Instrumentation Leo de Galan Laboratorium voor Analytische Scheikunde Technische Hogeschool Delft de Vries van Heystplantsoen 2 2628 RZ DELFT, the Netherlands
NEW DIRECTIONS IN OPTICA
ATOMIC SPECTROMETRY
Soon after its invention a successful method of analysis goes through a phase of rapid growth and exaggerated expectations before it recedes to a more balanced position in the analytical domain. Flame and furnace atomic absorption spectrometry (AAS) and inductively coupled plasma-atomic emission spectrometry (ICP-AES), as we know them now, were introduced 20 to 30 years ago, developed into commercial instruments within a decade after their first description in the scientific literature, and have now reached a state of developmental equilibrium. It is undeniable that these techniques have continued to develop, but recent advances have been largely technical and cosmetic. The emphasis on automation and software has made life much easier and has significantly reduced the demand for manpower, but it has not enlarged the analytical scope of the techniques. Many initial promises have been fulfilled, but some shortcomings persist even today. At this point it would be easy to formulate the ideal method that determines all elements from the sub-partsper-billion level to the 100% level, simultaneously, with high precision and accuracy, and at minimal cost. Clearly, no single method can possibly match such unrealistic expectations. In this article the author has, therefore, chosen the more modest approach of identifying some weak points in available technology and analyzing possible 0003-2700/86/0358-697A$01.50/0 © 1986 American Chemical Society
remedies. In several cases current developments are reviewed, and novel instruments proposed in the literature are evaluated. In other cases, the problems have hardly been addressed and thus may pose a challenge for future research.
Multielement AAS Textbooks on AAS stress that it is a single-element technique, because a separate hollow-cathode lamp is needed for each element. It is interesting to determine whether this is a fundamental restriction or the result of technical limitations. Provisions for automatic repositioning of up to 10 lamps offer only a partial solution, as the method remains sequential. It demonstrates, however, that the technical problems relating to the light source are surmountable. Multielement lamps can be constructed, or the radiation of many lamps can be combined with appropriate optics. Naturally, the monochromator can and must be replaced by a multichannel detector. A more fundamental objection can be raised to the use of hollow-cathode lamps for multielement AAS, either simultaneously or sequentially. Given the logarithmic relation between intensity and concentration, the dynamic range of the absorption process at a given wavelength is restricted to at most two decades (0.01 to 1 au) and usually less. With a paucity of useful absorption transitions for some ele-
ments, this means that sample dilution must correspond to the sample composition. As a result, different samples or even different elements in one sample may require different dilutions. The problem can be alleviated if we can shift the primary wavelength from its usual position at the peak of the atomic absorption line profile to the wings of this profile. One possibility is the use of the Zeeman technique, better known for its capabilities for background correction. If we use a more moderate field, we can shift the line partly from its peak position to obtain an analytical curve with lower sensitivity (slope). An elegant solution is offered by an ac-driven magnet around the furnace atomizer. When data are collected at three rather than two field strengths, two backgroundcorrected signals of different sensitivity are obtained and the dynamic range of AAS can be extended by another decade. A more far-reaching alternative is offered by a continuum lamp in combination with a high-resolution monochromator (Figure 1). The extensive work of O'Haver has amply demonstrated that simultaneous, multielement AAS is technically feasible and that the effective dynamic range can be extended to five decades by shifting the wavelength to the wings of the absorption line profile (1). A limitation remains in the far-UV region, where the intensity of current contin-
ANALYTICAL CHEMISTRY, VOL. 58, NO. 6, MAY 1986 • 697 A
(a)
L, ; /
1 Power
L2 Furnacey
Ίο
Scanner controller
f
I M-,
LI.
QP EG
supply
Ρ
Echelle monochromator Amplifier ]—>|
1
M,
Oscilloscope
Multiplexer Gas inlet A/D
Computer Isolation
=3TTC Teletype
Sample Strip chart
Line printer
Ouartz window
Heat supply
Disk
Figure 1. Two approaches to simultaneous determinations with the graphite furnace (a) Simultaneous multielement atomic absorption with a continuum source ( 1). EIMAC: continuum xenon arc source; L: lens; QP: quartz plate; EG: échelle grating; P: prism; M: mirror, (b) FANES (cf. Reference 4, courtesy of Kontron Spektralanalytik, Eching, F.R.G.)
uum sources is too low to provide adequate detection limits (Table I). However, a remaining problem is the atomizer. If elements are to be determined simultaneously, they must be atomized under a single set of uniform conditions. Possibly, such compromise conditions can be formulated for high-temperature flames. However, in terms of analytical performance, such an instrument would meet stiff competition from ICP-AES, especially when the cost of the latter dimin-
ishes, as will be discussed later. For their superior sensitivity electrothermal atomizers (ETAs) are a better choice for multielement A AS, but compromise conditions are much more difficult to formulate here. Until this critical issue has been decided, the fate of multielement AAS remains uncertain. Improved electrothermal atomization
Even though a uniform temperature program for all elements may be a re-
Table I. Detection limits (ng/mL) of established and novel methods Element
Au Β Cd
Co Cr Cu Fe Mn Pb Ti V
Zn
ICP-AES
ICP-MS
20 5 2 4 4 3 5 0.5 30 2 4 2
0.2 2 0.7 0.8 0.2 0.3 0.5 1.3 0.4 0.5 0.6 4.3
ETA-AAS
SIMAAC
0.1 0.005 0.05 0.02 0.02 0.03 0.01 0.02 10 1 0.001
FANES
0.06 0.03 0.02 0.9 0.1 0.7 0.1
0.9 0.4
0.08 0.02 0.09 0.06 16 0.12 0.04
ICP-AES: Inductively coupled plasma-atomic emission spectrometry ICP-MS: Inductively coupled plasma-mass spectrometry ( 10) ETA-AAS: Electrothermal atomization-atomic absorption spectrometry using hollow-cathode single-element sources SIMAAC: Simultaneous multielement atomic absorption using a continuum source and graphite furnace atomization ( 1) FANES: Furnace atomization nonthermal emission spectrometry (4)
698 A • ANALYTICAL CHEMISTRY, VOL. 58, NO. 6, MAY 1986
mote prospect, there are other aspects of electrothermal atomization that warrant improvement. A wider range of elements, the ability to cope with a large excess of matrix, and, in general, more trust in the result of the analysis are clearly desired. No doubt, all three aspects relate to the construction of the furnace—its geometry and base material. Important steps toward this goal have already been made (Figure 2). Interestingly, the introduction of the platform technique and tubes coated with pyrolytic graphite resulted from fundamental studies on the volatiliza tion and the dissociation processes in the furnace. Similarly, improved in strumentation and performance may be expected from a better insight into background correction techniques and the temperature distribution along the furnace. These examples demonstrate the benefits to be gained from well-ex ecuted basic research. There can be little doubt that the deuterium lamp will soon be obsolete and will be re placed by the more powerful Zeeman and pulsed lamp (Smith-Hieftje) techniques for background correction. Similarly, one can expect that future instruments dedicated to ETA-AAS will use controlled heating based on true furnace temperature, preferably referring to the inside wall. Even then, there will still be a need for an inert furnace material that can be used up to 3300 Κ with constant performance during 500 firings. As glassy carbon has not lived up to ini tial expectations, hope is now centered
Figure 2. The Zeeman background correction technique and the inserted platform have contributed much to improve the reliability of graphite furnace AAS; in the future such furnaces may be produced from total pyrolytic graphite
on total pyrolytic graphite (2). Meanwhile, there is ample room for research into high-temperature chemistry phenomena occurring at the surface and in the atmosphere of the furnace. The necessarily limited and often ambiguous thermodynamic studies executed up to now could benefit tremendously from mass-spectrometric probing and surface analysis techniques. Unfortunately, these are expensive and rather fundamental experiments that cannot be easily financed. Perhaps this is also the place to speculate on the possibility of absolute (i.e., calibration-free) analysis. As indicated by L'vov, ETA-AAS is in the best position to achieve this goal. Indeed, characteristic concentrations measured with a good instrument conform to predictions based on pure physics to within 20%. This is certainly an impressive scientific accomplishment and demonstrates the simplicity of the furnace as a model source of atoms. Nevertheless, the practical significance should not be overrated. Obviously, a similar or better accuracy can be obtained with a single calibration. More important is the question of whether the characteristic concentration (i.e., the analytical sensitivity) remains the same in the presence of a matrix. Indeed, improved resistance to interferences is one of the most desirable properties expected from future furnace materials, such as pyrolytic graphite. A real disadvantage of electrothermal atomization is the long cycle time. The drying, ashing, atomization, cleaning, and consecutive cooling of the furnace take between two and three minutes. Omission of the ashing
step brings some relief but does not appear to be generally applicable. A drastic reduction in analysis time is possible only when the desolvated sample is flash volatilized upon introduction into a preheated furnace. Past proposals such as aspiration of sample solutions or inserted probes do not meet these requirements. Surely, the excess water is unacceptable if we are to retain analytical sensitivity, and the heating of even a small probe takes time. A principally better approach is the hanging-droplet technique, whereby sample droplets are suspended on a carbon thread that is pulled through the hot furnace. If the droplets are dried before they enter the furnace, the cycle time is indeed reduced. Unfortunately, the technique is far from practical, but the physical considerations were correct. It demonstrates that the problem is solvable, in principle. It just needs a fresh look to develop a system that is practical to use. Furnace emission spectrometry
The long cycle time of current electrothermal atomizers emphasizes the importance of simultaneous, multielement furnace analysis. Given the difficulty of simultaneous AAS, it is natural that efforts have been made to combine the furnace with emission techniques. Although the registration of emission signals from the furnace itself is certainly the simplest approach, the low temperature during the release of volatile elements leads to poor detection limits for high-energy transitions (e.g., Cd, As). Naturally, the literature abounds with proposals to sweep the vapor released from an electrothermal atomizer into a plasma
700 A • ANALYTICAL CHEMISTRY, VOL. 58, NO. 6, MAY 1986
source. Both microwave-induced plasmas (MIPs) and ICPs have been used for this purpose. Unfortunately, the gain in sensitivity over straightforward ICP-AES is less than hoped for, and the long transport lines create their own problems. Better results can be expected from the two-furnace design of Freeh (3). The sample is volatilized from one furnace into a second furnace preheated to its maximum temperature. Although designed to separate volatilization and dissociation processes in furnace absorption, the combined furnace could have potential for furnace emission. The same principle underlies the approach taken by Falk (4). In this approach, the vapor released from a graphite furnace is postexcited in a radio frequency hollowcathode discharge. The entire assembly operates at a low pressure of a few torr, which accounts for a high effective excitation temperature in the nonthermal discharge. This phenomenon is expressed by the acronym FANES, coined by Falk for his technique—furnace atomization nonthermal emission spectrometry (Figure 1). Impressive detection limits have been reported for many elements including some nonmetals (Table I), but much more research is needed to establish this technique's simultaneous multielement potential. As in multielement AAS, uniform atomization conditions are a prerequisite. Also, the fear of enhanced interferences arising from the nonthermal hollow-cathode discharge should be alleviated. High-resolution ICP
Spectral interferences are still the major source of error in analyses with the ICP when it is used as a source of optical radiation in AES. Surprisingly, in this age of automation, we must still rely on human judgment to recognize and avoid them. Fancy software will help by projecting spectra and calculating corrections, but we are still a long way from the situation in which a novel sample can be safely trusted to automated initiation and optimization of the analysis. At first sight, higher resolution seems the answer to spectral interference problems. Indeed, échelle spectrometers are available, and a Fourier transform UV spectrometer has been described for this purpose. Unfortunately, spectral interferences do not diminish in proportion to the vastly superior resolution offered by these instruments. In fact, soon after we surpass the resolution of 0.01 nm provided by conventional grating instruments, we meet the natural boundary set by the physical width of spectral lines emitted by the ICP, which varies from 0.003 to 0.02 nm. It is therefore
doubtful whether high-resolution in struments warrant the additional ex pense. It would appear that a much sim pler solution is possible. Overlapping spectral curves are not unique to ICP-AES but are well-known in IR spectrometry, neutron activation anal ysis, and energy-dispersive X-ray fluo rescence spectrometry, to name only a few. Curve resolution techniques de veloped successfully in these areas should be equally profitable in ICP-AES. Improved data processing could also solve the related problem of background correction, which is still performed either on-line or off-line at discrete wavelengths selected by the operator. Recently, Taylor has adapt ed software routines from neutron ac tivation analysis to assess the spec trum around an analysis line in each individual sample (5). Superior back ground correction and removal of in terfering lines were achieved at the cost of a longer time needed to scan the wavelength region of interest. However, Taylor's suggestion that multichannel (e.g., photodiode array) detectors could overcome the time dis advantage seems premature. Current photodiodes are inferior to the photomultiplier tube in dynamic range and in sensitivity, certainly in the far UV. A more fundamental objection is the pixel size of 25 μπι, which corresponds nicely to the width of the entrance slit but does not allow collection of many intensity data across the spectral line as is needed for curve resolution. If we are to collect spectral data with 0.001-nm step size, we need either a very high dispersion monochromator or much smaller pixels of 1 μπι. The latter are not yet available. ICP-AFS and ICP-MS When we realize that spectral prob lems in ICP-AES arise from the den sity of emission spectra, then we should look at alternative registration systems. Despite their high selectivity, absorption techniques appear useless because the current weak primary sources cannot overcome the intense radiation of the ICP. Also, we would give up multielement capability. Atomic fluorescence spectroscopic (AFS) techniques retain multielement capability and offer remarkably empty spectra. Unfortunately, the primary source of radiation is again their weak point. Continuum sources would be preferable because they stimulate all elements simultaneously, but even the brightest continuum source available is much too weak. Hollow-cathode lamps are cheap enough to arrange a collection of 20 lamps around the dis charge as is done in the only commer cial ICP-AFS available. Unfortunate
ly, hollow-cathode lamps have limited intensity, especially for the ion lines that are most useful in ICP analysis. Indeed, when used for AFS, the ICP discharge is run under uncommon conditions to promote the atom frac tion over the ion species. Even then, detection limits are inferior to I C P AES for many elements. Lasers are the obvious answer to our prayers, but a dependable, tunable, low-cost laser that reaches the far UV is still not available. Given the high degree of ionization of most elements in the ICP, it is not surprising that mass spectrometric de tection receives much attention. The work of Gray in the United Kingdom and Houk in the United States (6) has been remarkably rapidly transformed into commercial instruments. Poten tially, the mass spectrometer promises cleaner spectra and a higher and more uniform sensitivity (Table I). Inevita bly, the first applications currently re ported in conferences present a more sobering view, but there is still cau tious optimism that the expectations will be borne out. Undoubtedly, the interface between the atmosphericpressure, high-temperature ICP and the low-pressure quadrupole mass spectrometer plays a crucial part. Ap parently, the efficiency by which the small pinhole extracts ions from the ICP is influenced by the sample com position, which gives rise to chemical interferences not known in ICP-AES. This problem seems surmountable. Molecular fragments may pose more difficulties, as a simple calculation shows. If matrix constituents are at omized to 99.99%, then the 0.01% frac tion of molecules still forms a large ex cess in comparison with the trace con stituents determined at the parts-permillion level. Because molecules are very weak emitters of radiation, their presence goes unnoted in ICP-AES. However, in ICP-MS the correspond ing molecular ions are detected with equal sensitivity and may thus give rise to interfering mass overlap in the ICP-MS spectrum. Consequently, both the interface and the ICP itself must be optimized for virtually com plete dissociation of matrix constitu ents. ICP-MS is still under rapid devel opment, and it is far too early for a de finitive assessment. For one thing, the currently high price of the instrument must come down before ICP-MS be comes a real alternative to ICP-AES. Excitation sources for AES Despite the popularity of the ICP, other sources for the atomization and excitation of dissolved samples still re ceive attention, although a true alter native has not yet emerged. Apparent
702 A • ANALYTICAL CHEMISTRY, VOL. 58, NO. 6, MAY 1986
ly, it is difficult to match, let alone beat the two major features of the ICP: low detection limits for many ele ments and high resistance against ma trix effects. Incidentally, these favor able properties of the ICP have evolved empirically and are gratefully accepted rather than satisfactorily ex plained. The many, and often repeti tive, studies of spatial distributions of temperature, electron concentration, and sample constituents contribute little to our theoretical understanding. Consequently it would be naive to ex pect a superior discharge to emerge from such studies. Curiously, potential alternatives to the ICP are generally recommended for cost-effectiveness. Of these, the dc arc discharge had the longest history before it evolved to its present config uration of three electrodes arranged as a Y. Detection power seems adequate, but interferences remain a weak point requiring the addition of ionization buffers. An interesting new proposal is an arc discharge generated inside a sil ica tube by three to six externally placed electrodes (7). Further devel opment of this design must be await ed. When run on argon at atmospheric pressure, the MIP can accept a nebu lized solution and detect metals. How ever, detection limits are barely ade quate, and it is difficult to see how matrix effects can become tolerable unless the power is raised from the current 100 W to appreciably higher values. The difference with the ICP would then become marginal, in per formance and in price. Perhaps this is also the place to mention the nonspectroscopic but an alytically related technique of laserenhanced ionization (LEI). When an analyte is atomized in a flame and ir radiated by an intense beam of reso nance radiation from a tunable laser, its ionization is greatly enhanced and a current can be detected with elec trodes placed around the flame. Again, detection limits are good, but interfer ences from other ions in the flame pose problems. More important from a practical point of view is the laser source, which increases the cost and makes the technique sequential rather than simultaneous. If cost-effectiveness is indeed a ma jor concern, then more can be expect ed from novel torch designs for the ICP proper (8). By carefully restrict ing the annular spacing between the two outer tubes, the so-called high-ef ficiency torches can be operated on 1 kW power and less than 10 L/min of total argon. When the torch is cooled externally with water or air, the argon demand can be reduced to a value as low as 1 L/min (Figure 3). In either
Figure 3. An air-cooled torch with an ICP sustained by 1 L/min total argon and 600 W rf power
case, the analytical performance matches that of currently available torches. In the future, external cooling may be obviated when a suitable hightemperature ceramic is found to re place the silica outer tube. It is expect ed that these designs will soon be in corporated into commercial low-cost ICP instruments. Sample introduction into the ICP
If users, manufacturers, and experts of the ICP agree on one thing, it is on the need for better introduction de vices for dissolved samples. The exist
ing pneumatic nebulizer-chamber combinations are wasteful of sample and noisy. The unfortunate reappear ance of the internal standard tech nique in atomic emission is an omi nous sign. Its success in improving the analytical precision of ICP analysis from 2% to below 0.5% identifies the sample introduction step as the major source of imprecision, but it accommo dates rather than solves the problem. To be sure, alternatives to the inad equate cross-flow and concentric neb ulizers have been proposed. However, the V-groove Babington nebulizer (Figure 4) offers no better precision or efficiency, although it is superior in some other respects. The ultrasonic nebulizer has probably generated more frustrated scientists than any other piece of atomic spectrometric equipment. There are, however, alter native nebulization principles to be tested and better aspiration chambers to be constructed. If anything, this would appear to be a fruitful topic for research. On the more general theme of sam ple transfer to an atomic spectrome ter, wide perspectives are opened by the use of flow injection techniques. In their simplest form these techniques guarantee a regular flow of carrier liq uid of constant viscosity. With only slightly more complex manifolds we can provide for the addition of matrix modifiers, ionization buffers, and in ternal standards. By including an ex ponential dilutor we can provide alter native ways of calibration. Solid samples
The prospect of being released from the tedious and time-consuming sam ple dissolution step in an analysis re currently induces studies into the pos sibility of direct introduction of solid samples. It is really not surprising that success has been limited. If, as we have seen above, we already have
problems in introducing liquid sam ples into the ETA and the ICP, how can we expect to cope with the much greater problems of solid samples? It is instructive to recall the by now al most forgotten carbon arc, where a few milligrams of a coarsely ground solid sample were heated to 4000 Κ in the arc electrode to evaporate freely—and erratically—over a period of several minutes. Current atomizers are lack ing in either the temperature or the residence time or both. As a result, volatilization and, hence, atomization remain incomplete, depending on the composition and the size of the parti cles. Calibration against carefully matched standards becomes the rule, unless the particle size is reduced to about 1 μτη. For most samples grind ing to such small dimensions is more arduous and time-consuming than dis solution. Indeed, electrothermal atomizers can cope with only a few milligrams of the most volatile, usually organic sam ples. Even then, sample inhomogeneity and dispensing problems account for poor precision. In the ICP, slurry techniques in conjunction with Bab ington nebulizers have occasionally been used with some success, but have limited applicability. Free evaporation from carbon cups lifted into the dis charge region of a high-power ICP gives rise to spectacular pictures but doubtful analytical results. A better approach is to separate the evaporation of solids from their exci tation. Lasers and high-current sparks have been used to generate a vapor that is swept into an ICP for consecu tive excitation and detection. Al though an already available ICP can be exploited in this way, it is a little foolish to buy one for this purpose. Sputtering techniques are used much more successfully in specially designed instruments, such as the glow dis charge and, of course, the spark spec-
Argon Sampl
Sample
Argon
V-groove Concentric Cross-flow
Sample-*Argon
Figure 4. Nebulizers for the ICP have evolved from cross-flow to concentric and V-groove models, but further improvement is needed 704 A • ANALYTICAL CHEMISTRY, VOL. 58, NO. 6, MAY 1986
trometer itself. Harrison and, more re cently, Caroli (9) have reported inter esting results for solid samples excited in a low-pressure hollow-cathode dis charge. This approach warrants fur ther attention. And let us not forget that X-ray fluorescence spectrometry is an excellent method for analyzing solid samples. Nonmetals
All optical atomic spectrometric methods of analysis primarily address metallic elements. Again, it is interest ing to analyze whether the exclusion of nonmetals is attributable to funda mental restrictions or to technical lim itations. Atomization should not be difficult. If metals can be dissociated from their chemical environment, then the halogens, phosphorus, and sulfur should also be atomized to a high degree. Carbon, hydrogen, nitro gen, and oxygen associate mutually more strongly. However, for these common elements the background lev els from impurities in the feed gas and from the solvents usually prohibit low limits of detection. Invariably, spectral transitions from nonmetals have a high excitation ener gy. Excitation temperatures of 6000 K, reached in (approximately) thermal sources operated at atmospheric pres sure, are too low to generate sufficient emission intensities. Indeed, the de tection limits for I, P, and S in the ICP are substantially higher than the nanograms-per-milliliter values cited for most metals. Incidentally, ICP-MS offers no respite, because ef ficient ionization of nonmetals is pre vented by their ionization energies, which are also high. Much higher electronic excitation temperatures can be reached in non thermal sources, especially when oper ated at reduced pressure. For exam ple, effective excitation temperatures as high as 100,000 Κ have been report ed for MIPs operated in a few torr of He at low power (50 W). Unfortunate ly, the kinetic temperature of this and similar sources is very low (
