temperature Platform Furnace STPF concept

w xby Slavin et al. 12 before it could become thedriving force for AAS in the 1980s.

But at a time when AAS was in use in everylaboratory, and research groups all around theworld were unveiling atomization and interfer-ence mechanisms of GF AAS, and the increasingknowledge made this technique more and morerugged, there was again a renowned scientist pre-

w xdicting the forthcoming end of AAS. Hieftje 13applied a third order polynomial to the annualnumber of publications on AAS, and predictedfrom the extrapolation of that function that ‘withthe current rapid growth in ICP atomic emissionand ICP mass spectrometry and with new incur-sions being made by methods based on glow-dis-charge lamps, AAS is heading for difficult times.If current trends continue, I would not be sur-

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]2094 2083

prised to see the removal of commercial AASinstruments from the marketplace by the year2000.’ By that time, however, AAS was alreadypregnant again, and the name of this new off-

Ž .spring was flow injection FI , a technique theinput of which can be considered revolutionary in

w xalmost all aspects of AAS analysis 14 . It hasbeen shown repeatedly that FI is far more thanan elegant sample handling and sample introduc-tion technique, and the potential of FI has not atall been exploited completely, as will be discussedlater in this paper.

Obviously only a few milestones in the develop-ment have been discussed in this review of thelast 45 years of AAS } if 1954, when the firstatomic absorption spectrometer was exhibited in

w xMelbourne 10 , is taken as the starting point.Zeeman-effect background correction, simultane-ous multi-element AAS, and the introduction ofsolid-state detectors should at least be mentionedas other important contributions. But now, at theend of this millennium, isn’t it time to look fornew goals, new challenges, and send AAS toretirement in the same way as this author wassent to retirement, because AAS is an ‘estab-lished technique’ that does not justify any moreresearch. Obviously this author has a differentopinion and is expecting a whole series of excitingnew developments, both in the field of applicationand in instrumentation. AAS is clearly pregnantagain.

2. New fields of application

When new fields of application are discussed, itis necessary also to consider alternate techniquesthat might be capable of solving the same prob-lem equally well or even better. It must also bekept in mind that there is no single analyticaltechnique that can solve all problems, and thateach technique performs best only in its optimumworking range, and that the results may deteri-orate quickly when it is used outside this range.The practical analyst looks for a solution to hisanalytical problems and, if he can choose, heselects the analytical technique that offers the

best solution with respect to simplicity, time andcost. Obviously, analytical aspects such as accu-racy, and frequently also detection power, play anequally important role.

Inductively coupled plasma mass spectrometryŽ .ICP-MS is undoubtedly the technique that ismost en ¨ogue today for trace element analysis,and there is no doubt about its detection power,

Ž .its multi-element -mass capability, and its speedof analysis. The possibility of doing isotope analy-sis, in addition, offers a unique field of applica-tion that is proprietary to MS. On the other hand,ICP-MS is undoubtedly one of the relativelyexpensive techniques, both in purchase price andin running cost. ICP-MS also requires consider-able operator’s skill, and is certainly not free ofinterferences, particularly in the presence of com-plex and concentrated matrices. Because of itspopularity and competitiveness, ICP-MS will bethe prominent technique for comparison with thenew fields that AAS may enter soon.

2.1. FI on-line preconcentration and separation forGF AAS

It has been mentioned earlier that the poten-tial of FI has not been fully exploited for itsapplication in AAS. One of these fields isundoubtedly FI on-line preconcentration and sep-aration for GF AAS, using sorbent extraction inpacked microcolumns andror precipitation and

Ž .collection in knotted reactors KR . Although thefirst papers using these techniques already ap-

w xpeared in the early 1990s 15,16 , this idea wasnever really supported by instrument manufactur-ers. This is surprising because the combination ofFI and GF AAS offers ideal conditions for fullyautomatic ultra-trace analysis in the low ngrlrange in complex matrices under normal routinelaboratory conditions. Detection limits of GF AAS

w xin samples such as seawater 17,18 or high-purityw xreagents 19 are lowered by 2]3 orders of magni-

tude with a relatively minor instrumental require-ment, only an insignificant reduction in samplethroughput, compared to direct GF AAS analysis,and no need for clean-room facilities.

An example of the capability of this technique

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]20942084

Table 1Determination of four elements in doubly deionized waterŽ .DDW and in NASS 2 open ocean sea water standard

Ž .reference material National Research Council Canada usingflow-injection on-line sorbent extraction separation and pre-

aconcentration for GF AAS

Element Sample

DDW NASS 2 NASS 2found found certified

Cd 24"1 28"1 29"4Pb 18"2 41"2 39"6Cu 34"5 101"4 106"11Ni 73"12 282"14 257"27

a ŽAll values in ngrl; S.D. of ns10 data taken fromw x.Sperling et al. 17 .

is given in Table 1, which shows the determina-tion of four trace elements in a sea water refer-ence material by GF AAS after on-line sorbent-extraction preconcentration and separation. Theresults demonstrate first of all that the techniqueis capable of obtaining accurate results in thengrl range in a complex and concentrated matrix

such as sea water. Secondly it is shown that allcommonly occurring elements can be determined

Ž .in doubly deionized water DDW , making anon-line purification of reagents a necessity.Thirdly, by comparing the standard deviation ob-tained in the sea water sample and in DDW itbecomes apparent that it is no longer the matrixthat determines the precision, but obviously the

Žlamp flicker noise nickel is well known to be a.noisy lamp . This means the matrix has been

separated completely and has no more influenceon the determination.

FI on-line preconcentration and separationclearly brings GF AAS detection limits close tothose of ICP-MS, and it even surpasses the capa-bilities of the latter technique when the matricesare considered in which the detection limits areobtained. Obviously, FI on-line preconcentrationand separation could also be coupled to ICP-MS,but not with the same ease, considering the elu-

Ž .ents that are typically used organic solvents andtheir volume which is usually -0.1 ml. Last, butnot least, the running cost for the GF AAS

Ž .Fig. 1. Determination of thallium in rock samples by direct solids analysis. a Atomization signals are dependent on the tubeŽ .temperature; 1 } 11008C, sample sinters; 2 } 14008C, sample melts; 3 } 17008C, sample begins to volatilize. b Correlation of

Ž . Ž w x.the AAS signals with neutron activation analysis NAA from Welz 9 .

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]2094 2085

approach is probably an order of magnitude lowerthan that for the comparable ICP-MS system.

2.2. Solid sampling for GF AAS

Although Alan Walsh in his first approach pro-posed ‘that the sample is dissolved and then

w xvaporized in a Lundegardh flame’ 5 , which forgood reasons became the preferred technique inthe 1960s and 1970s, solid sampling is as old as

w xAAS 20 . Boris L’vov used a graphite furnace anda few crystals of sodium chloride in his firstexperiment back in 1957 to demonstrate the prin-ciple of AAS, and several other groups developeda variety of furnaces for solid sample analysis inthe following years. Fig. 1 shows an early applica-tion of solid sample analysis that was carried outin our laboratory using a prototype of what waslater called the HGA-70, and that was presentedat the International Atomic Absorption Spectros-copy Conference in Sheffield, 1969. Although theconditions were very primitive at that time, withno reliable tools for solid sample introductioninto the furnace and no means for backgroundcorrection, the results were surprisingly good.

The direct analysis of solid samples, using allkinds of furnaces and devices for sample intro-

duction was continued throughout the past 30years, as reviewed in a recent book, edited by

w xKurfurst 21 , but it was typically supported by¨isolated research groups only. This is surprisingbecause GF AAS is an ideal technique for directsolid sample analysis, because it is very flexiblewith respect to the sample size, which can rangefrom approximately 0.01 mg up to almost 100 mgw x22 , and also with respect to the form in whichthe sample is presented. This is among otherthings due to the way the sample is introduced,e.g. on a platform, the thermal pretreatment inthe graphite furnace, and the long residence timeof the atoms in the atomizer. The introduction ofa commercial system for automatic slurry sam-

w xpling 23 , based on the extensive work of Miller-w xIhli 24 , could have been the turning point, but

the acceptance was not as expected.More recently, however, there appears to be an

increasing market requirement for solid sampleanalysis, coming predominantly from the produc-ers of modern high-tech materials, such as hardmetals, superalloys and ceramic superconductors,etc. All these materials have two things in com-mon: their quality depends extremely on theirpurity with respect to a number of critical traceelements, and they are very difficult to bring into

Table 2Ž . Ž .Limits of detection in ngrg based on 3 s achievable in analysis of tungsten blue oxide by solid sampling ET AAS SS-ETAAS in

Ž . Ž .comparison with those of solution ET AAS sol-ETAAS , slurry-ETAAS, ion chromatography IC , coupled to ICP OES, and glowaŽ .discharge mass spectrometry GDMS

Element SS- sol- slurry- sol- sol- IC- GDMSb c dETAAS ETAAS ETAAS ETAAS ETAAS ICP OES

Ca 0.05 10 1 20 10 22 100Co 0.2 10 20 300 100 78 3Cr 0.4 80 20 120 70 28 100Cu 0.09 10 210 230 200 11 10Fe 1.7 200 4 80 ] 20 200K 0.3 10 1 30 20 260 20Mg 0.02 5 1 10 10 2.5 50Mn 0.07 10 ] ] ] 3.8 5Na 0.01 15 1 60 30 55 50Ni 0.4 20 10 120 130 50 10Zn 0.01 20 ] ] ] 17 40

aŽ w x.from Hornung and Krivan 27 .b Digestion procedure: HFrHNO .3c Digestion procedure: H O .2 2d Digestion procedure: NH .3

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]20942086

solution. It was particularly the group of ViliamKrivan in Germany who did excellent pioneeringwork in that field, demonstrating that detectionlimits of solid sampling GF AAS are superior toall other techniques available for this kind ofanalysis, including ICP-MS, simply because therisk of contamination associated with any acid

w xdigestion technique, is avoided 22,25]27 . A typi-cal example is shown in Table 2 for the detectionlimits achievable for a number of elements in theanalysis of powdered tungsten trioxide and tung-sten blue oxide, using a variety of analytical tech-

w xniques 27 .Even more important for the practical analysis

is that the efforts of Krivan and his group haveresulted in the recent introduction of commercialequipment that automates solid sampling GF AASand makes it accessible for routine applicationŽ .Fig. 2 . It has been shown that modern furnacetechnology using platform atomization in a trans-versely heated graphite tube, and integratedabsorbance for signal evaluation, has increasedaccuracy dramatically, making solid sample analy-sis possible with calibration against simple aque-

w xous standards 22,25]27 .

2.3. Speciation analysis

In the 1960s, determination of summationparameters such as total heavy metals in water bysulfide precipitation and gravimetry, was replacedby the determination of the individual elements,as it was recognized that the various heavy metalsdiffer significantly in their toxicity. One of thedriving forces at that time was the introduction ofAAS which made these determinations easy. Inthe meantime it is well established that the totalcontent of an element does not give sufficientinformation, as several elements may be essentialfor humans or animals, and may be toxic at thesame time, depending not only on their concen-tration, but even more importantly on their oxida-tion state or the chemical form in which they arepresent. A typical example is chromium, which isessential in its trivalent form, but carcinogenic inits hexavalent compounds. Another example isarsenic, the inorganic forms of which are wellknown to be highly toxic, whereas organic com-

Fig. 2. Fully automatic solid sampling accessory SSA 61Ž .Analytik Jena, Germany for GF AAS with integrated mi-crobalance and automatic determination of sample weight.

pounds, such as arsenobetain, which are found inhigh concentration in some seafood, are non-toxic,and may even be essential.

A limiting factor for the introduction of specia-tion analysis into routine use is undoubtedly thatmost of the research work in that field is currentlydone using high-pressure liquid chromatographyŽ .HPLC as a separation technique, and ICP-MSfor detection. This combination, although verypowerful, is far too expensive for routine use,considering that a fast multi-element system istypically waiting 10]20 min for a few species of asingle element arriving at the detector. There isno doubt that this kind of research has to becarried out in order to gain all the knowledgenecessary for speciation analysis. However, forroutine purposes, a good percentage of speciationanalysis can be done without prior separation of

w xthe individual compounds by chromatography 28 .A simple procedure has, for example, been pro-posed to distinguish between ‘toxic’, i.e. inorganic,mono- and di-methylated arsenic, and non-toxic

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]2094 2087

Ž .arsenobetaine using hydride generation HGw xAAS 29 . FI on-line column preconcentration

and separation is typically selective for one oxida-tion state of an element only, and may hence beused for the separation of redox species, as wasdemonstrated for the differential determination

Ž . w xof Cr VI and total chromium in water 30 .Another example of speciation analysis by on-lineseparation and preconcentration is shown in Fig.3 for an activated alumina microcolumn, which at

Ž .pH 2 selectively preconcentrates Cr VI , whereasŽ . w xCr III is retained selectively at pH 7 31 . There

is no question that this field of routine speciationanalysis is far from being completely explored.

In addition, AAS with a quartz tube atomizer isan extremely attractive detector for gas chro-

Ž .matography GC , as already proposed back inw x1976 by Van Loon and Radziuk 32 . This door

was, however, closed by instrument manufactur-ers with the introduction of digital electronics inthe 1980s that no longer permitted the recordingof a series of peaks over an extended period oftime, i.e. the duration of a chromatogram. Itshould not be difficult to open this door again,and AAS could then become a sensitive, highlyspecific, and not too expensive detector for GCw x33 , and after post-column derivatisation, even

Fig. 3. Effect of the pH value of various buffer solutions onŽ . Ž .the preconcentration of Cr III and Cr VI on activated

alumina, expressed as the absorbance signal obtained afterŽ w x.elution from Sperling et al. 31 .

w xfor HPLC 34]36 . This aspect of AAS as anattractive detector for chromatography in specia-tion analysis will be brought up one more time in

Žconnection with diode laser AAS see Section.3.2 .

3. Instrumental developments

As in the previous chapter, this author does notintend to consult a crystal ball to see the distantfuture, but simply wants to analyze the ongoingresearch for its potential to become attractive forroutine application within the next few years.Obviously, as the instrumental developments, evenmore than the applications, depend on the deci-sion of instrument manufacturers to invest in aparticular field, it is not easy to predict if thosedevelopments that appear most promising to thisauthor, will really become commercial products.

3.1. FI gradient ratio calibration

Flow injection has been discussed several timesalready in this paper as a revolutionary techniquefor sample handling, sample introduction, and foron-line preconcentration and separation for AAS.However, there are many more application possi-bilities for this technique, particularly because ofthe very high repeatability of the transportproperties and the dispersion of the sample bolusin the FI system. It has been demonstrated thatthis high repeatability of all events over time canbe used successfully to correct for a number oflimitations of conventional AAS, including thelimited linear working range, and the influence of

w xconcomitants on the analyte signal in F AAS 37 .With this ‘gradient ratio calibration’ the entiretransient signal produced by the sample bolus isstored in a computer and used for calibration.The ratio of the absorbance of the calibrationsolution and the test sample solution is formed atthe reading frequency of the spectrometer. If nointerference is present, and the measurementsolutions are all in the linear range, the ratio willnot change during the entire measurement. If themaximum absorbance of a test sample solution isoutside the linear range, however, the absorbance

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]20942088

ratio changes as soon as the linear range isexceeded. This is recognized and corrected by thecomputer. The same situation occurs if an inter-ference is present which decreases with increas-

Ž .ing dilution dispersion , as shown in Fig. 4a forthe interference of phosphate on the determina-tion of calcium. If the computer program is usedto extrapolate the signal ratio over the entireabsorbance profile against absorbance As0, the

Žinterference can be eliminated by calculation Fig..4b .A major significance of this procedure is cer-

tainly that not only can non-spectral interferencesin many cases largely be eliminated by calcula-tion, but more importantly that it can recognizeinterferences. The possibility is thus given of issu-ing a warning signal that permits the analyst tothoroughly examine the trueness of the resultdetermined by the computer. This opens com-pletely new perspectives for quality assurance withrespect to trueness of analytical results. Tech-niques like this one, which essentially only

require software development, and which belongalready in the area of expert systems, shouldbecome more and more part of future instru-ments, as computer capacity is no longer a limit-ing factor.

3.2. Diode laser AAS

Tunable dye lasers have been proposed as radi-ation sources for AAS already some 20 years agow x38 . However, in spite of their extremely goodspectroscopic properties } they can be set tovirtually any atomic line between 213 nm and 900nm with a bandwidth corresponding to the natu-ral width of an atomic line } they have notfound their way into common use because of anumber of practical and economic reasons: theselaser systems are expensive, frequently unreliable,and difficult to operate. In contrast to dye lasers,

Ž .diode lasers DL appear to be more suitable toŽ .one day replace hollow cathode lamps HCL and

Ž .electrodeless discharge lamps EDL . Semicon-

Ž . Ž .Fig. 4. a Transient FI signals of measurement solutions containing 16 mgrl Ca calibration solution with and without a3y Ž . Ž .concentration of 0.01 molrl PO test sample solution ; b signal ratio of the calibration solution to the test sample solution as a4

Ž w x.function of the absorbance of the test sample solution from Sperling et al. 37 .

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]2094 2089

ductor laser diodes are nowadays mass producedfor compact disc players, laser printers, opticaldata storage systems, and telecommunicationequipment, and hence they are cheap, reliable,easy to operate and they have long lifetimes. Anumber of these diodes have excellent spectros-copic properties which make them attractive

w xsources for spectrochemical analysis 39 .Firstly, the power of presently available

commercial DLs is between one and several or-ders of magnitude higher than what is providedby the best HCLs. In addition, DLs show excep-tional stability, both in terms of wavelength andintensity. These two factors together make it pos-sible to measure extremely low absorbances ifoptimal experimental conditions are realized andthe fundamental shot-noise limit is achieved. Buteven in conventional ‘routine’ atomizers, such asflames and furnaces, detection limits wereachieved that were 1]2 orders of magnitude lower

w xthan those obtained with HCLs 39,40 .Secondly, the typical linewidth of a commercial

DL is approximately two orders of magnitude lessthan the width of absorption lines in flames andfurnaces. This makes possible the expansion ofthe dynamic range of calibration to high analyteconcentrations by measuring the absorption onthe wings of the absorption line, where optically

w xthin conditions prevail 40 . In addition, a DL,under normal operating conditions, emits one sin-gle narrow line, which dramatically simplifies thespectral isolation of the absorption signal. Onedoes not need the monochromator which is nec-essary with HCLs for isolation of the analyticalline from the other spectral lines emitted by theHCL, and the photomultiplier could be replaced

w xby a simple semiconductor photodiode 39,41 .Thirdly, the wavelength of DLs can be easily

modulated at frequencies up to GHz by modula-tion of the diode current. Wavelength modulationof the DL with detection of the absorption at thesecond harmonic of the modulation frequency, 2f ,

Ž .greatly reduces low-frequency flicker noise inthe baseline, providing improved detection limitsw x39,40 . In addition, wavelength modulation of theradiation source provides an ideal correction ofnon-specific absorption and significantly improvesthe selectivity of the AAS technique.

The major limitation of DL AAS at this pointin time is that, although a commercial blue laserdiode was introduced earlier this year, the lowerwavelength limit for mass-produced DLs is ap-proximately 630 nm, which means that even withfrequency doubling in non-linear crystals, the im-portant wavelength range for AAS of 190]315 nmcannot be attained yet with this technique. How-ever, this need not necessarily be a major limita-tion for the successful introduction of DL AASinto routine application, as this technique, in theopinion of this author, is ideally suited for dedi-cated instruments for special purposes and aslow-cost detectors for GC or HPLC. One examplefor such an application is the HPLC]DL AAS

Ž . Ž .system for speciation analysis of Cr III rCr VIw xproposed by Groll et al. 42 ; another example is

the tungsten coil atomizer DL AAS system forthe determination of aluminum and chromium

w xdescribed by Krivan et al. 41 , which is shown inFig. 5. Another very interesting aspect is that withDLs as radiation sources, the determination ofnon-metals such as halogens, sulfur, or even noblegases comes within reach of AAS. All these ele-ments have long-lived excited states from whichstrong absorption transitions can be induced by

w xthe red and near-IR radiation of LDs 39 . Zybinw xet al. 43 , for example reported about a determi-

nation of chlorine by GC-microwave inducedplasma-DL AAS with a detection limit some twoorders of magnitude lower than the best valuesobtained by optical emission spectrometry.

In the opinion of this author, DL AAS shouldat this point in time not be considered a competi-tor of the conventional AAS with HCLs andEDLs, but as a very attractive expansion of thecapabilities of AAS that opens entirely new fieldsof application at relatively low cost. The first-generation instruments and modules for DL AAS

w xthat became commercially available recently 44offer an excellent opportunity to exploit the po-tential of this technique not only in research butalready in routine application.

3.3. Continuum source AAS

In his early work of 1952, Walsh essentiallyŽ .ruled out the use of a continuum source CS for

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]20942090

Ž . Ž .Fig. 5. Schematic diagram of a diode laser atomic absorption spectrometer with a tungsten coil atomizer: 1 laser module; 2Ž . Ž . Ž .Peltier-element operated heat sink; 3 laser diode; 4 collimating lens system; 5 optical non-linear media for second harmonic

Ž . Ž . Ž . Ž . Ž . Ž .generation; 6 filter; 7 quartz absorption cell; 8 gas inlet for Ar]H ; 9 ceramic stopper; 10 copper electrodes; 11 tungsten2Ž . Ž . Ž . Ž . Ž . Ž . Ž . Ž .coil; 12 gas outlet; 13 quartz stopper; 14 diaphragm; 15 filter; 16 photomultiplier; 17 lock-in amplifier 2f detection ; 18

Ž . Ž . Ž . Ž . Ždigital storage oscilloscope; 19 computer; 20 power supply; 21 modulation voltage; and 22 laser module controller fromw x.Krivan et al. 41 .

AAS because the required resolution of approxi-mately 2 pm was far beyond the capabilities ofthe best spectrometer available in his laboratoryat the time. He concluded that ‘one of the maindifficulties is due to the fact that the relationshipsbetween absorption and concentration depend onthe resolution of the spectrograph...’ if a regular

w xmonochromator is used 10 . Hence, line sourcesŽ .LS , such as HCLs and EDLs were used almostexclusively for routine application of AAS overthe past 45 years, as their stable, narrow-lineemission guarantees high analyte specificity, evenwith the use of low-resolution monochromators,as well as good detection limits, and a workingrange of 2]3 orders of magnitude. The fact, how-ever, that the use of LSs in essence makes AAS a

Žone-element-at-a-time technique although the si-multaneous determination of more than one ele-

.ment has in the meantime been realized has

stimulated researchers again and again to investi-w xgate the feasibility of CS AAS 45 .

Only recently, however, have these effortsshown convincing results with the availability of

w xhigh-resolution echelle spectrometers 46 , solid-´w xstate array detectors 47,48 , and continuum

sources that have a sufficiently high emissionintensity within the small spectral interval underconsideration over the entire wavelength range ofAAS. Fig. 6 depicts such a system with a Xenonshort-arc lamp, a double-echelle monochromator´Ž .DEMON , and a CCD detector, providing aspectral resolution of approximately 2 pmrpixelw x49 . The resolving power of this setup is impres-sively demonstrated in Fig. 7, which shows theintegrated absorbance spectrum around thecadmium resonance line at 228.802 nm.

Obviously, instruments for CS AAS can bedesigned in several different ways, and their fea-

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]2094 2091

Ž . Ž . Ž .Fig. 6. Experimental setup for CS AAS with double echelle monochromator DEMON ; 1 Xenon short-arc lamp; 2 hollowŽ . Ž . Ž .cathode lamp for emission measurements; 3 off-axis ellipsoid mirrors; 4 longitudinal Zeeman graphite furnace module; 5

Ž . Ž . Ž . Ž . Žentrance slit; 6 off-axis parabolic mirrors; 7 Littrow prism; 8 deflection mirrors and intermediate slit; 9 echelle grating 75. Ž . Ž w x.groovesrmm, blaze 768 ; and 10 linear CCD array detector from Heitmann et al. 49 .

tures depend on their particular design. The setupshown in Fig. 6 is still of the monochromatortype, i.e. designed for the determination of oneelement at a time. However, it provides a varietyof advantages over conventional LS AAS. Firstly,

the atomic absorption can not only be measuredŽat the center of the absorption line with maxi-

. Žmum sensitivity but also at its wings with.reduced sensitivity , thus greatly increasing the

dynamic range to approximately 5]6 orders of

Fig. 7. Time-integrated absorbance spectrum around the Cd resonance line at 228.802 nm, measured with CS AAS and DEMON;Ž w x.spectral resolution 2.3 pmrpixel from Becker-Ross et al. 50 .

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]20942092

magnitude in concentration or mass. This,together with detection limits that according to

w xrecent reports 45 tend to be even better thanthose obtained with LS AAS, and with the expec-tation of further improvement, eliminates one ofthe classical disadvantages of AAS } its limiteddynamic range.

Secondly, the setup shown in Fig. 6 with a CCDdetector provides a lot of information about thespectral neighborhood around the analytical line,which eventually results in a much more reliableand accurate background correction, as essen-tially any pixel or set of pixels may be used forthat purpose, depending on the nature of thebackground. In addition, as this wavelength-resolved absorbance is measured over time witheach pixel, a three-dimensional absorbance pat-tern is obtained eventually, as shown in Fig. 8 forthe background signal of 5 ml of urine in thevicinity of the palladium line at 247.642 nm. Thiskind of structured background makes a correctionwith a deuterium lamp impossible, and presents aproblem even for Zeeman-effect background cor-

w xrection 50 . In the case of the CS AAS, however,

it was possible to eliminate the backgroundalmost completely by a least-squares backgroundcorrection with the normalized spectra of PO and

w xNa 50 .To extend the capabilities of CS AAS to its full

capability, that is simultaneous multi-elementAAS, requires the replacement of the one-dimen-sional array detector by a two-dimensional multi-array detector. This would obviously have a sig-nificant impact on the cost and the complexity ofthe whole instrument, if all the above-describedfeatures are to be retained. A sacrifice in infor-mation obtained by significantly reducing thenumber of pixels per array could be a com-promise which, however has yet to be investi-gated. However, as also pointed out by Harnlyw x45 , this author believes that the advantages ofCS AAS over LS AAS, even for the determina-tion of one element at a time, with respect todetection limit, dynamic range, spectral informa-tion, and background correction capabilities, areso convincing that it is about time for CS AAS toreplace conventional LS AAS step by step. Thesimultaneous multi-element version will then fol-

Fig. 8. Absorption spectrum of an undiluted human urine sample, spiked with 25 pg Pd, recorded with CS AAS and DEMON.Ž w x.Linear dispersion: 2.6 pmrpixel; time resolution: 16.7 msrscan from Becker-Ross et al. 50 .

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]2094 2093

low with the increasing acceptance of CS AAS,and as prices for two-dimensional array detectorsdrop.

4. Conclusion

Atomic absorption spectrometry is withoutdoubt pregnant, and will give birth to new off-spring very soon, provided the instrument manu-facturers are prepared to give their assistance.Some of the application-oriented innovationscould be incremented at a comparably small ef-fort; others, such as solid sampling for GF AASare already available, and only wait for their moregeneral acceptance. Diode laser AAS is still asmall niche market, but there is hope that themost rewarding fields of application for this tech-nique, i.e. dedicated instruments, detectors forchromatography, and determination of non-metals, will be exploited soon and introduced intothe market properly. Obviously, the manufactur-ers of laser diodes have to recognize the impor-tance of the analytical and sensor market beforefurther progress can be expected. Continuumsource AAS, finally, has the greatest potential inthe opinion of this author, and it only requiresthe decision of a courageous instrument manufac-turer to re-design the image of AAS, making ityoung, bright, and prosperous again.

References

w x Ž .1 G. Kirchhoff, Ann. Phys. Poggendorf’s Annalen 109Ž .1860 275.

w x Ž .2 G. Kirchhoff, R. Bunsen, Philos. Mag. 20 1860 89.w x Ž .3 G. Kirchhoff, R. Bunsen, Philos. Mag. 22 1861 329]349.w x Ž .4 A. Walsh, Spectrochim. Acta Part B 35 1980 643]652.w x5 C.Th.J. Alkemade, J.M.W. Milatz, A double-beam

method of spectral selection with flames, Appl. Sci. Res.Ž .B4 1955 289]299.

w x6 C.Th.J. Alkemade, J.M.W. Milatz, J. Opt. Soc. Am. 45Ž .1955 583]584.

w x7 A. Walsh, The application of atomic absorption spectraŽ .to chemical analysis, Spectrochim. Acta 7 1955

108]117.w x8 B.V. L’vov, Twenty-five years of furnace atomic absorp-

Ž .tion spectroscopy’, Spectrochim. Acta Part B 39 1984149]157.

w x9 B. Welz, M. Sperling, Atomic Absorption Spectrometry,3rd ed, Wiley-VCH, Weinheim, New York, 1999.

w x10 A. Walsh, Atomic absorption spectroscopy } stagnantŽ .or pregnant?, Anal. Chem. 46 1974 698A]708A.

w x11 B.V. L’vov, Electrothermal atomization } the way to-wards absolute methods of atomic absorption analysis,

Ž .Spectrochim. Acta Part B 33 1978 153]193.w x12 W. Slavin, D.C. Manning, G.R. Carnrick, The stabilized

Ž .temperature platform furnace, At. Spectrosc. 2 1981137]145.

w x13 G.M. Hieftje, Atomic absorption spectrometry } has itgone or where is it going?, J. Anal. At. Spectrom. 4Ž .1989 117]122.

w x14 B. Welz, M. Sperling, Flow injection: the ultimate ap-proach to automation in analytical atomic spectroscopy,

Ž .Pure Appl. Chem. 65 1993 2465]2472.w x15 Z.-L. Fang, M. Sperling, B. Welz, Flow-injection online

sorbent extraction pre-concentration for graphite fur-nace atomic absorption spectrometry, J. Anal. At. Spec-

Ž .trom. 5 1990 639]646.w x16 Z.-L. Fang, L.-P. Dong, Flow-injection online co-precipi-

tation pre-concentration for electrothermal atomic ab-Ž .sorption spectrometry, J. Anal. At. Spectrom. 7 1992

439]446.w x17 M. Sperling, X.-F. Yin, B. Welz, Flow-injection online

separation and pre-concentration for electrothermalatomic absorption spectrometry. I. Determination ofultra-trace amounts of cadmium, copper, lead and nickel

Ž .in water samples, J. Anal. At. Spectrom. 6 1991295]300.

w x18 M. Sperling, X.-F. Yin, B. Welz, Flow-injection onlineseparation and pre-concentration for electrothermalatomic absorption spectrometry. II. Determination ofultra-trace amounts of cobalt in water, J. Anal. At.

Ž .Spectrom. 6 1991 615]621.w x19 M. Sperling, X.-P. Yan, B. Welz, Electrothermal atomic

absorption spectrometric determination of lead in high-purity reagents with flow-injection on-line microcolumnpreconcentration and separation using a macrocycle im-mobilized silica gel sorbent, Spectrochim. Acta Part B

Ž .51 1996 1875]1889.w x20 J.N. Lockyer, Studies in Spectrum Analysis, Appleton,

London, 1878.w x Ž .21 U. Kurfurst Ed. , Solid Sample Analysis, Springer-¨

Verlag, Berlin, Heidelberg, New York, 1998.w x22 K.-Ch. Friese, V. Krivan, A solid-sampling system for a

transversely heated graphite furnace and its applicationto trace element analysis of high-purity tantalumpowders by atomic absorption spectrometry, Spec-

Ž .trochim. Acta Part B 53 1998 1069]1078.w x23 G.R. Carnrick, G. Daley, A. Fotinopoulos, Design and

use of a new automated ultrasonic slurry sampler forgraphite furnace atomic absorption, At. Spectrosc. 10Ž .1989 170]174.

w x24 N.J. Miller-Ihli, Automated ultrasonic mixing accessoryfor slurry sampling into a graphite furnace atomic ab-

Ž .sorption spectrometer, J. Anal. At. Spectrom. 4 1989295]297.

( )B. Welz r Spectrochimica Acta Part B: Atomic Spectroscopy 54 1999 2081]20942094

w x25 M. Lucic, V. Krivan, Solid sampling electrothermalatomic absorption spectrometry for analysis ofaluminium oxide powders, J. Anal. At. Spectrom. 13Ž .1998 1133]1139.

w x26 M. Hornung, V. Krivan, Determination of trace impuri-ties in high-purity tungsten by direct solid samplingelectrothermal atomic absorption spectrometry using a

Ž .transversely heated graphite tube, Anal. Chem. 70 19983444]3451.

w x27 M. Hornung, V. Krivan. Solid sampling electrothermalatomic absorption spectrometry for analysis of high-pur-ity tungsten trioxide and high-purity tungsten blue oxide,

Ž .Spectrochim. Acta Part B 54 1999 1177]1191.w x28 B. Welz, Speciation analysis. The future of atomic ab-

Ž .sorption spectrometry, J. Anal. At. Spectrom. 13 1998413]417.

w x29 T. Guo, J. Baasner, D.L. Tsalev, Fast automated de-termination of toxically relevant arsenic in urine by flowinjection-hydride generation atomic absorption spec-

Ž .trometry, Anal. Chim. Acta 349 1997 313]318.w x30 M. Sperling, X.-F. Yin, B. Welz, Differential determina-

Ž .tion of chromium VI and total chromium in naturalwaters using flow injection on-line separation and pre-concentration electrothermal atomic absorption spec-

Ž .trometry, Analyst 117 1992 629]635.w x31 M. Sperling, S.-K. Xu, B. Welz, Determination of

Ž . Ž .chromium III and chromium VI in water using flowinjection on-line preconcentration with selective adsorp-tion on activated alumina and flame atomic absorption

Ž .spectrometric detection, Anal. Chem. 64 19923101]3108.

w x32 J.C. Van Loon, B. Radziuk, A quartz tube furnace-atomic absorption spectroscopy system for metal specia-

Ž .tion studies, Can. J. Spectrosc. 21 1976 46]50.w x33 O.F.X. Donard, S. Rapsomanikis, J.H. Weber, Specia-

tion of inorganic tin and alkyltin compounds by atomicabsorption spectrometry using electrothermal quartz

Ž .furnace after hydride generation, Anal. Chem. 58 1986772]777.

w x34 X.-R. Zhang, R. Cornelis, J. DeKimpe, L. Mees, Arsenicspeciation in serum of uraemic patients based on liquidchromatography with hydride generation atomic absorp-tion spectrometry and on-line UV photo-oxidation di-

Ž .gestion, Anal. Chim. Acta 319 1996 177]185.w x35 D.L. Tsalev, M. Sperling, B. Welz, On-line microwave

sample pre-treatment for hydride generation and coldvapour atomic absorption spectrometry. Part 1. The

Ž .manifold, Analyst 117 1992 1729]1733.w x36 D.L. Tsalev, M. Sperling, B. Welz, On-line microwave

sample pre-treatment for hydride generation and coldvapour atomic absorption spectrometry. Part 2. Chem-

Ž .istry and applications, Analyst 117 1992 1735]1741.w x37 M. Sperling, Z.-L. Fang, B. Welz, Expansion of dynamic

working range and correction for interferences in flameatomic absorption spectrometry using flow-injection gra-dient ratio calibration with a single standard, Anal.

Ž .Chem. 63 1991 151]159.w x38 H. Falk, Analytical capabilities of atomic spectrometric

methods using tunable lasers: a theoretical approach,Ž .Prog. Anal. Spectrosc. 3 1980 181]208.

w x39 K. Niemax, A. Zybin, C. Schnurer-Patschan, H. Groll,¨Semiconductor diode lasers in atomic spectrometry,

Ž .Anal. Chem. 68 1996 351A]356A.w x40 A. Zybin, C. Schnurer-Patschan, M.A. Bolshov, K.¨

Niemax, Elemental analysis by diode laser spectroscopy,Ž .Trends Anal. Chem. 17 1998 513]520.

w x41 V. Krivan, P. Barth, C. Schnurer-Patschan, An elec-¨trothermal atomic absorption spectrometer using semi-conductor diode lasers and a tungsten coil atomizer:

Ž .design and first applications, Anal. Chem. 70 19983525]3532.

w x42 H. Groll, G. Schaldach, H. Berndt, K. Niemax, Measure-Ž . Ž .ment of Cr III rCr VI species by wavelength modula-

tion diode laser flame atomic absorption spectrometry,Ž .Spectrochim. Acta Part B 50 1995 1293]1298.

w x43 A. Zybin, C. Schnurer-Patschan, K. Niemax, Wavelength¨modulation diode laser atomic absorption spectrometryin modulated low-pressure helium plasmas for element-selective detection in gas chromatography, J. Anal. At.

Ž .Spectrom. 10 1995 563]567.w x44 LaserSpec Analytik, Munchen, Germany, 1998.¨w x45 J.M. Harnly, The future of atomic absorption spec-

trometry: a continuum source with a charge coupledŽ .array detector, J. Anal. At. Spectrom. 14 1999 137]146.

w x46 H. Becker-Ross, S. Florek, U. Heitmann, R. Weisse,Influence of the spectral bandwidth of the spectrometeron the sensitivity using continuum source atomic ab-sorption spectrometry, Fresenius J. Anal. Chem. 355Ž .1996 300]303.

w x47 S. Florek, H. Becker-Ross, T. Florek, Adaption of anechelle spectrograph to a large CCD detector, Fresenius

Ž .J. Anal.Chem. 355 1996 269]271.w x48 H. Becker-Ross, S. Florek, Echelle spectrometers and

charge-coupled devices, Spectrochim. Acta Part B 52Ž .1997 1367]1376.

w x49 U. Heitmann, M. Schuetz, H. Becker-Ross, S. Florek,Measurements on the Zeeman splitting of analyticallines by means of a continuum source graphite furnaceatomic absorption spectrometer with a linear charge-

Ž .coupled device array, Spectrochim. Acta Part B 51 19961095]1105.

w x50 H. Becker-Ross, S. Florek, U. Heitmann, M. Schuetz,Continuum-source atomic absorption spectrometry }

the second dimension in absorbance measurements.Fifth Rio Symposium on Atomic Spectrometry, Cancun,Mexico, 1998.