The Future of Plasma Spectrochemical Instrumentation* Plenary Lecture GARY M. HIEFTJE Department of Chemistry Indiana University Bloomington IN 47405 USA Every decade seems to bring with it some important novelty in atomic spectrochemical instrumentation. During the past decade however the changes seemed to be more evolutionary than revolutionary. Emission spectrometric instrumentation has become less expensive and more capable with the introduction of advanced user interfaces lowered detection limits that now approach those for electrothermal atomization in AAS axial-viewing options that provide greater stability and higher sensitivity but at the cost of elevated interferences and multichannel detector arrays that enable an entire emission spectrum to be viewed at once. Similarly MS instrumentation has evolved to a simpler less expensive form even as capabilities have increased.In atomic MS the origins of several troublesome interference effects have been identified and substantially reduced and strategies have been devised to reduce the severity of isobaric overlaps (spectral interferences). Because of these trends sales of emission-based instrumentation have remained brisk while those for atomic mass spectrometers have risen dramatically. The combination of relatively high sales and higher capabilities has encouraged a number of new instrument manufacturers to enter the plasma spectrochemical market while others have been forced to drop out to consolidate or to be acquired. In attempting to project the future of atomic spectrochemical instrumentation it is a safe bet to assume that past trends will continue. Revolutionary changes are of course much more difficult to forecast.Nevertheless some guidance can be derived from reviewing the limitations of current sources and detection techniques that are used for plasma spectrometry. Such a review reveals that detection limits in emission measurements are usually constrained by background noise levels whereas those in MS are bounded by the efficiency of sample utilization and by the transmission of the mass spectrometer and of the interface that separates it from the plasma source. Also although the origin of matrix interferences in both atomic emission and atomic mass spectrometry are not fully understood it seems clear that what is needed is better control of sample introduction atom formation and the plasma environment that fosters atomic excitation and ionization.Further it is obvious that more information must be derived from each atomic spectrometric measurement in order to learn more about sample speciation and to enable the instrument better to monitor its own operation. These needs and likely trends argue strongly for a higher degree of dimensionality in atomic spectrometric measurements. Higher dimensionality represented in other areas of analytical chemistry by the so-called ‘hyphenated techniques’ such as GC-MS can be achieved in atomic spectrometry by using sources and sample-introduction techniques in tandem (either in series or in parallel) by combining emission MS and AF measurements and by employing multi-dimensional calibration and sample-recognition algorithms.For example it can be * Presented at the 1996 Winter Conference on Plasma Spectrochemistry Fort Lauderdale FL USA January 8-1 3 1996. 1 Journal of I Analytical I Atomic Spectrometry shown that MS resolution greater than 300 000 can be achieved by means of a relatively simple moderate-resolution mass spectrometer as long as it is preceded by suitable sample-introduction apparatus. Further interference effects and sample-utilization efficiency might be drama tically increased by introducing sample solutions in the form of discrete droplets or as puffs of sample vapour. These and other examples taken from the author’s laboratory and from laboratories of others illustrate these various trends and future projections. Keywords Plasma emission spectrometry; plasma mass spectrometry; inductively coupled plasma; sample-introduction efjiciency ; time-of-jlight mass spectrometry; multidimensional methods; chemometrics It is always dangerous to project the future.It is of course a relatively simple matter to track current trends and to extra- polate them. This approach perhaps the safest has been employed in at least one book series’ and has in retrospect demonstrated a fair degree of reliability. Another approach which was taken in some of our own earlier reviews2y3 applies particularly well to the fields of analytical science and chemical instrumentation. It involves defining an ‘ideal’ device instru- ment or technique and determining criteria by which existing systems fall short. It is then a relatively safe bet to assume that the greatest emphasis in the near future will be placed upon overcoming the most serious of the shortcomings. A third technique to forecast the future is to record notable advances in related scientific or technical areas with the assumption being that those advances will eventually be transferred.Unfortunately none of these established means of forecasting is able to deal with a true ‘breakthrough’. By definition true innovation cannot be predicted. For that reason no one fore- saw for example magnetic resonance imaging matrix-assisted laser desorption/ionization in MS or the deuterium-lamp method for background correction in AAS. Still an informed observer might have foreseen the widespread use of AAS as an analytical technique the attractiveness of ICP-AES and the rapid growth of ICP-MS simply by keeping abreast of the physics literature and by listening attentively to lectures at major international symposia.The forecasting exercise therefore seems to be worthwhile. Predictions in atomic spectrometry are complicated by the blurring of disciplinary boundaries that are affecting all of science. For example it is not simple to decide whether electrospray ionization (ESI) coupled with MS is a form of atomic spectrometry or not. To be sure the method can provide the same sort of information as can ICP-MS and additional details as well. However in its present state of development ESI does not provide complete and unambiguous information about the qualitative and quantitative elemental composition of a sample.Indeed the data it provides are more akin to those offered by ion chromatography than by ICP-MS. Journal of Analytical Atomic Spectrometry September 1996 Vol. 11 (61 3-621) 61 3Similarly the increasing emphasis on speciation in atomic spectrometric analysis precludes a clear definition of the boundaries of the field. Ordinarily speciation is considered to include an indication of the oxidation state of an element; however it might also require information about the degree of complexation the number and type of associated atoms and the sort of bonding in which an atom is involved. Does it then require in its ultimate embodiment also a complete definition of the atomic composition and layout of a sample? If so it would seem to be more competitive with X-ray crystallography than with current techniques such as ICP-AES ICP-MS and AAS.If this degree of speciation is to be the supreme goal of atomic spectometry perhaps our attention should be directed more towards secondary ion mass spectrometry or sputtered neutral mass spectrometry than to these traditional methods. Conveniently both these latter powerful (but expensive) methods and traditional ones can be represented by the scheme shown in Fig. 1 which embodies perhaps the ‘ultimate’ method of atomic spectrometric analysis. In this arrangement a solid liquid or gaseous sample is decomposed quantitatively into its constituent atoms. Those atoms are then sorted by type and by isotope and each of the isotopes subsequently counted.In an ICP mass spectrometer equipped with a laser-ablation accessory for example the laser beam and the ICP together provide the atomization a mass spectrometer performs atom (really ion) sorting and a high-sensitivity low-background detector counts the ions individually. Unfortunately in such an instrument atomization is not quantitative spatial resolution is not possible at the atomic level not all atoms are ionized and transmission losses preclude the detection of every atomic ion. Ideally one might wish to employ a ‘Maxwell’s Demon’ in the scheme of Fig. 1. The demon would pluck atoms one by one from the sample of interest recording at the same time the location from which the atom was withdrawn. Each atom would than be sorted (perhaps by mass) and counted.Because the nearest neighbours of every atom would be recorded and because the demon might be able to register information about bonding and oxidation states at the same time the atom is withdrawn a complete description of a sample would be possible. On the other hand some applications might require greater speed than that of which the Maxwell’s Demon is capable. In such a situation the atomization process might resemble more closely the game-starting ‘break‘ that occurs in pocket billiards (pool). Here the atoms of the sample would be simultaneously liberated and scattered each type (element) Atomic ions fl Atomize C. - Mn U B r Count Atomic Ions Fig. 1 In the ultimate atomic spectrometric instrument the sample would be decomposed quantitatively into its constituent atoms with the location of each atom in the sample being recorded along with its oxidation state degree of complexation and nearest neighbours.The isolated atoms would then be separated by type and by isotope and the individual isotopes counted being directed to its own ‘pocket’ and the atoms counted as they arrive. Interestingly the Maxwell’s Demon concept above is one component of the emerging field of nanotechnology. Ordinarily nanotechnology is viewed as involving the assembly of complex macroscopic materials from individual atoms or molecules. The human body for example is an exquisitely refined nanotechnology factory. However nanotechnology can also involve the disassembly of materials a process which some view as a prerequisite for nanotechnological fabrication.In this concept a nanotechnological ‘disassembler’ would analyse (disassemble) a given sample on an atom-by-atom basis while keeping a nanoscopic record of the position and bonding of each atom. This record or programme stored perhaps in molecular form just as is a biological code in the human genome would then be used to construct perfect replicates of the original object or material. This second (fabrication) step would be carried out by a second type of nanotechnological device an ‘assembler’. Perhaps this nanotechnological direction is where atomic spectrometry will eventually head. For the near term however it is probably more prudent to consider whether and how existing methods can be modified or improved. There is already a substantial array of highly capable sources including glow discharges microwave plasmas inductively coupled plasmas electrospray ionization and others. Similarly a host of alter- native detection methods exists including AAS AES atomic mass spectrometry (AMS) and AFS.In most practical applications of these methods sample preparation plays a crucial role. In many cases this preparation step consists of dissolving the sample and adding to it such things as internal standards or ‘spikes’ to enable standard additions to be used. In other cases sample preparation consists of casting a solid into a suitable form and grinding or other modification of its surface. In still others such procedures as fusion etching or even ion implantation are utilized. Not infrequently sample preparation is the rate-limiting step in performing an elemental analysis. Unfortunately despite the importance of preparing a sample the subject cannot be treated here.Indeed an adequate examination of sample- preparation options for the future deserves a review of similar scope by itself. Considerations will therefore be constrained to likely directions that will be taken in the areas of source and spectrometer development and in signal processing. To begin the most significant shortcomings of the commonly used methods for atomic spectrometry will be reviewed briefly. Trends in both atomic spectrometric instrumentation and in instrumentation overall will then be considered in an effort to project what new instrumental advances are likely.Considering detection limits first it will be seen that improvements in AES will depend upon reducing background emission whereas MS will require gains in sample-utilization efficiency and in instrument transmission. Methods by which other figures of merit can be improved will also be addressed. Figures of merit to be emphasized in this discussion include precision and instrument speed. Other trends that seem likely will then be outlined and will emphasize an increase in the dimensionality of atomic spectrometric instrumentation. Added dimensions might derive from the coupling of atomic methods with others (e.g. ‘hyphenation’) or from the use of multiple sources com- bined atom-detection schemes or multi-spectral correlation. Finally alternative strategies for future plasma spectrometric instrumentation will be suggested.SHORTCOMINGS OF CURRENT TECHNIQUES Lists of the most troublesome shortcomings that currently plague ICP-AES ICP-MS and GD techniques are compiled in Tables 1-3. Of course a person’s orientation and field of 61 4 Journal of Analytical Atomic Spectrometry September 1996 Vol. 11Table 1 Most significant current shortcomings of ICP-AES Modest detection limits (pg 1-') Drift and limited precision Matrix interferences Poor semiquantitative performance Limited for non-metals No isotopic information Sample preparation Micro-samples inconvenient Table 2 Most significant current shortcomings of ICP-MS Matrix and spectral interferences Drift and limited precision Difficulty with micro- or transient samples High cost complexity Need for sample preparation Table 3 Most significant shortcomings of glow discharges Modest detection limits Solution samples inconvenient Microsamples impractical Slow (5-30 min per sample) Complex spectrum (AES or MS) Limited for non-metals interest will dictate which entries will appear on lists such as these.However most would agree that in general even better detection capability improved precision freedom from inter- ferences better semiquantitative performance full elemental and isotropic coverage and freedom from sample preparation would be desirable in all atomic spectrometric techniques. Further most would appreciate the capability to examine micro- or solid samples directly and to obtain spatial resolution in solid specimens. Although one technique might at present meet certain of these needs better than others all could stand improvement.Also the very fact that several alternative methods for atomic analysis coexist and compete testifies convincingly that no single method has a decisive advantage over the others. Clearly a major portion of the fundamental and technical effort in atomic spectrometry in the near future is likely to be devoted to overcoming the limitations listed in Tables 1-3. However the degree to which such goals can be reached will depend upon present and future developments in instrumen- tation particularly of the type useful for atomic spectrometry. Some of these trends will be examined in a bit more detail in an effort to decide which might be of particular importance in atomic spectrometry.TRENDS IN INSTRUMENTATION AND IN ATOMIC SPECTROMETRY In Table 4 are listed a number of trends that can be identified in the development of chemical instrumentation. Of course additional trends could be named; here only those have been included that are especially applicable to atomic spectrometry. It is interesting to compare this list with the one compiled in Table 5 which applies particularly to trends in atomic spectro- Table 4 Trends in instrumentation Faster More specialized Smaller (nano) Spatial resolution More sensitive Microsampling More precise More intelligent Lower reagent consumption Greater dimensionality Integrated ('total analysis systems') Table 5 Present trends in atomic spectrometric instrumentation More sensitive More precise Faster (?) Fewer interferences Less expensive Smaller (?) More automated Fewer vendors (less competition) Increased speciation metric instrumentation.The disparity between these tables will form the basis of a discussion in a later section. Few would dispute that atomic spectrometers over the past few years have become considerably more sensitive and precise. Also especially in AMS both matrix and spectral interferences have been reduced in severity although a number of trouble- some effects still exist. Unfortunately interferences can still pose serious problems in ICP-AES. Interestingly there seems to have been little emphasis on making atomic spectrometers operate more rapidly. Most commonly integration times on the order of 10-60 s are employed although longer times are required when a GD source is used.In part this limited speed could be a result of the relatively long wash-out time of most nebulizer-spray chamber combinations. Similarly modern atomic spectrometers are not a great deal smaller than earlier systems. Although manufacturers have introduced tabletop systems for AMS and AES developments in miniaturization that are actively underway in other fields make one wonder whether additional substantial reductions in size are not possible. Largely because of technological advances instruments have however been made simpler in construction less expensive as a consequence and more user-friendly mainly because of a higher level of automation. This increased emphasis on economy and the realities of a limited market have forced a number of vendors out and have resulted in the acquisition of others.Interestingly despite a high degree of interest in and demand for speciation few commerical packages to perform such an analysis are yet available. It would be useful to evaluate each of the trends listed in Tables 4 and 5 to determine which trends are most likely to be important in atomic spectrometry in the future and to track each trend to its eventual conclusion. Unfortunately space precludes such a comprehensive approach. Instead the sensi- tivity issue will be considered first a trend which appears in both Tables 4 and 5 and that seems to form the focus of much recent work in AES and AM$. The issues of precision and instrumental speed will then be discussed briefly. DETECTION LIMITS IN ATOMIC SPECTROMETRY A fairly simple comparison of ICP emission and mass spectrometry signals reveals the factors that are responsible for governing detection capability.Assume for example that a 1 pg ml-l solution of a chosen element is aspirated into each instrument. Experience shows that a typical mass spectrometer will produce a signal level of roughly 106-107 counts s-I while the emission spectrometer will generate a photocurrent of roughly 10-6A. With a photomultiplier gain of lo6 this photocurrent corresponds to a photon detection rate of 6 x lo6 counts s-'. Interestingly the two techniques produce approximately the same signal level! Where they differ is in the background level. Background count rates expected in AMS seldom exceed 10 counts s-l whereas the background photocurrent in emission spectrometry is of the order of lo-* A which corresponds to a photon detection rate of 6 x lo4 counts s-'. Because the background noise is pro- portional to the square root of the background count rate the signal-to-background noise ratio (S/Nb) of the mass spectro- metric measurement will be lo6 while that for the emission Journal of Analytical Atomic Spectrometry September 1996 Vol.I1 61 5determination will be only lo4. This disparity provides ICP-MS with detection limits that are about two orders of magnitude lower than those of€€ered by ICP-AES. In particular for a given S/Nb (usually three) at the detection limit the mass spectrometer would be expected to yield a sub-ngl-' limit compared with the 0.1 pg1-l detection capability of the emission instrument.These values are roughly what is experienced. Overall then it is the background level rather than the signal magnitude that is responsible for the difference in detection capability of the two techniques. The corollary is similarly clear that MS detection limits can be improved mainly by raising the fraction of atoms in the sample that are eventually detected; in contrast the emission spectrometer could benefit by the same advances but also by lowering the spectral background. Achieving Lower Detection Limits in Plasma Source AES From the foregoing analysis it seems that the most direct route to improving detection limits in AES is to reduce the background level. Of course it is just this end that is sought in the use of on-axis viewing of an ICP.The 'throat' of the discharge is both cooler and less intense than the outer or toroidal region of the plasma. Furthermore end-on viewing enables a greater depth of the centre of the plasma to be viewed so localized instabilities can be averaged out. The result is a lower more stable background reading. Unfortunately with this reduction in background comes a concomitant increase in the severity of inter-element inter- ferences and the likelihood of complications induced by intact aerosol droplets or particles that persist into regions high in the discharge. An alternative means of reducing background intensites in plasma emission spectrometry is to operate the discharge at reduced pressure. Because the recombination continuum in any plasma results from a two-body process (the recombination of an ion and an electron) the frequency of the recombination events and the intensity of the continuum they produce should drop with the square of the ambient pressure.In contrast analyte number densities should drop only linearly with pressure so a gain in S/Nb should be achieved. Further reduced-pressure plasmas are ordinarily more diffuse and stable than those at atmospheric pressure so background fluctuations should be even smaller. Unfortunately when the pressure in a plasma is reduced its ability to volatilize and atomize particulate aerosols or polyatomic species is com- promised rendering the discharge more complicated to use and more susceptible to matrix interferences. It would seem more appropriate therefore to use a reduced- pressure discharge for atomic excitation (or ionization) but to use an auxiliary source to atomize the sample.This 'tandem- source' concept is one that has been discussed in detail earlier4 and does not require elaboration here. Suffice it to say that several alternative tandem sources are attractive. The first source of such a tandem pair could alternatively be one that employs rare-gas sputtering wall volatilization or an atmospheric- pressure plasma for sample atomization. The resulting atoms can then be fed directly (in the case of wall volatilization or rare-gas sputtering) or through a differentially pumped interface (in the case of an atmospheric-pressure atomization source) into a reduced-pressure source for excitation or ionization.Additional tandem-source combinations have been suggested ele~ewhere.~.~ Attainable Detection Limits in Plasma Source Mass Spectrometry As was discussed above improving detection limits in plasma source mass spectrometry is probably achievable only by increasing the fraction of ions that are ultimately detected; background count rates appear unlikely to be lowered signifi- cantly. Accordingly let us first follow the fate of atoms in a sample as they are converted into free atoms in the plasma ionized extracted into the mass-spectrometer interface trans- mitted through the spectrometer and finally detected. Following these steps will allow an assessment of where ion losses occur and which of those loss mechanisms is likely to be improved in the future.In general this discussion will be applicable to most mass spectrometers that are in current use (particularly those that employ a quadrupole mass filter) although specific details will pertain to a time-of-flight mass spectrometer (TOFMS) that has been developed in this laboratory.6-10 In an earlier evaluation," the losses in analyte-ion popu- lations that occur as a sample species moves from the initial solution to the detector in an atomic mass spectrometer were traced. In brief it was found that a 100-fold loss occurs in the aerosol-generation process because of the inefficiency of most nebulizer-spray chamber combinations. However virtually all of the resulting aerosol that enters the plasma and all of the sample vapour it produces ultimately is extracted into the first stage of the vacuum interface to the mass spectrometer.In contrast only about 1 % of that extracted beam passed through the skimmer largely because of geometric (solid angle) con- siderations. For a 1 mg 1-1 analyte-atom concentration in solution delivered at a flow rate of 1 ml min-l into a nebulizer the resulting flow of analyte ions through the skimmer therefore corresponds to roughly lolo analyte atoms s-'. For atoms of moderate ionization energy most will be charged. Yet only about one in lo4 of them is eventually detected in most quadrupole-based instruments. Of course recent advances in interface design have raised the efficiency of these processes somewhat but usually by only an order of magnitude or so. The overall consequence is that only between lo6 and lo7 analyte-ion counts s-l are registered at the dectector even though the initial analyte-atom flow into the nebulizer approxi- mates 1014 atoms s-'.The losses in analyte-atom flux corre- sponding to factors of between lo7 and lo8 can surely be improved. A more detailed analysis of these losses is possible with data recently acquired from the TOFMS instrument in this laboratory. In that instrument it can be assumed that the sample-introduction efficiency constrained mostly by losses in the aerosol-generation process is approximately 1 YO similar to that in most other ICP-MS instruments. Also the efficiency of transmission through the skimmer is unlikely to be better and can therefore be estimated to be 1%. This instrument employs an orthogonal-extraction geometry so transmission losses in both perpendicular sections of the ion beam must be considered.It is estimated that the transmission efficiency of the primary beam is about lo% and measurements have shown7 that the fraction of the remaining ions that are detected after transiting the TOFMS flight tube is of similar magnitude. In addition the current instrument still suffers somewhat from a duty factor which is less than unity; in particular only a little more than 10% of the ions in the primary beam are able to be extracted and ultimately detected because of the finite time it requires for a TOF mass spectrum to be recorded. These estimated losses are summarized in Table 6. Table 6 Losses in an ICP-TOFMS instrument Sample-introduction efficiency = lo-' Skimming efficiency = Ion-optic throughput (estimated) = 10- ' TOF transmission efficiency = lo-' TOF duty factor = lo-' Total losses accounted for = lo-' 61 6 Journal of Analytical Atomic Spectrometry September 1996 Vol.1 IIt is useful to compare the total accumulated losses in Table6 a factor of lo7 with detection limits that have so far been realized with the same instrument. Interestingly detection limits all fall near the level of 107-108 atoms for most sample- introduction schemes that have been explored with the ICP-TOFMS. These methods include the use of ten input ion pulses into the TOFMS instrument from a continuously aspirated sample solution,8 a flow-injection plug and electro- thermal volatilization into the ICP. These values are compiled in Table 7.Because detection limits are approximately at the level of lo7 atoms and because losses of about the same order of magnitude can be ascribed in total to instrumental components or sections virtually all losses in the instrument can be accounted for. It is therefore relatively straightforward to assess in which of these areas gains are likely. Improvements in Sample-utilization Efficiency Advances in sample-utilization efficiency should benefit not only plasma source mass spectrometry but also emission-based techniques. Several attractive possibilities exist including the use of in situ laser ablation,13 more efficient electrothermal vaporization a glass-frit14 or higher-efficiency ultrasonic nebulizer nebulizer starvation and introduction of the sample solution in the form of discrete isolated droplet^'^.'^ or dried monodisperse particle^.'^*'^ Liu and Horlick have shown that signal increases as high as 1000-fold are possible in ICP-AES through use of in situ laser ab1ati0n.I~ In this technique a solid sample is placed high in the ICP torch just below the plasma ‘fireball’.An ablating laser beam focused onto the surface of the sample then generates a cloud of atomic vapour immediately upstream of the discharge. As a consequence sample-transport efficiency into the plasma is virtually loo% yielding an extremely intense but very brief emission signal. Importantly because this brief signal pulse lasts less than 1 ms it is attractive only if coupled with a multichannel detection system such as a direct- reading emission spectrometer an emission spectrograph or a TOFMS instrument.With such instrumentation increases in signal levels up to a factor of lo3 should be possible. Although a number of high-efficiency approaches to nebulization have been introduced and evaluated it seems likely that their higher efficiency could derive as much from a low aerosol density as from any intrinsic properties of the nebulizer of spray-chamber design. Recent work from the research group of Olesiklg has shown that even a conventional glass-concentric nebulizer and a Scott-type spray chamber are capable of delivering sample-introduction efficiencies approaching 90% merely by restricting the sample-solution flow rate into the nebulizer. In part the increased efficiency might derive directly from the production of a finer aerosol which is known to occur under conditions of nebulizer ‘starvation’.However it could also be a result of a greatly reduced aerosol density and a consequent reduction in droplet- droplet collisions. In turn fewer droplet-droplet collisions will cause less coalescence and the resulting loss of large droplets. Table 7 Current limits of detection (LOD) in ICP-TOFMS Importantly this higher efficiency requires extremely low sample-solution flow rates approaching a few pl min-’. As a result the technique is particularly attractive when sample volumes are precious or when it is desired to couple the detection method with for example microbore liquid chroma- tography or chip-based sample-processing systems. These latter systems will be described in more detail later.Virtually 100% of a sample solution can be utilized if it is introduced into a plasma in the form of discrete or dried particle^.'^.'^ Even in the earliest studies in which flame emission spectrometry was employed,15 detection limits as low as 10’ atoms of sodium were obtained and precision levels as high as 0.01% RSD could be achieved by integrating the signal from multiple droplets. Achieving even better detection capability should be possible by using the same system or its dried-particle with an ICP and mass spectrometer. Of course this technique (as with in situ laser ablation) produces a transient sample pulse and will be best coupled with a multichannel emission spectrometer or a rapid-scanning mass spectrometer such as a TOF system.Higher Skimming Efficiency Although skimming efficiency is constrained largely by the geometry of the conventional ICP-MS interface improvements are possible if a low-density reduced-pressure ion source is located in the first vacuum stage of the mass spectrometer. Ion sources can be conceived that maintain most ions on axis and therefore should allow the ions to be more efficiently extracted into the lower-pressure zones of the interface. Of course such a source would not be particularly appropriate for atomizing a sample and would therefore probably be most useful if coupled with a preliminary atomization source. Such a tandem combination was described earlier. Improving Ion-optic Throughput Recent studies by Douglas2’ have shown that the transmission efficiency of an ion-optic train can be improved merely by forming a ‘brighter’ ion source and by reducing the ion density in the beam.In this context a ‘bright’ source implies that the ions must be introduced into the optical system in a very narrow cone and appearing to derive from a very small region in space. Such conditions could be met by a suitably designed reduced-pressure source such as the one just described. Furthermore reductions in ion-beam density would be possible through use of sample-introduction systems that minimize the sample-ion flux but which utilize the available sample ions more efficiently. A ‘starved’ nebulizer system would be one system appropriate for this task. If these sample ions are then carried in a relatively low-density plasma-ion beam trans- mission efficiency could approach unity in the primary ion beam in our TOFMS system.Similarly the transmission of our TOFMS instrument itself should be able to be improved by roughly 10-fold simply by reducing the spread of the ion beam as it travels down the flight tube.7 Indeed most workers in the field feel that a Sample introduction mode Nebulizer multi-element* Nebulizer 10 pulses ETV multi-element Sample volume 170 pl 25 nl 10 ml LODt (atoms) 8 x lo’ ( 5 x 107) 2 x 107 3 x 109 (5 x 106) LODt/mol I-’ (8 x 10-14) 6 x lo-’’ 8 x 3 x 10-9 (2 x 10-13) * Signal-integration time of 10 s taken in flow-injection mode. t Values in parentheses are intrinsic detection limits taken from ref. 12. Journal of Analytical Atomic Spectrometry September 1996 Vol.11 61 7TOFMS instrument should be capable of transmitting and detecting virtually all the ions introduced into it. Also plans are already underway in this laboratory to improve the duty factor of our system. A trivial method which will result at least in a doubling of the duty factor is to employ a shorter flight tube. At present the resolving power offered by the instrument is in excess of what is required for routine AMS. Shortening the flight tube would therefore not result in an unacceptable loss in resolution but will offer both increased flight-tube transmission and a shortened time for acquiring a mass spectrum. Because more mass spectra can be acquired each second the duty factor will increase proportionally. Ultimate Levels of Detection in Plasma Source Mass Spectrometry With all these possibilities for improvement it seems reasonable to expect gains in detection efficiency of analyte ions in the relatively near future of a factor between lo4 and lo5.The resulting detection limits should then approach 100-1000 atoms in the sample to be analysed. It is reasonable to question whether such extraordinary detection capabilities are necessary or even desirable in the real world. After all the likelihood of sample contamination increases as detection limits drop; furthermore at some level virtually every element can be found in every sample solution. However it must be recalled that these detection limits have been cited in terms of the number of detectable atoms (or moles). Therefore they could be exploited either to measure vanishingly low elemental concentrations in a sample of moderate volume (or size) or to determine moderate concen- trations in an extremely tiny sample.The latter option might prove to be the more attractive. The ability to use extremely small sample volumes opens the way to a number of novel and attractive sample-processing alternatives. Many of these options including on-chip sample processing microchromo- tography and others all offer increased speed of analysis smaller equipment closed flow channels to prevent con- tamination and the conservation of chemical reagents a matter that is likely to be of increasing concern in the next century. Of course reduced sample requirements also open the field to the analysis of a greater range of sample types including those from the clinical environment those resulting from biotechnology efforts and those in the nanotechnology arena.IMPROVED PRECISION I N ATOMIC SPECTROMETRY The dominant source of imprecision in plasma source emission or mass spectrometry ordinarily arises from the plasma itself and from the sample-introduction equipment. Peristaltic-pump pulsations temperature-induced drift of the nebulizer and spray chamber plasma tail-flame waver the presence of intact aerosol droplets or solute particles in the discharge and in the case of MS measurements inhomogeneities in the plasma volume that is sampled all constrain precision levels to between 1 and 5% RSD. However because all these sources of fluctuation affect signals from different analyte species in much the same way a considerable improvement in precision can be obtained by a ratioing technique such as internal standardization.Clearly for this ratioing or normalization to work as well as possible the spectroscopic physical and chemical features of the analyte element and its internal standard must be matched as closely as possible. When MS is employed this match can be virtually perfect if an isotope of the analyte element can be employed as the internal standard. This method embodied in isotope dilution techniques can result in tremendous gains in precision. Unfortunately these gains are possible only if the analyte species and its internal standard (or isotope) are measured at the same time. With a sequentially scanned system the tem- poral offset that exists between the measurement of the analyte and internal-standard signals can render the ratioing process imperfect so compensation for relatively rapid fluctuations is impossible.To be sure peak hopping in plasma source mass spectrometry can be extremely fast so precision levels can be improved considerably. However the number of such peak hops that can be achieved in a given measurement interval constrains the number of analyte-internal standard pairs that can be measured at once. The problem is particularly acute when a transient sample is introduced such as that which would be generated by several of the sample-introduction methods just suggested. The conclusion for the future must be that simultaneous or virtually simultaneous instruments will become most attractive.Again such systems include two-dimensional detector arrays for emission spectrometry and any of several alternatives for MS. These alternatives include TOFMS an ion trap a Fourier- transform mass spectrometer and a sector-field system equipped with a multichannel detector array. Unfortunately none of these MS alternatives is yet commercially available. Among them the FTMS seems least attractive in part because of its cost and in part because of the relatively low ion densities that it can contain. This latter complication which the FTMS shares with the ion trap might restrict dynamic range to unaccept- able levels. PROBABLE TRENDS IN ATOMIC SPECTROMERIC INSTRUMENTATION A comparison of Tables 4 and 5 reveals that a number of trends that exists in the development of chemical instrumentation overall are not yet being actively pursued in the field of atomic spectrometry.These missing areas are highlighted in Table 8. Several of the trends listed in Table 8 require no elaboration. However it seems clear that atomic spectrometric instruments like most others will become targeted towards increasingly specialized markets in the future. Because of this trend a number of instrumental requirements will be relaxed. It is of course difficult to design an instrument that is capable of meeting the needs of all users. If a particular class of users strives for a certain group of elements (such as in the metals industry) expects a constrained range of concentrations (as in process-control applications) or must deal with only a single sample form (i.e.solutions) lower-cost special-purpose instru- mentation can be developed. However for this approach to be attractive to vendors the market for a particular application must be sufficient. Accompanying a more specialized or targeted market will come a higher level of integration of the entire analytical operation. A component of this integration will be coupling the output of atomic spectrometric instruments with data provided by complementary analytical methods so the output of an analytical laboratory will be more akin to the solution of a specific problem rather than a mere list of concentrations. Similarly it seems likely that the marriage between sample processing and the atomic spectrometric instrument is likely to be a closer one with particular sample-processing approaches being used for specific customers markets or sample types.Table 8 Probable trends in atomic spectrometric instrumentation More specialized Greater dimensionality Spatial resolution Microsampling Integrated (‘total analysis systems’) Lower reagent consumption 61 8 Journal of Analytical Atomic Spectrometry September I996 Vol. 11The need for spatial resolution and microsampling will similarly be restricted to particular markets or users that are likely to grow in importance in the future. Also the cost of chemical-waste disposal will undoubtedly increase and will encourage conservation of reagents. Integrated-analysis micro- sampling systems will therefore probably grow rapidly in importance and might be implemented by means of the on-chip devices mentioned earlier and discussed in more detail below.A particularly interesting aspect of instrumentation develop- ment outside the field of atomic spectrometry is a trend towards higher dimensionality. This trend represented most obviously by the ‘hyphenation’ that is associated with combi- nations such as GC-MS is leading to measurements that are extraordinarily information-rich. It would seem that multi- dimensionality might be useful in atomic spectrometry also and could provide such advantages as added selectivity lower levels of interference improved speciation sampling con- venience applicability to alternative sample types better pre- cision and detection limits and added confidence in sample identification.Overall multidimensionality should provide more information capability and flexibility. Multidimensional Atomic Spectrometry The components that are common to all atomic spectrometric instruments are portrayed in Fig. 2. Interestingly any of these components can be made multidimensional by employing alternative embodiments of it in either a parallel or serial fashion. For example the sample-processor module could consist alternatively of an aerosol-introduction system an FI device a chromatograph a laser-ablation accessory or others. Moreover in highly flexible instruments several of these modules could be employed in parallel to increase the flexibility and capability of the system. However several of these alternative sample-processing modules produce transient peaks many of which have a duration of less than 1 ms.For the rest of the atomic spectro- metric instrument to accommodate such sample-processing units it must respond rapidly and ideally simultaneously to all elemental concentrations in the sample. In the absence of this capability precision and detection limits will be sacrificed in a multi-element mode. A trade-off will then have to be made between broad elemental coverage or low detection limits and high precision. As was mentioned before the need for this type of flexibility highlights the importance of truly multichannel emission or mass spectrometers. The second module in Fig. 2 the source can also be multidimensional. For example it could alternatively be an ICP a GD an MIP a high-voltage spark or another source.Indeed there have already been commercial offerings in which an ICP or GD could be coupled to the interface of a mass spectrometer. However the tandem-source approach that was already mentioned is another form of multidimensional source that is particularly attractive for the next generation of atomic spectrometric instruments. If properly configured such tandem combinations could yield truly multidimensional information. Just as the combination of GC and MS renders sample identification much simpler by separating and identifying components along two orthogonal (ie. independent) axes Fig.2 Block diagram of an atomic spectrometric instrument. In future-generation instruments each of the blocks would be modular with several alternative devices or schemes for each block being able to operate in parallel or in series.The resulting higher-order instrument should provide far more capability and flexibility than is now available chromatrography and mass spectrometry a tandem source can improve instrument performance. An example of this capability can be found in recent results obtained with our TOFMS system in which the source con- sisted of a tandem combination of an electrothermal vaporizer and an ICP. In this arrangement sample solutions are loaded in the conventional way into a graphite furnace and the solvent driven off from the sample deposit. The sample is then ashed briefly and the temperature of the furnace subsequently ramped to a high volatilization temperature. As has long been experi- enced in AAS it was found that different elements volatilize at distinct times on the temperature ramp.During this atomization process TOF mass spectra were acquired in rapid succession (at a spectral-generation rate of approximately 20 kHz). The result was a two-dimensional map in which the vertical axis consisted of atomization time (related to atomization temperature) and the horizontal axis displayed the mass spectrum. In this two-dimensional map relatively volatile elements such as cadmium appear earlier than those that are more refractory. As a result it became possible to separate otherwise isobaric interferences such as the 112 and 114 isotopes of cadmium and tin. Interestingly to resolve these isotopes mass spectrometrically would require a resolving power of 93 000 for the 112 isotopes and 190 000 for the 114 isotopes.Yet the separation was achieved by this two-dimen- sional technique with a mass spectrometer whose nominal resolving power was less than 2000. Even greater capability is realized by separation on the basis of volatility differences of the 113 isotopes of indium and cadmium. The mass-spectral resolving power required to separate these two species exceeds 300 000. Perhaps a more real-world application of this capability is the separation of ArO’ and Fe’ both of which appear at m/z 56. In most AMS determinations this overlap constitutes one of the most troublesome interferences. However it can be avoided entirely by driving the solvent completely from the sample before the iron is volatilized. Obviously a great many other interferences can be similarly avoided.Other benefits can also be derived from the use of a tandem source especially when it is combined with MS detection. For example either of the two sources in the tandem pair can be modular. Probably the most attractive scheme is to employ the same ionization source (perhaps in a reduced-pressure environment) but to couple to it a group of first sources each of which is tailored for a particular sample type. There could be one first source for the atomization of aerosol-based samples another for solid samples a third for providing spatial resolution in solid samples another for FI applications one for a chromatographic interface a device for handling micro- samples etc. Further the first source in the tandem pair could be modulated so it produces on alternative half cycles fully atomized sample material and merely fragmented sample vapour both of which would pass in alternating fashion into the second (ionization) source.Such a device would yield on one half cycle an atomic mass spectrum and on the other a fragmentation mass spectrum. If these spectra could be generated rapidly enough through use of a sufficiently fast mass analyser such as in TOFMS more unambiguous infor- mation about an entering sample would be available. Such a device would aid greatly for example the identification of eluting constituents from a liquid chromatograph. Multidimensionality is already fairly common in the atom- detection module of an atomic spectrometer. It is routine for example to employ a slew-scan monochromator as an ‘n + 1’ channel in a direct-reading emission system.Also at least one manufacturer has already announced an instrument that couples emission and mass spectrometric detection. An additional scheme that has not been pursued actively however is the combination of MS and fluorimetric detection. Journal of Analytical Atomic Spectrometry September 1996 Vol. 11 61 9This last combination could take several forms but its main advantage would be the ability to provide an extremely high detection efficiency for every analyte atom or ion. Conventional MS detectors are inherently destructive so there is only one chance to detect each ion. As a result it is impossible to distinguish between a single ion and an extraneous detector event.Although such events are relatively infrequent (1-10 per second in a typical instrument) they preclude the detection of less than ten ions or so. In contrast fluorescence is a non- destructive process and each atomic ion is capable of pro- ducing hundreds of millions of detector events per second. The result is truly single-atom detection. One possible combination of fluorimetric detection and MS would be similar in design to an earlier commerical instrument. However in the new arrangement the atoms or ions to be detected would be extracted through a vacuum interface similar to that employed in a modern atomic mass spectrometer. Each extracted ion would then pass through a region irradiated by a high-intensity laser beam. Although some background counts from scattered laser radiation would no doubt exist the passage of an atomic species would be signaled by a brief burst of additional photons.This ‘photon-burst’ mode of detection2’ provides not only the ability to detect single atoms but also a much higher degree of selectivity than could MS by itself. Because it is a truly multidimensional approach MS resolution could be coupled with optical resolution to achieve a much higher degree of selectivity than would otherwise be available. It has been estimated that resolving powers as high as 1014 could be achieved in this fashion.21 Of course such a scheme would compromise multi-element detection unless an array of simultaneous laser beams could be employed. Indeed such capabilities would not seem too far beyond the horizon in view of recent advances that have been made in the development of blue and frequency-doubled diode lasers.Lastly the signal-processing block in the diagram of Fig. 2 might also be multidimensional.2 Likely multidimensional schemes include multi-line and regression-based calibration such as has already been used by a number of commercial manufacturers. However new means of displaying and pro- cessing atomic spectrometric information will no doubt be devised. Similarly element ‘profiles’ of an incoming sample will no doubt be generated; coupling such information with data from other methods will probably be critical in developing special-purpose laboratories that are intended to provide answers to analytical questions rather than simple uninterpreted data.A GLIMPSE OF THE POSSIBLE FUTURE The preceding sections cover relatively straightforward extra- polations of current trends in atomic spectrometry. However is seems appropriate also to cast a bit farther afield to see what the future might bring. For some time physicists have employed crossed laser beams as optical ‘traps’ and optical ‘tweezers’ to manipulate particles in sizes ranging from bacteria tomm spheres. The same type of technology might be profitably applied to analytical atomic spectrometry. For example it is possible to dispense a single microdroplet of sample solution having a volume of less than 1 n1.22 Such a microvolume might be directed into an insoluble liquid with a refractive index different from that of the micro- volume itself. In this medium the droplet could be manipulated by use of laser tweezers and introduced into a selected sample- processing device.Alternatively such a droplet might initially contain not the sample solution but rather a chelating reagent that serves to preconcentrate analyte species from the sur- rounding volume. Similarly a single microscopic ion-exchange particle could be introduced into a larger volume of sample solution and the preconcentrated sample so generated used in a subsequent sample-processing device. To handle such microscopic samples it would be attractive to employ some of the on-chip t e ~ h n o l o g y ~ ~ - ~ ~ that has been used recently in FI analysis in capillary electrophoresis and in sample processing itself. In such devices microchannels are etched into a suitable substrate most commonly silicon or quartz.Electro-osmosis or electrophoresis is then employed to pump samples reagents or diluents about the chip. Injection of a sample plug is possible for example by crossing a sample- containing channel with one filled with a carrier. Because the electrokinetic pump can be switched merely by the application and removal of high voltages a sample plug can be positioned in a desired location (at the junction between the two channels) and carried off in a different direction (if desired) by the carrier stream. In a similar fashion reagents can be added to the sample solution it could be diluted internal standards or standard-additions aliquots could be injected and serial dilution could be implemented. The sample so processed could then be introduced into a multidimensional atomic spectrometer. Lastly is seems likely that future generations of atomic spectrometric instrumentation will become more and more intelligent2 Efforts have been underway for some time to clarify the origins of interference effects in sources such as the ICP.With this information and with knowledge of how an interferent affects the spatial distribution of analyte in a discharge and the characteristics of the discharge itself it should be possible to design an instrument that is truly self- diagnostic one that monitors its own output by means of two- dimensional images spectral information and other sources of data and to feed back control signals to its various inputs. This feedback would be intended both to overcome inter- ferences and to achieve the highest possible signal-to-noise ratios the first time a sample is introduced.Accessible inputs to such an intelligent instrument would include the sample- solution concentration reagents that could be added to it gas flow rates source power alternative methods for sample processing spectrometer resolution and dwell time and detector characteristics. Information available to achieve the necessary feedback would include not only spatial and spectral infor- mation but also details about the sample and its processor that are not now being used. Obviously with these many alternatives available the field of atomic spectrometry holds great promise for the future. The coming of the new millennium should bring even greater challenges and opportunities for its users and students.This work was supported in part by the National Science Foundation through grant CHE 90-20631 and by the National Institute of Health through grant lROl GM 53560. REFERENCES Naisbitt J. and Aburdene P. Megatrends 2000 William Morrow New York 1990. Hieftje G. M. Spectrochim. Acta (Special Supplement) 1989 44 113. Hieftje G. M. Fresenius’ J. Anal. Chem. 1990 337 528. Borer M. W. and Hieftje G. M. Spectrochim. Acta Rev. 1991 14 463. Borer M. W. and Hieftje G. M. J. Anal. At. Spectrom. 1993 8 339. Myers D. P. and Hieftje G. M. Microchem. J. 1993 48 259. Myers D. P. Li G. Yang P. and Hieftje G. M. J. Am. SOC. Mass. Spectrom. 1994 5 1008. Myers D. P. Li G. Mahoney P. P. and Hieftje G. M. J. Am. SOC. Mass. Spectrom. 1995 6 400. Myers D. P. Li G. Mahoney P. P. and Hieftje G. M. J. Am. SOC. Mass. Spectrom. 1995 6 411. 620 Journal of Analytical Atomic Spectrometry September 1996 Vol. 1 110 11 12 13 14 15 16 17 18 19 20 21 Myers D. P. Mahoney P. P. Li G. and Hieftje G. M. J. Am. SOC. Mass. Spectrom. 1995 6 920. Hieftje G. M. J. Anal. At. Spectrom. 1992 7 783. Falk H. Spectrochim. Acta Part B 1994 49 1373. Liu X. R. and Horlick G. Spectrochim. Acta Part B 1995 50 537. Layman L. R. and Lichte F. E. Anal. Chem. 1982 54 638. Hieftje G. M. and Malmstadt H. V. Anal. Chem. 1969 41 1735. Bastiaans G. J. and Hieftje G. M. Anal. Chem. 1973 45 1994. French J. B. Etkin B. and Jong R. Anal. Chem. 1994 66 685. Olesik J. W. and Hobbs S. E. Anal. Chem. 1994 66 3371. Olesik J. W. Ohio State University personal communication 1996. Douglas D. J. presented at the 1995 Chemical Congress of Pacific Basin Societies Honolulu HI USA December 17-22 1995 paper ANYL 226. Keller R. A. Los Alamos National Laboratory personal communication 1994. 22 Shabushnig J. G. and Hieftje G. M. Anal. Chim. Acta 1981 126 167. 23 Ramsey J. M. presented at the 1995 Chemical Congress of Pacific Basin Societies Honolulu HI USA December 17-22 1995 paper ANYL 234. 24 Manz A. Verpoorte E. Busch M. Malone M. Erbacher C. Spielmann A. and Widmer H. M. presented at the 1995 Chemical Congress of Pacijic Basin Societies Honolulu HI USA December 17-22 1995 paper ANYL 38. 25 Harrison D. J. Chiem N. Tang T. and Fluri K. presented at the 1995 Chemical Congress of Pacifc Basin Societies Honolulu HI USA December 17-22 1995 paper ANYL 40. Paper 6/003830 Received January 17 1996 Accepted March 28 1996 Journal of Analytical Atomic Spectrometry September 1996 Vol. 11 621