JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1992 VOL. 7 783 Towards the Next Generation of Plasma Source Mass Spectrometers* Plenary Lecture Gary M. Hieftje Indiana University Department of Chemistry Bloomington IN 4 7405 USA In this paper the present state of plasma source mass spectrometry is reviewed with special emphasis placed on the strengths and weaknesses of currently available systems. Attention is then directed towards basic and applied studies that are underway throughout the world to reduce the remaining shortcomings of the technique. Such efforts include the design and evaluation of novel sources and mass spectrometers modifications in ion- optic and interface systems attempts to understand and overcome isobaric interferences (spectral overlaps from oxides and other polyatomic species) techniques for stabilizing plasma sources and mass spectrometers in an effort to reduce instrumental drift and methods for improving precision both in routine analysis and isotope determination situations. It is argued that plasma source mass spectrometers of the future might be simpler yet more powerful than those now in use.Vacuum pump requirements might be lessened tandem sources will offer new flexibility and capability and simultaneously reading mass spectrometers will speed analyses make interfacing to chromatography devices more practicable and improve precision. Keywords Plasma source mass spectrometry; inductively coupled plasma mass spectrometry; tandem source; instrumentation Plasma source mass spectrometry (PSMS) particularly in the guises of inductively coupled plasma mass spectrometry (ICP-MS) and glow discharge mass spectrometry (GDMS) has become a widely accepted tool for elemental analysis.There are already several hundred commercial ICP-MS instruments alone that are being used in applications that span the spectrum of those customary in elemental analysis. These commercial instruments and those that have been assembled or modified in academic and governmental laboratories have already undergone several stages of evolution and have incorporated newer vacuum-pumping systems ion-optic configurations detectors and interface designs. The most recent commercial offerings provide phenomenal detection limits approaching the parts-per- quadrillion level for one instrument broad linear working ranges excellent performance in a semiquantitative mode high abundance sensitivity and a high degree of user friendliness.It is therefore appropriate to question whether dramatic new developments in PSMS are likely to occur. Will the fundamental studies being carried out in many laboratories throughout the world yield any findings of direct benefit to the practising analyst? Can new plasma sources ion-optic configurations detector arrangements or mass spectrometer alternatives improve performance markedly make the instruments easier to use or more maintenance-free or render them less susceptible to isobaric overlaps and sample-matrix interferences? To answer this question the strengths and weaknesses of existing PSMS systems are considered in an effort to discover where improvement is possible and where it is most needed and a number of current trends and areas of intense study are examined in order to discern the direction that future developments might take.Finally with these considerations in mind it might be possible for us to project how future PSMS designs might look and what they might provide. Sensitivity and Detection Limits in PSMS Low detection limits are the feature of modern PSMS instruments that is most frequently cited as a dominant *Presented at the 1992 Winter Conference on Plasma Spectro- chemistry San Diego CA USA January 6- 1 I 1992. strength. It is therefore useful to examine what factors of the instrument performance are responsible for such low detectable levels and what might be done to improve them further. To begin the efficiency of the ICP as an ion source is examined to determine whether it would be appropriate to consider supplanting it with an alternative. A typical ICP employed for mass spectrometry will have the general configuration depicted in Fig.1. Here a pneumatic nebulizer is coupled to a traditional Scott-type spray chamber and the resulting tertiary aerosol is fed into the central tube of an ICP. Ions ultimately generated in the plasma tail flame are then extracted into the first stage of a mass spectrometer interface via a sampling orifice. Although some arrangements might utilize an ultrasonic nebulizer or some other alternative to the pneumatic system and although temperature-regulated spray cham- bers and desolvation systems are frequently employed the following considerations should remain generally valid.10" analy\e atoms s \+/ Sampling cone 1018 argon3 / atoms cm 0 0 1 ppm at 1 cm3 min ' =10"atoms s I 10' analyt atoms cm Fig. 1 Evaluation of the characteristics of a typical sample introduction system shows that most or all of the analyte sent into an ICP finds its way into the first stage of the ICP-MS interface784 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1992 VOL. 7 Suppose that a 1 ppm solution of a chosen analyte element is aspirated into the pneumatic nebulizer at a sample- solution flow rate of 1 cm3 min-' and carried into the central tube of the ICP torch by a carrier flow of argon at 1 dm3 rnin-'. If it is also assumed that the aerosol exiting the spray chamber comprises approximately 1 O/o of the solution being aspirated a typical value and if it is posited that only moderate dilution of the resulting atoms takes place in the ICP tail flame it can be calculated that the tail flame will contain approximately 1 x l O9 analyte atoms ~ m - ~ .Further- more it can be shown that the atmospheric-pressure tail flame of the ICP will contain about 1 x loi8 argon atoms ~ m - ~ if the ICP gas temperature is 6000 K. Thus if the initial concentration of the solution of the analyte element is 1 ppm the ratio of analyte atoms to argon atoms in the ICP tail flame is approximately 1 x Moreover because there are approximately 1 x loi5 argon ions in the ICP the ratio of analyte atoms to argon ions will be This same value will pertain to the ratio of analyte ions to argon ions for most elements since ionization for many species is virtually complete. These species contained within the ICP are then ex- tracted into the moderate-pressure (typically 133 Pa) region of the mass spectrometer first stage and form there a supersonic expansion.Straightforward calculations' show that the gas flow through this orifice will consist of approximately 1 x lo2' argon atoms s-l a flow correspond- ing roughly to 3 dm3 min-l at standard temperature and pressure. Because the ratio of analyte species to argon atoms should not be altered by this extraction process this sampled beam will contain a flux of approximately 1 x loi2 analyte atoms (or ions) s-' as indicated in Fig. 1. These simple considerations lead to an extremely signifi- cant conclusion. As shown in Fig.1 a sample flow of 1 cm3 min-l containing 1 ppm of analyte corresponds to a flux of 1 x 1014 analyte atoms s-' entering the pneumatic nebulizer; the 1 O/o efficiency of the nebulizer which has been assumed drops this transport to 1 x 10I2 atoms s-l. In other words virtually every analyte atom that enters the ICP will eventually be extracted into the first stage of the ICP-MS interface. The mass spectrometer and its interface are now consi- dered in order to assess in additional detail where signal losses might arise. As revealed in Fig. 2 the supersonic expansion formed by the sampled plasma gases is skimmed through a second orifice (in the 'skimming cone') with relatively low efficiency. The modelled ICP tail flame at atmospheric pressure (Po) contains about 1 x lo9 analyte atoms and the extracted supersonic beam carries a flux of approximately 1 x 1021 argon atoms s-' and 1 x 10I2 analyte atoms s-l as outlined in Fig.1. However calcula- tions' show that only 1% of the species in that beam enter the second stage of the mass spectrometer. That is the analyte flux into the second stage will consist of approxi- mately 1 x 1O'O analyte atoms (or ions) s-I carried with about 1 x 1019 argon atoms s-' and about 1 x loi6 argon ions s-l. Yet even this moderate ion throughput suffers tremendous loss when it is compared with the final signal levels that are realized in typical ICP-MS instruments. It is common experience that an analyte concentration of 1 ppm will produce a final MS signal of approximately 1 x lo6 counts s-l as listed in Fig.2. Thus if most of the analyte species that are sent into the ICP-MS interface are ionized only about 1 in lo4 of those present in the second stage of the interface is ever detected and only about 1 in every 1 x lo6 in the ICP survives the extraction ion-separation and detection processes. Clearly substantial inefficiencies result from skimming the supersonic expansion and in passing that skimmed volume into the third stage of the interface through the ion optics and mass spectrometer and producing a detectable signal. ..-.+# Space-charge ,.,._ dominated region Sampling cone 1 ppm or 10' analyte atoms cm-3 Po Fig. 2 Illustration of the sources of inefficiency in a typical ICP- MS interface.The greatest losses occur in the skimming of the supersonic expansion (where only 1 ion in 100 is transmitted) and in the third stage of the system (where only 1 ion in 10000 is detected). Adapted from a drawing prepared by D. M. Chambers From these rough calculations several important conclu- sions can be derived. Firstly it seems unlikely that any other atmospheric-pressure ion source will be more efficient than the ICP. Of course substantial gains might be derived through use of more efficient approaches for sample introduction; the model used here figures a 99% loss in the analyte introduced into it. Direct sample insertion laser ablation ultrasonic nebulization and a host of other approaches are superior and the improved detection limits that such techniques provide offer evidence of this.Secondly it must be recognized that a reduced-pressure source might offer analyte throughput efficiencies higher than the ICP. Because a 100-fold loss in analyte throughput occurs at the skimmer orifice an ion source that takes the place of the supersonic expansion in the interface of an ICP mass spectrometer could probably be sampled more effici- ently. Modified GD sources or other reduced-pressure units are already being examined and are discussed in greater detail later. Thirdly it is clear that the most commonly employed mass spectrometer for ICP-MS a quadrupole mass filter is a rather inefficient device.2 Not only does the quadrupole suffer a fairly low transmission efficiency as the foregoing calculations indicate but it is capable of measuring only a single mass at a time.As a result requiring the same number of counts from 50 different elements or isotopes will take 50 times as long as it would on an instrument that measured all masses simultaneously. Lastly the ratio of argon ions to analyte ions in the ICP tail flame must be borne in mind when alternative ion- sampling schemes or mass spectrometric configurations are considered. The calculations described above indicate that even a solution with a relatively high concentration of the analyte element (1 ppm) will produce an analyte ion concentration in the ICP that is only times that of the argon ion number density. Because many ion optic and mass spectrometer arrangements are limited in dynamic range or by space charge (coulombic repulsion) problems,JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1992 VOL.7 785 they might not be readily applicable to sampling the ions from an ICP. For example it is commonly held that a Paul (r.f.) ion trap can contain no greater than 1 x lo6 ions at a time. If those ions were extracted directly from an operating ICP the trap would probably contain only a single analyte ion even if the initial concentration of the solution were as high as 1 ppm. Given the million-fold loss in analyte ion throughput calculated above it might seem surprising that ICP-MS is capable of providing such low detection limits. If the mass spectrometer and its interface are as inefficient as they appear to be why are ICP-MS detection limits so much lower than those found in ICP emission spectrometry? Again the answer to this question can be found through a simple calculation.By definition the detection limit is that amount or concentration of an analyte that can be reliably distin- guished (at a desired confidence level) from a blank or background signal. Therefore detection limits depend not only on the magnitude of a signal but also on the level of background or blank noise. In turn background noise levels in ICP-MS are far lower than those encountered in ICP emission spectrometry. It was stated above that a solution with a concentration of 1 ppm produces an ICP-MS signal of approximately 1 x lo6 counts s-l. In contrast experience shows that a 1 ppm solution will generate a photocurrent in emission spectrometry of approximately 1 pA or if photon counting is utilized a signal of 40 x lo6 counts s-l.Thus the signal in emission is actually as much as 40 times that in mass spectrometry. However in ICP-MS commonly encoun- tered background count rates are between 1 and 10 counts s-l whereas in emission measurements the back- ground is more of the order of 1 x lo-* A or about 0.4 x 1 O6 counts s-l. This far higher background in emission determinations is caused both by continuum emission from the ICP and also by the fact that photodetectors for emission spectrometry are inherently noisier than ion detectors utilized for MS. This higher level of detector noise derives directly from the fact that photons in the ultraviolet or visible spectral regions are far less energetic than is an ion accelerated onto a detector surface.Because of this energy difference the ‘work function’ that is required in a photon detector must be far lower than that in a detector intended for ions. As a consequence thermally generated noise (t hermionic emis- sion) is a far more serious problem in the photon detector. This unavoidably higher background level in the emission photodetector is exacerbated by continuum emission from the ICP generated by ion-electron recombination and depending upon the spectral region being viewed by molecular emission features. An added complication is that the background level produced by an ICP is not completely stable but fluctuates somewhat because of inherent instabilities in the ICP tail flame in the sample introduction system and by the possible presence of intact or recently evaporated aerosol droplets that enter the tail flame.These considerations lead to important conclusions for both emission and mass spectrometry. Better detection limits in emission spectrometry are likely to be realized only by using a source of lower background emission and to a somewhat lesser extent of greater stability. In contrast gains in the detection capability of PSMS are likely to be achieved only by improving the efficiency of sample utilization or throughput in the mass spectrometer and its interface. Other Shortcomings of PSMS Low detection limits are hardly the only consideration that dictates the use and attractiveness of a technique for Table 1 Key problem areas in PSMS Isobaric overlap (spectral interferences) Sample matrix interferences Long-term instability (drift) Orifice clogging (high salt samples) Cost (compared with ICP emission) Limited precision (~0.5%) Difficulty with transient samples (e.g.FIA) Instrument maintenance elemental analysis. A review of limitations in both ICP-MS and GDMS3 suggests that the most significant shortcomings are those listed in Table 1. Most users would agree that some of the most serious errors in either ICP-MS or GDMS arise from isobaric overlap of atomic spectral peaks with those of polyatomic species. By and large the polyatomic ions of greatest abundance in ICP-MS are oxides whereas the most troublesome overlaps in GDMS are from argides. Other matrix interferences are also troublesome in both techniques. Perhaps the most serious is the so-called ‘mass- dependent interferen~e’,~-~ which affects virtually every element.In brief this type of interference results in a loss of analyte signal which increases in severity as the mass of the interfering element increases. Also the interference is greater for light elements than for heavier ones; hence the term ‘mass-dependent’. Long-term instability is also a problem in most specially constructed and commercial PSMS instruments. Although attempts have been made to stabilize both the plasma source and the mass spectrometer difficulties remain. Some of the drift is no doubt caused by the clogging of sampling and skimmer orifices by sample material. Understandably such clogging is most severe when solution concentrations in ICP-MS are high.Precision is also lower in PSMS than is desirable. Although it is possible to approach levels near 0.5% relative standard deviation (RSD) when an internal standard is employed more common figures are 1-5% RSD. Transient samples also cause difficulty in all current PSMS instruments a result completely of the fact that mass filters are sequentially scanned devices. Whether a quadru- pole mass filter or sector arrangement is used each isotope or element of interest is measured at a time different from others. If a limited measurement interval is available a trade-off must therefore be made between examining a large number of elements or isotopes at reduced sensitivity and precision or restricting a measurement to only a few elements or isotopes in order to improve the signal-to-noise ratio.With a transient sample the luxury does not exist for observing all the spectral peaks of interest for as long as would be desired. Finally improvement is possible in a number of practical areas including instrument costs and maintenance. A mass spectrometer is clearly a more complicated instru- ment in many respects than is an emission spectrometer dictating its higher price and greater level of maintenance. Yet these characteristics are hardly desirable and can be accepted only if other attributes of the method outweigh them. With this narrative as a backdrop let us now review developments that are underway and others which might be on the horizon to overcome the shortcomings of PSMS listed in Table 1. In some cases the most fruitful path for improvement will be fundamental study and characteriza- tion of a particular shortcoming or the characteristics of an instrument that displays it.In other situations the most direct path to improving instrument performance will be to modify the ion source detector ion optics or mass spectrometer.786 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1992 VOL. 7 Fundamental Path to PSMS Improvement Because it is 1992 just 500 years after Christopher Columbus landed in the New World it is not inappropriate to reflect on his accomplishments. An interesting parallel exists between the experiences of Columbus and the development of some chemical instrumentation and tech- niques in that they both rely heavily and in some cases excessively on serendipity. While seeking a shorter route to India Columbus stumbled onto a hitherto unrecognized land of immense potential.While searching for ways of obtaining stronger signals users of PSMS sometimes happen upon conditions that lead to reduced inter-element effects and improved precision. Of course neither approach to geographical or scientific discovery is particularly elegant and neither can be guaranteed to succeed. A more satisfactory and certainly satisfying approach would be to identify clearly a desired goal (a new continent reduced interferences or whatever) to explore alternative avenues for attaining the goal and to pursue the approaches in decreasing order of probability of success. In scientific research it is therefore better to clarify the underlying reasons for poor sensitivity inter-element interferences or unacceptable drift and through that understanding to overcome the problems.With this better understanding instrument performance will be more predictable and difficulties that are encountered can be overcome through rational adjustment of operating parameters rather than through empirical optimization. That this fundamental approach can be effective is found in a group of recent p~blications~-~ that were intended to elucidate the origin of mass dependent interference in ICP- MS. The suite of studies was aimed at clarifying the fate of an ion as it leaves the ICP tail flame is extracted through the sampling orifice of an ICP mass spectrometer is skimmed and is transmitted by the ion optics and mass spectrometer.The tools in the investigation included retarding plates placed in the second (1 33 mPa) and third (1.3 mPa) stages of the mass spectrometer interface in order to measure ion kinetic energies and a Langmuir probe which could map the electrostatic characteristics of the supersonic beam extracted into the first stage of the interface. The characteristics and composition of the expansion and the resulting ion beam were then measured and mapped as a number of controlled operating para- meters were varied. Among such parameters were the ICP central-gas flow the aerosol and solvent vapour load in the ICP the pressure in the first vacuum stage and the configuration of the ion optics used in the interface. The studies all verified that the mass-dependent interfer- ence was indeed a result of coulombic repulsion in the dense ion beam that was produced in the interface.However the finding that space charge was greatest in the second stage of the interface was unexpected. In that zone the gas density is sufficiently low that electrostatic repulsion can have a significant effect on ion trajectories yet high enough that charged species are brought into extremely close proximity to each other. Furthermore ion-optical elements that are commonly placed in the mass spectro- meter second stage force ions in the beam even closer to each other where the coulombic effects (space charge) become dramatic. These findings suggest that straightforward solutions to reducing the mass dependent interference in PSMS include relatively simple modifications to the ion optics.In particu- lar for the mass spectrometer used in this laboratory,IO removal of the ion optics in the second stage was sufficient to reduce the interferences to a negligible level. Removing the ion optics in the second stage of the interface had a dramatic effect on ion throughput. To compensate for the resulting loss in efficiency the remain- ing ion optics were retuned slightly and the ‘photon stop’ that is used in most interfaces was removed. Surprisingly removing the photon stop did not increase background levels caused by ‘photon noise’ significantly since the ion detector is placed well off axis and ions are deflected into it by means of two charged plates. In essence the ‘photon stop’ in the mass spectrometer served really as an ‘ion stop’ rather than a photon attenuator.It will be interesting to determine whether similar behaviour is found in commer- cially available and in other laboratory-constructed instru- ments. Regardless of how widely applicable these findings are the point remains clear. By a more thorough understanding of the origins of an interference novel approaches ,might be devised to overcome it. It is therefore important that laboratories throughout the world seek a better understand- ing of why the other shortcomings listed in Table 1 exist and how they might most efficiently be overcome. Improvement Through Instrument Redesign As suggested in the previous section improvements in the performance of PSMS might be achieved by modifying the plasma source the mass spectrometer the ion detector or the PSMS interface.The benefits that might be gained from modifications to the ion source are examined first. Better Ion Sources for PSMS Alleviation of isobaric overlaps from polyatomic species can perhaps be accomplished best by changes in the ion source. Indeed success in this area has already been dramatic and has resulted mostly from the use of mixed-gas plasmas and the adoption of more efficient solvent-removal systems.11J2 For example the addition of nitrogen or other molecular gases serves to raise the thermal temperature of an ICP and to dissociate polyatomic clusters that it would otherwise form. In addition removing the solvent from an aerosol to be sent into an ICP serves to remove the principal source of oxygen responsible for the production of oxide- containing ions. With a cryogenic desolvation device it is possible even to remove volatile substances such as hydrogen chloride and the polyatomic interferences they cause.12 It is possible to tailor the plasma chemistry in additional ways to reduce the severity of polyatomic ion interferences.For example it has been shown that xenon can be added to the central-gas flow of an ICP where it will undergo charge exchange with a number of troublesome polyatomic ions.12 Through this charge-exchange process the polyatomic species are neutralized and are no longer detected. Other source-based methods for reducing isobaric inter- ferences would include the adoptian of modified ICP torches to reduce the entrainment of atmospheric gases.Such torches might include extended outer (coolant) tubes a more linear or laminar outer-gas flow or might be of larger diameter to help isolate the central analyte-contain- ing channel from outside influences. More dramatic alterations in source design might result in even more significant improvements in performance. Especially appealing are the features of a ‘tandem’ s o ~ r c e . ~ J In the tandem arrangement two independent sources are coupled to produce a combination that offers the best characteristics of each. The first source in the pair would be designed to vaporize and atomize the sample directed into it; atoms it produced would then be ionized by the second source. Because the functions of sample atomi- zation and atomic ionization are separated each of the two sources can be independently optimized for its intendedJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY.SEPTEMBER 1992. VOL. 7 787 function. Thus the first source should be able more effectively to atomize the sample fully so polyatomic ions would be in far lower abundance. A tandem source would also lend itself better to feedback stabilization than would a single device. It has already been shown how emission signals from an ICP can be stabilized by use of an emission-based feedback signal.14 A reference element added intentionally to a sample solution provides the feedback signal which in turn can control r.f. power to the ICP in a manner that stabilizes emission signals from it. Although the same technique is being applied currently to stabilize an ICP-MS instrument the concept would be even more effective if applied to a tandem source.Feedback signals from for example both atomic and polyatomic ion channels could be employed to optimize the atomization behaviour of the first source while signals from ions of high ionization energy and from doubly charged ions could be used to enhance and stabilize the performance of the second (ionization) source. Significantly this optimization could be performed automatically under computer control and could be tailored to each individual sample so the resulting mass spectrometric signals would be stronger stabler and more interference-free than is now the case. The two sources in a tandem pair might operate by any of a number of mechanisms; a few are listed in Table 2.Interestingly many of these and other combinations have already been explored for use in tandem sources. Indeed the GD is itself a tandem source with its atomization being performed by a mechanical sputtering phenomenon while ionization occurs either by charge transfer from metastable species (Penning ionization) or by electron impact. Addi- tional details about the tandem source concept can be found e l s e ~ h e r e . ~ J ~ J ~ Hopefully workers will be able to conceive their own tandem source combinations that are even more attractive than those that have already been described in the literature. Because the second (ionization) source in a tandem pair need only ionize atoms directed into it it can be relatively simple. A particularly attractive device is the atmospheric sampling GD described by McLuckey et af.16 Such a system could accept atoms produced by a range of alternative atom sources.In fact because the tandem-source scheme lends itself to modularity any sort of atom generator could be coupled to the ionization source. That is one first source could be employed when it is necessary to atomize conductive solids directly another when liquids are to be analysed another for non-conductive solids one for gase- ous samples another for microsamples one that provides three-dimensional spatial profiling in solids a unit that accepts the effluent from a chromatographic column one to employ flow injection and others. Furthermore it might be possible to employ a first (atomization) source that is modulated in amplitude.13J5 In this arrangement the first (atomization) source would be operated in an alternating fashion at two operating levels one of which is intended to atomize the sample fully and the lower of which is intended merely to fragment the sample.Table 2 Tandem source mechanisms Source I Source 2 (atomization) (excitation/ionization) gas Energykharge transfer Hot high-pressure (metastable species ions etc.) Mechanical Hot surface Chemical reaction Radiative ablation Radiative (sputtering) Electron impact/collision With a fairly fast mass spectrometer attached to the ionization source the tandem pair would thus produce in rapid and alternating sequence an atomic and a fragmenta- tion mass spectrum so the identity and structure of a sample could be determined more unambiguously.Such a combination would be a powerful tool for detection of the effluent from a chromatography column or for identifying the species in which particular chemical elements are contained. As a final point in the consideration of new ion sources for elemental mass spectrometry it must be borne in mind that the characteristics of the ion source are likely to dictate in large measure what the interface to the mass spectro- meter must accomplish and how it must be designed and also what requirements are placed on the mass spectro- meter itself. For example the novel electrospray ion source developed by Blades et al.l7-I9 and modified by Agnes and Horlick20121 operates in an extraordinarily simple and direct fashion. It is necessary merely to apply a high voltage to the solution to be analysed.Charges accumulate on the surface of the solution and lead to its electrostakic disruption. Because many of the charges detach from the surface in the form of elemental ions or polyatomic clusters in which they are included they can be directed into a mass spectrometer for detection. If the interface leading to the mass spectro- meter is suitably designed many of the polyatomic frag- ments are decomposed so a relatively clean atomic mass spectrum can be produced.20*2* Not only does this new approach lead to simplicity in source design and the potential ability to determine the chemical species in which the elements are originally contained but the source can also operate with an extraor- dinarily low gas flow. A low gas flow reduces tremendously vacuum pumping requirements and could conceivably lead to simplified designs of mass spectrometers just as has been the case in the development of mass spectrometric detectors for gas chromatography.Improved Mass Spectrometers for PSMS Benefits will be (and are being) derived from redesign of the mass spectrometer interface and by substituting alternative types of mass spectrometers for those now commonly employed. A clear example of how suitable mass spectro- meter design can overcome some of the problems listed in Table 1 is the use of a high-resolution double focusing mass spectrometer to resolve isobaric overlaps. At a resolving power of approximately 8000 it becomes possible to separate the overlap of ArCl from As; many other useful examples exist.It might be possible to alleviate polyatomic ion interfer- ences in even more subtle ways. Many believe that an abundance of polyatomic species is created in the stagnant zone that lies near the surface of the sampling cone in the PSMS interface. In this relatively cool stagnant region recombination and ion-molecule reactions no doubt take place and might lead to complications in a mass spectrum. Of course if the beam expansion from the sampling orifice were ideal the ion-source gases that entered the sampling orifice would not contain any of the polyatomic species that form in the stagnant boundary layer. In a real interface however the situation might be more similar to that depicted in Fig. 3 in which traces of the polyatomic species near the sampling cone are carried into the interface.Luckily it would seem that such polyatomics would be in greater abundance in the periphery of the expanding plume of ion gases rather than in its centre. Furthermore they might exhibit a different kinetic energy distribution than species skimmed from the centre of the supersonic expan- sion. As a result it might be possible to discriminate against788 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1992 VOL. 7 Fig. 3 Schematic illustration of the entrainment of external gases into the supersonic expansion of the mass spectrometer first stage. Polyatomic ions might form in the stagnant zone that lies along the outside surface of the sampling cone and would be carried into the mass spectrometer along with the plasma gases them on the basis of their kinetic energies or possibly through use of a more spatially selective configuration of the ion optical system.In fact recent results from the laboratory of Houk et allZ indicated that a spatially selective ion optical system might be helpful in lessening polyatomic ion interferences. Other approaches to reducing the severity of polyatomic interferences include the use of collisional dissociation either in a moderate-pressure gas cell or with a surface. The first approach has been explored with some success,22 but to this author's knowledge surface-induced dissociation has not yet been examined to aid PSMS. It might be possible also to reduce the deleterious effects of instrument drift by suitable redesign of the mass spectrometer or by adoption of different mass spectrome- tric configurations. Obviously a straightforward method to reduce drift is to stabilize the mass spectrometer itself.Manufacturers already go to some lengths to overcome temperature-induced instability and to minimize instru- mental variation. However these approaches are unlikely to be fully effective since the dominant source of long-term variation in a PSMS instrument is no doubt the plasma source. In such a situation the only reliable ways to overcome drift are either to stabilize the source as documented earlier or to employ internal standardization. In all likelihood both schemes will be necessary since some long-term variation derives from the sample introduction system and from changes in plasma-source throughput whereas others are a consequence of unstable atom pro- duction and atomic ionization.The latter sources of drift will have to be compensated for completely by stabilizing the source; only the former can be compensated for completely by internal standardizati~n.~~ For an internal standard by be fully effective its characteristics must match those of the species to be determined and its variation must be correlated with that of the analyte. Furthermore even if these conditions are both met the analyte peak and that of the internal standard must be measured either simultaneously or in sufficiently rapid sequence that instrumental drift is insignificant during the intervening period. If instrument response varies for any reason between the times when an analyte signal and an internal standard are measured changes in the two signals will not faithfully track each other and complete compensa- tion will not be possible.This behaviour was documented clearly in a study by F ~ r u t a * ~ who examined the precision of isotope ratio measurements as a function of the dwell time on each isotope. A peak hopping approach was employed and the isotope signals of interest were measured as the rate of peak hopping was increased. It was found that precision continu- ously improved until the dwell time per peak was as low as 10 ps beyond which point the precision of the computed isotope ratio was limited by counting statistics. In other words Furuta found that it was necessary to switch back and forth between the two peaks at a rate of 1 x lo5 Hz in order to derive the best precision from the mass spectro- meter. If this situation prevails for other sequential-scanning mass spectrometers they will never be able to provide high precision in a complete multi-element analysis.An operator will have to choose between measuring a spectral peak and its internal standard in sufficiently rapid sequence to obtain high precision or examining several elements but with a greater interval before the internal standard peak can be examined. Because of this greater interval precision will suffer. For these reasons the most attractive desigvs for future atomic mass spectrometer systems are those that offer simultaneous readout of all the isotope peaks of interest. Such systems would not only yield improved precision in accordance with the foregoing arguments but would lend themselves more naturally to the measurement of signals from transient samples such as those produced by laser ablation flow injection and chromatographic systems.Other limitations of current quadrupole mass spectromet- ers also indicate that new directions be taken. These limitations include limited resolution the relative low transmission efficiency described earlier and only moderate stability. Which mass spectrometer alternatives should be consi- dered? Those which are already being explored and which offer some appeal include the Paul ion trap Fourier transform mass spectrometry (FTMS) time-of-flight mass spectrometry (TOFMS) and a sector instrument coupled with a diode-array readout.Preliminary work'with an ion trap has already proved its effectivenes~.~~ However to be attractive from a practical point of view the ion trap will have to be employed with some form of mass-selective pre-filter because of the limited dynamic range mentioned earlier.26 The purpose of the pre- filter would be to exclude argon ions from the trap so its limited dynamic range could be exploited fully. A promis- ing candidate for such a pre-filter would be a quadrupole device operated in a notch filter mode.z7 Even with this preliminary filtering however the appeal of the ion trap might be somewhat limited in that the most abundant analyte species that it contains should be no more than lo6 times the concentration of the least abundant ions. Importantly the ,ion trap unlike the quadrupole mass filter or scanned sector-based mass spectrometer produces a mass spectrum that is derived from ions extracted simultaneously from the source.Thus internal standardiza- tion should be more effective. Furthermore because the ion trap can be scanned at high speed it should be applicable to the measurement of transient samples. An FT mass spectrometer is attractive as a spectrometer for elemental analysis because it provides truly simulta- neous measurement of all masses and offers extraordinarily high r e s o l u t i ~ n . ~ ~ ~ ~ ~ Yet like the ion trap an FTMS offers limited dynamic range and is at present extremely expen- sive. Furthermore if it is to provide high-resolution mass spectra its time of analysis can be relatively long. A TOFMS is appealing for use in elemental MS in part because of its great speed simplicity high transmission efficiency and its ability to extract ions simultaneously from a source.However shortcomings of the TOFMS would seem to be the low duty factors under which it operates its limited resolution and the relatively complex electronics that are ordinarily needed for its operation. However the shortcomings of TOFMS are not as serious as they might initially seem. At first it might seemJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1992 VOL. 7 789 foolhardy to apply TOFMS to a continuously operating ion source such as those commonly employed for atomic spectrometry. After all the TOFMS instrument operates by extracting an extremely brief (typically 20 ns) pulse of ions from a source accelerating them down a flight tube and recording their times of arrival.It is ordinarily not possible therefore to extract another aliquot of ions from the source until the full mass spectrum is recorded a time interval of the order of tens of microseconds. Thus the duty factor of a TOFMS is something of the order of or less. Still a TOFMS offers almost unity transmission efficiency. Fur- thermore it records a complete mass spectrum from each ion pulse extracted from the source; these factors alone would make the TOFMS at least competitive with instru- ments such as quadrupole mass filters and sector-based mass spectrometers. However the TOFMS offers other strengths also. It is extraordinarily fast and also extracts all ions at the same time from an operating source.Thus it should be far more powerful when applied to transient atom sources and should be effective also in reducing the complications of source drift if internal standardization is employed. Moreover as has already been demonstrated experimentally a TOFMS can be operated in a right-angle or in a ‘multiplexed’ c~nfiguration~~ that can raise the duty factor to levels of between 0.1 and 1 .O. In the first of these schemes an ion beam extracted from an operating source forms a supersonic expansion just as is ordinarily found in an ICP-MS interface. Because ions in the expan- sion have only a moderate velocity they fill the expansion volume at only a moderate rate. Once that expansion volume is filled the narrow ion beam can be pulsed in a direction perpendicular to its original route and accelerated in that perpendicular direction down the flight tube.During the time it takes to record the mass spectrum from those ions the initial expansion volume can be refilled with ions that will eventually form the next input pulse to the TOFMS. Importantly this right-angle configuration has the advantage also of accelerating the ions in a direction perpendicular to the initial supersonic expansion. Because the kinetic energy distribution of ions in that perpendicular direction is extremely narrow resolution even in a linear TOFMS is good; a resolving power of approximately 500 has already been demonstrated for a corona discharge source.31 In the second (multiplex) scheme to increase the TOFMS duty factor the flight tube is essentially shared by a number of input pulses simultaneously.Termed ‘Fourier transform TOFMS’,3Z the method can achieve virtually unity duty factor performance from a TOFMS. Another attractive combination for potential use in atomic mass spectrometry couples a sector based mass spectrometer with a diode-array detector. Atomic spectro- metrists are accustomed to using photodiode arrays but for multichannel readout in emission spectrometry. In such applications a trade-off is usually required between the examination of a broad spectral range at only moderate resolution or the measurement of a narrow spectral range but with only a few spectral lines being determined. Of course neither option is completely satisfactory and a number of elegant schemes have been devised in an attempt to overcome the limitation. A linear diode array would seem to be almost ideally suited for atomic mass spectrometry. Moderately priced diode-array formats contain 256 5 12 1024 2048 or 4096 diodes in a row all equally spaced on approximately 25.4 pm centres.Conveniently an atomic mass spectrum con- tains only about 250 atomic mass units of interest and in most applications half-mass resolution is all that is needed. Thus even a 1024 element diode array would enable the measurement of four points over each atomic mass-spectral peak of interest The detector would be simultaneous integrating and could be read out at high speed if a transient sample were being measured. This diode-array approach has long been recognized as viable by people in the mass spectrometry c ~ r n m u n i t y .~ ~ ~ ~ ~ However there are several complications to the approach. Firstly a linear photodiode array such as that generally used in atomic emission spectrometry is not by itself a particularly effective detector for high-energy ions. As a result some kind of ion-to-photon conversion interface is needed. In most arrangements this conversion is a multi- step process which involves colliding the ion of interest with a microchannel plate imaging the resulting pulse of electrons onto the phosphor-coated surface of a fibre optic array and allowing the fibre optic to guide the resulting photons to an individual pixel on the photodiode array. Thus several conversion steps are needed from ions to electrons to photons and back again to an electronic signal in the diode array.In each step losses occur noise is generated and spatial (Le. spectral) resolution can be Sacrificed. Surely improved schemes can be found. A second compromise in the sector-diode array combina- tion is the limited mass range that can ordinarily be covered. Although there are more than enough pixels on an inexpensive linear diode array to cover the atomic mass range of interest the mass display produced by a sector instrument is not linear. Most studies have therefore used arrays to cover only modest mass ranges commonly spanning only a factor of two in atomic mass units (e.g. from mlz to 2mlz). Still novel mass spectrometers sector arrangements multiple diode arrays successive accelerat- ing voltages or switched magnetic fields might be utilized to cover a broader mass range.Conclusion From the foregoing narrative it should be clear that many opportunities exist for overcoming the limitations of PSMS that are listed in Table 1. Fundamental studies that are already underway should serve to characterize more fully how and where polyatomic species are formed in PSMS instruments and how they might be minimized. Such studies might also be able to define better where other sorts of matrix interferences arise and what the sources of long- term instability might be. Similarly the development investigation and adoption of novel ion sources interface configurations ion-optic arrangements and mass spectrometers might aid in over- coming spectral and matrix interferences increase instru- mental stability and precision and make PSMS instruments applicable to samples of different kinds.At the same time cost might be lowered instruments might be reduced in size and their maintenance simplified and the over-all capabili- ties of PSMS enhanced. To achieve these ends future instruments used in PSMS might incorporate a number of the features outlined in Table 3. Perhaps most ,importantly a next-generation system should provide simultaneous measurement of all masses and isotopes of interest in order to enhance precision through isotope dilution or internal standardiza- tion increase sample throughput and make the system more amenable to coupling with an atom source that produces transient signals. Ideally such instruments should be able to operate at either moderate or high resolution depending upon the incidence of spectral interference and the need for high sensitivity.Despite the simplicity of the ICP and GD it would seem that tandem sources have an important role to play in future PSMS instruments. Through proper feedback they could be made more stable and could be configured to reduce the incidence of polyatomic interferences applied to samples of many different kinds by means of a modular design and conceiv-790 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1992 VOL. 7 Table 3 Next-generation PSMS instruments Simultaneous detection of all isotopes Fast Precise Transient sampling Moderate or high resolution Tandem source Stable Fewer polyatomics Flexible (modular) Atomic or molecular Feedback stabilized Reasonable cost ably be employed in a modulated configuration to produce simultaneously atomic and molecular mass spectra.Whether or not a tandem source is employed it is certain that the next generation of PSMS systems should incorpor- ate advanced diagnostics to enhance the characteristics of sample introduction and to stabilize the instrument. Such feedback approaches are becoming increasingly sophisti- cated and powerful and could be incorporated readily even into many current computer-operated units. If these goals are all met it would seem likely that PSMS instruments could dominate the field of atomic spectrome- try. If such were the case economies of production would no doubt yield a reduction in instrumental cost. In turn lower priced instruments would increase the competitive advantage of PSMS over atomic absorption or emission approaches making the cycle repeat itself.It will be interesting to see how far this cycle carries us as we approach the new millenium. 9 Chambers D. M. Ross B. S. and Hieftje G. M. Spectrochim. Acta Part B 1991 46 785. 10 Ross B. S. and Hieftje G. M. Spectrochim. Acta Part B 199 1,46 1263. 1 1 Lam J. W. and McLaren J. W. J. Anal. At. Spectrom. 1990 5 419. 12 Houk R. S. Hu K. and Clemons P. S. paper presented at the Winter Conference on Plasma Spectrochemistry San Diego CA USA January 6-1 1 1992 (paper 1L13). 13 Hieftje G. M. Spectrochim. Acta 1990 44 (Spec. Suppl.) 113. 14 Marks M. A. and Hieftje G. M. Appl. Spectrosc. 1988 42 277. 15 Hieftje G. M. Fresenius’ J.Anal. Chem. 1990 337 528. 16 McLuckey S. A. Glish G. L. Asano K. G. and Grant B. C. Anal. Chem. 1988 60 2221. 17 Blades A. T. Jayaweera P. Ikonomou M. G. and Kebarle P. J. Chem. Phys. 1990 92 5900. 18 Blades A. T. Jayaweera P. Ikonomou M. G. and Kebarle P. Int. J. Mass Spectrom. Ion Proc. 1990 101 325. 19 Blades A. T. Jayaweera P. Ikonomou M. G. and Kebarle P. Int. J. Mass Spectrom. Ion Proc. 1990 102 251. 20 Agnes G. and Horlick G. presented at the 1992 Winter Conference on Plasma Spectrochemistry San Diego CA USA January 6- 1 1 1992 (poster ThP9). 2 1 Horlick G. paper presented at the 1992 Winter Conference on Plasma Spectrochemistry San Diego CA USA January 6-1 1 1992 (paper PL5). 22 Duckworth D. C. and Marcus R. K. Appl. Spectrosc. 1990 44,649. 23 Carre M. Poussel E. and Mermet J.-M. J. Anal. At. Spectrom. 1992 7 791. 24 Furuta N. J. Anal. At. Spectrom. 1991 6 199. 25 Gill C. G. Daigle B. and Blades M. W. Spectrochim. Acta Part B 1991,46 1227. 26 Koppenaal D. W. Barinaga C. J. and Smith M. R. presented at the 1992 Winter Conference on Plasma Spectro- chemistry San Diego CA USA January 6-1 1 (paperThP16). 27 Denton M. B. lnt.-J. Mass Spectrom. Ion Phys. 1981,37,241. 28 Barshick C. and Eyler J. J. Am. SOC. MassSpectrom. 1992 in the press. 29 Marcus R. K. Duckworth D. C. Glish G. L. McLuckey S. A. Buchanan M. Wise M. Pochkowski J. M. and Weller This work was Institutes of Health through grant GM 46853 by the National Science Foundation through grant CHE 90-20631 and by the Leco Comorat ion. in part by the References Vickers G. H. PhD Dissertation Indiana University July Brunee C. Int. J. Mass Spectrom. Ion Proc. 1987 76 121. Hieftje G. M. and Norman L. A. Int J. Mass Spectrom. Ion Proc. in the press. Beauchemin D. McLaren J. W. and Berman S. S. Spectro- chim. Acta Part B 1987 42 467. Tan S. H. and Horlick G. J. Anal. At. Spectrom. 1987 2 745. Vickers G. H. Ross B. S. and Hieftje G. M. Appl. Spectrosc. 1989,43 1330. Chambers D. M. Poehlman J. Yang P. and Hieftje G. M. Spectrochim. Acta Part B 1991 46 741. Chambers D. M. and Hieftje G. M. Spectrochim. Acta Part B 1991 46 761. 1989 pp. 64-70. R. R. paper presented at the 1992 Winter Conference on Plasma Spectrochemistry San Diego; CA USA January 6- 1 1 1992 (paper 53). 30 Dodonov A. F. Chernushevich I. V. and Laiko V. V. presented ,at the International Mass Spectrometry Conference Amsterdam The Netherlands August 26-30 199 1 (poster 31 Sin C. H. Lee E. D. and Lee M. L. Anal. Chem. 1991,63 2897. 32 Knorr F. J. Ajami M. and Chatfield D. A. Anal. Chem. 1986 58 690. 33 Boettger H. F. Giffin C. E. and Norris D. D. ACSSymp. Ser. 1979 102 291. 34 Sinha M. P. and Gutnikov G. Anal. Chem. 1991,63 2012. TUA-D20). Paper 2/00625A Received February 5 1992 Accepted May 20 1992