Characterization of Ionization and Matrix Suppression in Inductively Coupled CCold' Plasma Mass Spectrometry* Journal of Analytical Atomic Spectrometry SCOTT D. TANNER SCIEX 71 Four Valley Drive Concord Ontario Canada L4K 4V8 A parametric study of plasma power and central gas flow was carried out to study the transition from normal analytical conditions to cooler plasma conditions using an inductively coupled plasma mass spectrometer having a balanced load coil. 'Cold plasma' conditions (low power and high central gas flow) permit the determination of K Ca and Fe at trace levels. The effect of changing the position of the ground reference of the load coil was investigated. Trace element ionization is consistent with thermal ionization at low electron density. Ion- molecule chemistry (charge transfer) with NO+ or 02+ may be important at the cooler plasma temperature.Suppression of analyte signals by concomitant matrix elements is partially correlated with the ionization potentials of the matrix element. If the analyte ion signals are normalized to that for NO' the suppression of signals appears to be independent of the matrix element and a modest dependence on the ionization potential of the analyte element is apparent. For high concentrations of elements of low ionization potential an additional or enhanced mechanism of ionization is evident. The onset for this enhanced ionization is sharply defined by a characteristic ionization potential near 6.0 eV. The sensitivity for trace elements does not appear to be affected by this enhanced ionization.The appearance of the enhanced ionization is made evident by a change in the ratio of the NO+ and 02+ signals. Use of cold plasma conditions for the determination of K Ca and Fe in high-purity waters and acids is evident. It appears that the method may also be used for samples having moderate salt content if the analytical protocol includes measurement of the background ions NO+ and 02+. Keywords Inductively coupled plasma; mass spectrometry; secondary discharge; matrix efect ; easily ionized element eflect The appearance of oxide ions of refractory elements was observed in the earliest reports on the performance of induc- tively coupled plasma mass spectrometry (ICP-MS).ly2 With the boundary sampling used these oxide ions could dominate over atomic ions. The appearance of polyatomic ions can compromise the determination of isobaric (having the same mass-to-charge ratio) elements.When the interference appears at the most abundant isotope of the element to be determined the analyst may have to resort to measurements at a less abundant isotope with a concomitant loss of sensitivity. Progress in the reduction of these interferences has been substantial (for example through interface design3 and the use of mixed gas plasmas4) but polyatomic ions associated with primary plasma ions remain a problem. One of the most significant of these interferences is that of ArO" with the dominant isotope of Fe+ at m/z=56. Addition of N2 to the plasma gas can substantially reduce the interference at m/z= 56 but increases the interference at the less abundant isotope 54Fe+ by ArN+.4 In addition the primary plasma ions can interfere directly with the determination of certain elements * Presented at the 1995 European Winter Conference on Plasma Spectrochemistry Cambridge UK January 8-13 1995.(e.g. 40Ar+ with 40Ca+). Even for current instruments that show high abundance sensitivity (a measure of the residual signal at 1 u lower than the mass of a dominant ion which is a limitation in mass spectral resolution) the very large signal for 40Ar+ can interfere with the dominant isotope of K+ at 39 u. In addition the measurement at m/z = 39 is also compli- cated by the presence of 38Ar1H+ and that at the other important isotope of K is interfered with by 40Ar1H+. One solution to this isobaric interference problem is to analyse the sample under high mass spectral resolution (e.g.a magnetic sector mass ~pectrometer).'>~ At a resolution of m/Am=2500 the signal for ArO+ is resolved from that for 56Fe+ at about half peak height. Where the Fe is to be determined at very low trace levels yet higher resolution is required. An alternative approach for certain elements appli- cable to quadrupole-based instruments is to operate the plasma source under conditions of lower power and higher nebulizer flow (and perhaps increased sampling depth and aerosol desolvation). This approach was first reported by Jiang et ~ 1 . ~ for the determination of K isotope ratios. They observed that under these conditions the background mass spectrum changed from one dominated by Arf species to one dominated by NO'.This opened the possibility of determining the K isotope ratios without substantial interference from Ar + or ArH+. The calibration graphs obtained under these 'cooler' plasma conditions were found to show two linear segments intersecting at about 10 mg 1-1 (10 ppm) K. This change in response was ascribed to a change in the dominant plasma ion from NO+ to K+ that is a self-induced matrix effect at about 10ppm. Appearance of a matrix effect at such low concentrations suggests that these plasma conditions may be useful only for analyses of samples having a low salt content. The correlation of the matrix effect with the change of domi- nant plasma ion also inferred that the mechanism of ionization involved ion-molecule chemistry of the plasma ion.This is distinguished from normal plasma conditions for which ana- lyte ionization is dominated by electron impact,8 with ion- molecule chemistry (charge transfer) playing a significant role only in the formation of highly excited ion^.^^^^ The original report on using 'cooler' plasma conditions ended with a caution that such results appeared to be achievable only under con- ditions where a secondary discharge between the plasma and the sampling orifice was suppressed. Subsequently Sakata and Kawabata" recognized the impor- tance of this mode of operation for the determination of K Ca and Fe by ICP-MS. The insertion of a grounded metal shield between the load coil and the torch as suggested by Gray,12 attenuated the secondary discharge.Combining the 'Shield Torch' with conditions of lower plasma power higher nebulizer flow and larger sampling depth led to the suppression of polyatomic argide ions notably ArO+. That report com- pared the spectra with and without the electrostatic shield and mapped the response of certain background and analyte ions as a function of plasma power and carrier flow. Photographs of the region between the sampler and skimmer Journal of Analytical Atomic Spectrometry November 1995 Vol. 10 905demonstrated that the electrostatic shield suppressed the dis- charge that was otherwise present. It was noted that suppres- sion of the secondary discharge was of itself insufficient to attenuate the argide species to useful levels; operation under cooler plasma conditions was also required.Hence two sources for the argide interferences were identified (1) derived from the plasma itself (ascribed to capacitive coupling between the load coil and plasma resulting in ionization of polyatomic argides) and (2) formed within the first stage of the vacuum interface owing to the secondary discharge. Nonose et all3 studied the formation of polyatomic ions within the ICP and in the micro-plasma between the sampler and skimmer both with and without an electrostatic shield of the sort described by Gray' and by Sakata and Kawabata." Polyatomic ions were classified into two groups the first included NO' 0,' and metal monoxide ions and the second included the argide polyatomic ions. It was concluded that the former ions were characteristic of the source plasma (ICP) while the latter were formed in a secondary discharge within the vacuum interface.The dissociation equilibrium for ArX + species was consistent with a plasma temperature of 3500 K which the authors took to indicate equilibration within the vacuum interface. At about the same time Uchida and Itox4 reported that the electrostatic shield discussed above did reduce the secondary discharge but that some evidence of the discharge remained more noticeably at a plasma generator frequency of 40 MHz than at 27 MHz. They suggest that the residual secondary discharge may be responsible for the lower analyte monoxide ion signals observed at the higher frequency. In addition they note that air entrainment at large sampling depths may also lead to the formation of oxide ions of analyte species.Douglas and French'' reported that the intensity of the secondary discharge was reduced for a centre-grounded ('bal- anced') load coil configuration. This configuration may be obtained by various means including physical grounding of the load coil or using a balanced Colpitts oscillator.16 The effects of reducing the secondary discharge were claimed to include reduction in the intensity of doubly ionized species a narrowing of the kinetic energy distribution reduction of ions derived from erosion of the sampling aperture and increased orifice lifetime. This work was undertaken to evaluate the operation of an ICP-MS instrument having a balanced load coil configuration without a physical shield under 'cold' plasma conditions to permit the determination of K Ca and Fe at trace levels.In particular the signals of analyte and plasma background ions were monitored as a function of the plasma conditions (power and central channel gas flow) and of the balance of the load coil. It was furthermore the objective to characterize the effect of concomitant elements on the response of analyte and plasma ions (k the matrix effect). Analyte species which were treated as either trace elements or matrix elements in separate analyt- ical solutions were selected having a range of ionization potential mass and heat of vaporization. It was intended that characterization of the matrix effect would help to delineate the mechanism of ionization in the plasma and the importance of particle vaporization on analyte response.EXPERIMENTAL Instrumentation Experimental results were obtained on a Perkin-Elmer SCIEX Elan 6000 ICP-MS instrument. Since this instrument has not been described in the literature a schematic diagram of the major components is given in Fig. 1 and a functional descrip- tion of these components is provided below. The plasma rf generator is free-running (meaning that the Turbomolecular Turbomolecular Roughing Pump pump pump Fig. 1 Schematic diagram of the Perkin-Elmer SCIEX Elan 6000 ICP-MS system used in this work. The function and operation of the various components are described in the text frequency is automatically varied to maintain tuning) at a nominal frequency of 40 MHz. The three-turn load coil is a component of the Colpitts oscillator a schematic diagram of which is shown in Fig.2. The plasma potential (the dc potential of the plasma resulting from capacitive coupling between the load coil and plasma) is a function of the position of the ground reference along the coil as has been shown by Douglas and French." The load coil is 'balanced' (that is the position of the ground reference along the load coil is optimized) by adjustment of the capacitor plate CP1. Moving CP1 towards CP2 (referred to here as a negative displacement) moves the ground reference down the load coil away from the sampler. Therefore the plasma potential is a function of the position of CP1 and by inference the intensity of a secondary discharge can be minimized (or enhanced) by changing the displacement of CP1. For the experiments involving adjustment of the plasma potential the stand-offs that normally hold CP1 in position were removed along one edge and a threaded insulated rod was attached to this edge.The threaded rod extended through a hole drilled through the generator housing and was held in place with a nut on the outside of the generator. The position of CP1 was adjusted while the plasma was operating by adjusting the length of the threaded rod extending outside the generator housing which therefore moved the edge of CP1 attached to the threaded rod with respect to the fixed plates CP2 and CP3. Since CP1 was fixed in position along one edge this motion caused CP1 to tilt relative to the fixed plates CP2 and CP3 rather than move plane-parallel to CP2 and CP3. The capacitor position was measured as the extension of the threaded rod out of the generator housing.For all other experiments the capacitor was adjusted to give the largest ratio of Na' signal from a 10 pg 1-l (10 ppb) Na solution to background 40Ar' signal measured for a plasma power of 600 W and nebulizer flow of 1.08 1 min-l (ie. the 'cold' plasma conditions described below). This condition was also found to yield the largest Rh' signal for a plasma power of 1200 W OSCILLATOR COLPlrrS rf choke H-l Front (towards LOAD COIL Ground J- reference a s x j - POWER SOURCE Fig. 2 Schematic diagram of the Colpitts oscillator used in this work. Capacitor plate CP1 is adjustable between plates CP2 and CP3 and determines the ground reference of the load coil. Moving CP1 towards CP2 (denoted as a change in the negative direction) moves the ground reference away from the front of the load coil 906 Journal of Analytical Atomic Spectrometry November 1995 VoZ.10and nebulizer flow of 0.77 1 min-' (Le. the 'normal' plasma conditions described below). The sample was delivered at a rate of 1 ml min-l via a peristaltic pump to a cross-flow nebulizer. The double pass spray chamber was mounted outside the torch box and was maintained at ambient room temperature (about 21 "C). The resultant aerosol was not desolvated. For most experiments the entire injector (central carrier) gas flow was passed through the nebulizer (injector flow = nebulizer flow). For experiments in which ion signals were measured as a function of injector flow a constant nebulizer flow of 0.45 1 min-l was passed through the nebulizer (hence maintaining more-or-less uniform nebulization efficiency) and a make-up flow was added to the injector through a T downstream of the spray chamber; hence the total injector flow was the sum of the fixed nebulizer flow and the variable make-up flow.The nebulizer flow rate was metered using the instrument-native flow controller which was cross-calibrated using an external digital flow meter and the make-up flow (when used) was measured using an external digital flow meter. The alumina injector (central gas) tube had an id of 2 mm at its exit to the torch. The plasma (coolant) and auxiliary (intermediate) Ar flows were preset at 15 and l.Olmin-' respectively using the instrument-native flow controllers with a primary Ar input pressure of 50 psi.The sampling depth (distance from the end of the load coil to the sampler orifice) was set to 9mm and was not adjusted. The sampler ( 1.1 mm diameter orifice) and skimmer (0.88 mm diameter orifice) cones were made of nickel. The spacing between the sampler and skimmer was 6.9mm (stan- dard for this instrument). The background pressure in the interface region was about 3 Torr. The high vacuum chamber is differentially pumped by turbomolecular pumps. The higher pressure chamber (about 8 x lov4 Torr) contains the entrance ion optics. These optics consist of a grounded shadow stop at the base of the skimmer cone a cylinder lens and a grounded differential pumping aperture. The shadow stop intercepts unvaporized plasma particles preventing their deposition downstream in the ion optics.It also serves as an on-axis ground potential reference for the extracted plasma. The voltage applied to the cylinder lens is automatically tuned using the system software and is typically ramped in concert with the measured ion mass. The voltage applied to the lens is typically within the range +3 to +1OV and is found to optimize linearly with measured (transmitted) ion mass. To a first approximation the optimum lens voltage is approximately the kinetic energy that the ions gain from the supersonic expansion through the interface;17 hence the optimum linear scanning of lens voltage with meas- ured ion mass. For the 'cold' plasma conditions described below it was found sufficient to maintain the lens voltage at a constant +3 V irrespective of measured ion mass. The energy bandpass of these ion optics is narrow about 3 eV at half-height.18 This bandpass is comparable to the width of the ion energy distributions. Ions having kinetic energies signifi- cantly higher or lower than the applied lens voltage are not efficiently transmitted.Optimum sensitivity is obtained for ions having a small distribution of kinetic energies centred at the applied lens voltage. Ions created in the secondary discharge if present have substantially wider energy distributions and higher kinetic energy than those obtained from the supersonic expansion.15 Therefore these ion optics do not efficiently transmit ions that are generated in a secondary discharge. A further discussion of the ion optics can be found in ref.18. The higher vacuum chamber (about 1.5 x low5 Torr) down- stream of the differential pumping aperture contains the quadrupole mass analyser and the ion detector. The pressure in this chamber is accurately measured using a Bayert-Alpert (hot) ionization gauge tube. It is found that the analyser chamber pressure is a function of the plasma conditions as is to be expected [a cooler plasma is more dense and both the speed of sound (relevant for the flow through the sampler) and the terminal velocity (relevant for the flow through the skim- mer) are lower and so the flow through the interface is increased3]. In fact a reasonable estimate of the relative plasma temperatures for different plasma conditions can be obtained by measuring the relative analyser chamber pressure since the ratio of the zero-corrected (corrected for de-gassing) pressures is very nearly the inverse ratio of the squares of the absolute source plasma temperature^.^^'^ The mass analyser is a quadrupole mass filter with a capacitively coupled ac-only prefilter to enhance ion transport into the mass filter.The ion detector is an ETP Model AF210-Ml8E active film discrete dynode detector. A single ion impacting the surface of the first dynode yields both an analogue and a digital signal each of these having separately adjustable (automatically by the system software) gain factors. Both analogue and digital signals are simultaneously measured and archived. The system software cross-calibrates the dual signals. The digital gain channel is automatically protected against excessive ion current by hard- ware feedback shutdown circuitry.For the experiments reported here the gain of the analogue channel was set to saturate the analogue counter at 2 x lo9 ion counts s-l yield- ing a dynamic range (on the fly in a single scan) of approxi- mately 10'. Solutions The results reported here required the analysis of a number of related solutions. The distilled de-ionized water (DDIW) for background measurements and dilution was prepared in-house by distillation in glass followed by passage through a three- cartridge Millipore de-ionizer. The DDIW was usually pre- pared freshly. The background spectra reported here were obtained for 0.1% nitric acid ('Baker InstraAnalyzed' J. T. Baker Phillipsburg NJ USA) in DDIW.Analytical solutions were prepared as 10 pg I-' (10 ppb) of each analyte by serial dilution with 1 % nitric acid from 1000 mg 1-1 (ppm) standard Li Be B Na Al K Sc Fe Co Zn As Se Rh Pd Cd Sn Sb W T1 and Bi (obtained variously from SPEX Industries Edison NJ USA; Mallinckrodt Paris KY USA; Fisher Scientific Fairlawn NJ USA; and J. T. Baker Solutions). A series of matrix solutions was prepared with an additional 1 3 10 30 100 or 300mgl-' (ppm) of one of the analyte elements in the final dilution. Data are reported here for Li Na Al K Sc Zn Rh Cd T1 and Bi as the matrix elements for a total of 60 matrix solutions plus one 'clean' solution. For the experiments in which the rf generator capacitor position was varied and for those in which ion signals were measured as a function of plasma power and injector flow the analytical solution contained 10 pg 1-l (10 ppb) each of Li Be B Na Mg Sc Co As Rh Ba Ce Tb W Pb and U in 1% nitric acid.Most of these analyte elements were run separately or in groups to ensure that polyatomic interelement interferences did not occur to a significant extent over the plasma conditions studied. Also several of the more abundant isotopes were measured for each element where possible. Thermodynamic data for the elements determined as analyte ions and/or as matrix elements are given in Table 1. The elements cover a wide range of ionization potential atomic mass and heat of vaporization. It might be expected that the element is vaporized in the plasma as a molecule and that atomization follows.It may therefore be more appropriate to consider the heat of vaporization of the molecule. However the identity of the anion is not obvious (nitrate or oxide in nitric acid solution?) and the heats of vaporization of elemental nitrates and oxides do not appear to be readily available. Journal of Analytical Atomic Spectrometry November 1995 VoZ. 10 907Table 1 Thermodynamic properties of elements studied31 Element K Na Li A1 U TI sc Bi Sn Rh c o Fe W B Pd Sb Cd Be Zn Se As Mg Most abundant mass/u 39 23 7 27 238 205 45 209 116 103 24 59 56 184 11 106 121 114 9 64 80 75 Ionization potential/eV 4.339 5.138 5.39 5.984 6.08 6.106 6.54 7.287 7.342 7.46 7.644 7.86 7.87 7.98 8.296 8.33 8.639 8.991 9.32 9.391 9.75 9.81 Heat of vaporization/ kcal mol - 18.88 23.4 32.48 67.9 38.81 80.0 41.1 55.0 127.0 31.5 93.0 84.62 75.0 89.0 46.665 23.86 27.43 14.27 - - - ICP-MS Optimization modes could be obtained by loading the plasma parameter file (including plasma power gas flows and lens voltages) without further operator intervention. In the course of the parametric study presented here the plasma temperature clearly varied substantially as the plasma power and injector flow were adjusted over the wide ranges investigated.Consequently the kinetic energies that the ions gained from the supersonic expansion also varied throughout the course of the parametric experiments. As noted above optimum ion transmission is obtained when the applied lens voltage is comparable to the ion kinetic energy. Therefore the lens voltage should properly have been adjusted for each plasma condition (power and injector flow) set including the determination of the optimum mass-dependent voltage ramp.Some 78 combinations of power and injector flow were studied for each replicate of the parametric study. Even with only one lens voltage to optimize the time required for the experiment would have become prohibitive. Therefore the lens voltage was set to + 3 V throughout the parametric study. This decision holds the ramification that the ion transmission was biased towards plasma conditions yielding ion energies of the order of 3 eV which are cooler than those used under normal analytical conditions. The result is that the parametric curves are biased towards lower power and higher injector flows than they would have been had the ion lens been optimized for each condition.Two standard plasma conditions corresponding to the 'normal' analytical and 'cold plasma' conditions were adopted for most of this work. The plasma and instrumental settings used are RESULTS AND DISCUSSION indicated in Table 2. Initial optimization for the 'cold plasma' conditions was obtained by first setting the plasma power and nebulizer flows setting the lens voltage to + 3 V (fixed) and then adjusting the x-y position (horizontal and vertical pos- ition of the plasma relative to the sampling aperture) for maximum Na + signal and minimum background Ar + signal. The width of the central channel appeared to be more narrow under 'cold plasma' conditions than under 'normal' plasma conditions. These optimum 'cold plasma' operating conditions were confirmed daily.This procedure reproducibly yielded the best analytical conditions for K Ca and Fe and the sensitivity obtained for these elements was reproducible day after day. The appropriate 'cold plasma' conditions typically yielded sensitivities of the order of 1.8 x lo6 counts s-' per ppm for Fe and 30 x lo6 counts s-' per ppm for Li with continuum back- ground signals of less than 5 counts s-l and 40Ar+ (40Ca+) signals for a 0.1% nitric acid solution of less than 3000 counts s - l (less than 1800 counts s-' for DDIW); these ion signals were considered the target for appropriate optimiz- ation. The x-y optimization of the plasma relative to the skimmer as described here required no adjustment when the plasma conditions were returned for 'normal' analytical con- ditions; after set-up the 'cold plasma' and 'normal' operation Table 2 ICP-MS operating conditions Plasma rf power/W Argon gas flow rate/l min-l Plasma (coolant) Auxiliary (intermediate) Nebulizer (central channel) Sampling depth/mm Nebulizer Spray chamber Sample uptake/ml min- ' Desolvation Lens voltage Normal plasma 'Cold' plasma 1200 600 15.0 15.0 1 .o 1 .o 0.77 1.08 9.0 9.0 Cross-flow Scott-type at room temperature 1.0 None Linearly ramped + 3.0 V with ion mass +3.0 V at 7Li+ to +9.0 V at 238U+ (all ion masses) Background Spectra The mass spectrum obtained under 'normal' operating con- ditions for 0.1 YO nitric acid is shown in Fig.3(a). This spectrum was obtained following the matrix study reported below and only a cursory washout of the sample introduction system was attempted; hence some residual signals for the matrix elements Li Na Al Sc and Zn are evident.The dominant ions in the spectrum are O+ and Ar+ at m/z=16 and 40 (and the less abundant Ar+ isotopes at m/z=36 and 38). Other important ions include ArH' and the lower mass ions OH' H20+ or l80+ etc. which presumably derive from the solvent (dilute acid). Of particular note are the large background signals for Ar2+ at m/z=80 and ArO' at m/z=56. The Ar-derived ions Ar' ArH' Ar2+ and ArO+ interfere with the detection of 40Ca 39K 80Se and 56Fe the most abundant isotopes of these elements. For many applications these elements (with the exception of K) can be determined at their less abundant isotopes. In some instances desolvation or mixed gas plasmas can enhance the determination of 56Fe+.However for certain applications such as the analysis of high-purity acids for the semiconductor industry sufficiently low detection limits cannot be obtained under normal analytical conditions. The mass spectrum for the same solution obtained under 'cold plasma' conditions is shown in Fig. 3(b). The dominant background plasma ions are NO' 02+ and H30+ with most Ar-related ions greatly reduced in intensity. The very substan- tial H30 + signal is of itself a strong indication that the plasma is cool. This spectrum differs from that given in ref. 7 with the addition of important peaks in the mass range 16-20u and the persistence of ArH+ at m/z=41. These ions appear to derive from the solvent as no desolvation was used in the present work.The signal remaining at m/z = 56 is thought to be due primarily to contaminant Fe in the water and acid as its magnitude varied with the water distillation batch and with the source and concentration of the acid. This observation of variance in the background at m/z = 56 is the reason the ArO+ signal was not used as a determinant for appropriate 'cold plasma' optimum conditions. The relatively large signal at m/z=80 ascribed to ATz+ seems peculiar since the other 908 Journal of Analytical Atomic Spectrometry November 1995 Vol. 101o'O lo9 108 10' 106 lo5 lo4 lo3 lo2 r '; 10' 5 1 0 0 0 - 104,. 10-5 10' 106 lo5 lo4 lo3 lo2 10' 100 * [ej = 4 x 1 0 ~ ~ ; r = re- = 3500 A A . . . I . 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 Mass Fig.3 Spectra of the plasma background ions for a sample containing 0.1% HNOJ obtained under (a) the 'normal' and (b) 'cold' operating conditions given in Table2. The data were obtained using a simul- taneous dual-mode (pulse counting/analogue) detector. The analogue channel was set to saturate at 2 x lo9 counts s-'; the signals for Ar' ArH' and 0' have saturated the analogue channel for the higher power condition. Note that the vertical scales are logarithmic. Other than the plasma power and injector (carrier or nebulizer) gas flow the plasma conditions (including sampling depth sample introduction and load coil configuration) were the same for the two spectra argide ions (except for ArH') have been very substantially attenuated or eliminated. In a separate experiment this ion also showed an apparently unique dependence on spray chamber temperature.Whereas almost all the other ions (background atomic and polyatomic ions as well as analyte atomic ions) showed little variation with spray chamber tem- perature the m/z=80 ion varied by nearly two orders of magnitude over a temperature range from 0 to +44"C decreasing in intensity as the temperature was increased. It is notable also that the relative intensities of NO+ and 02' were a function of the nitric acid concentration with NO' increasing in magnitude at the expense of 02+ as the acid concentration was increased. For nitric acid concentrations below about 0.3% the increase in the NO+ signal is approxi- mately linear with acid concentration.Above this acid strength the NO' signal increases more slowly. Over the range 0.005-4% nitric acid the sum of the NO' and 02' signals remains relatively constant. It is concluded that the major source of NO+ under the experimental conditions was from the nitric acid rather than from air entrainment and that NO' derives from an ion-molecule reaction involving 02'. Finally the level of ionization in the source plasma appears to be 2-3 orders of magnitude lower under 'cold plasma' conditions relative to normal analytical plasma conditions. This con- clusion is drawn from a comparison of the total ion signals measured for the two plasmas which differ by about this magnitude (1O1O uersus 4 x lo7 s-l). With the extracted ion current reduced substantially it is to be expected that space charge effects in the ion optics are much less important.One of the effects of space charge is to reduce ion transmission efficiency,1g920 and so the ratio of the levels of ionization within the ICP is likely to be greater than the measured total ion signal ratio. Another important result of the reduction of the ion current and hence space charge is that space charge- related matrix effects should be much less significant for the 'cold plasma'. The pressure in the analyser chamber under 'normal' analyt- ical conditions was approximately 1.2 x Torr (corrected for zero-flow de-gassing) while under 'cold plasma' conditions this rose to approximately 2.3 x Torr. Taking the gas kinetic temperature of the source plasma under normal con- ditions as 5300K,21922 the ratio of pressures suggests a 'cold plasma' gas kinetic temperature of about 1450 K.This tempera- ture relates to the plasma directly in front of the sampling aperture which is extracted through the sampler. It is substan- tially lower than the dissociation equilibrium temperature for ArX' species determined by Nonose et all3 with a shielded torch configuration. This suggests that under conditions where the secondary discharge is suppressed the argide ions could derive from the source plasma rather than from within the interface. The cooler plasma temperature helps to explain the difference in lens optimization for the two plasma conditions. When the plasma temperature is cooler the mass-dependent kinetic energy gained from the expansion is less and the slope of a plot of kinetic energy uersus ion mass is smaller.22 Since the voltage applied to the lens appears to be comparable to the ion kinetic energy the reduced mass-dependence of the ion kinetic energy resulting from the cooler plasma results in a lower slope of optimum voltage with mass.For the cold plasma conditions reported here the slope was sufficiently small that a static lens voltage (+ 3 V) provided approximately optimum focusing for ions of all masses. Equilibrium ionization efficiencies for the elements have been calculated by H o ~ k ~ ~ assuming an electron density of 1015 cm-3 and ionization and electron temperatures of 7500 K. Polynomial fits of the electronic partition functions over the temperature range 1500-7000 K for most of the elements listed in Table 1 have been tabulated.24 Calculated degrees of ioniz- ation (DOI) using these partition functions extrapolated to the 7500K temperature assumed by H o u ~ ~ ~ are shown in Fig. 4 where the results are plotted against ionization potential.Similar calculations assuming that the ionization and electron temperatures are equilibrated at the gas kinetic temperature of 1450K derived above yield exceedingly low DOIs. The electron density is important in the calculation of equilibrium ion densities according to the Saha equation and reflects the rate of recombination which reduces the degree of ionization. At high electron density ( lo1' ~ m - ~ ) and low temperature 4 5 6 7 8 9 10 Ionization potentiallev Fig. 4 Degree of ionization plotted as a function of ionization poten- tial for two different plasma conditions of electron number density [e- 1 ionization temperature To and electron temperature T,-.The electronic partition functions used were taken from ref. 24 and were strictly valid only for the temperature range 1500-7000 K Journal of Analytical Atomic Spectrometry November 1995 Vol. 10 909(1450 K) the rate of recombination is much greater than the rate of ionization and the DO1 is consequently greatly reduced. Measured values of the electron density corresponding to the 'cold plasma' conditions have not been reported. Conventional optical methods appear to be limited to electron densities above about lOI3 ~ m - ~ . Values of this magnitude have been reported low in the plasma (7 mm above the load coil) for a 1.25 kW Ar ICP at relatively high central gas flow (1.2 1 min-1)?5 A similar value has been reported for a 0.9 kW plasma at a central gas flow of 1.0 1 min-' and 5 mm above the load It may be expected that under 'cold plasma' conditions where the plasma power is substantially less than that used for the above electron density measurements the electron density will be significantly lower.Furthermore the electron and ionization temperatures are probably not equilib- rated with the gas kinetic temperature. For example under the conditions for which the electron density noted above was measured the electron and gas kinetic temperatures were found to be approximately 3500 and 2700 K re~pectively.~~ The gas kinetic temperature derived from the ratio of pressures is an average temperature for the bulk plasma and other work has shown that the ions may be concentrated in the vicinity of a vaporizing particle.27 Therefore the ionization and electron temperatures may differ significantly from the bulk gas kinetic temperature.Calculated DOIs corresponding to an electron density of only 4 x 10" cm-3 and equilibration of the ioniz- ation and electron temperatures at 3500 K are also presented in Fig. 4. The temperature was chosen assuming local equilib- rium at the ArX' dissociation temperature determined by Nonose et (note that this temperature has been applied to the ICP rather than to the vacuum interface plasma) and the electron density was chosen to yield results which mimic the ionization potential-dependence of the sensitivity data presented below.Under these conditions Fe (ionization poten- tial = 7.87 eV) is approximately 15 % ionized. For the normal plasma the flow through the skimmer is about 1 x lOI9 s - ' . ~ ~ ' For a plasma temperature of 5300 K this corresponds to a sample of about 7.2cm3 s-' from the ICP. If the electron (and ion) density in the ICP is about l O " ~ m - ~ this corresponds to an electron (and ion) flow through the skimmer of about 7 x s-'. The total measured ion current at the detector is of the order of 10" s-' suggesting an over-all transmission efficiency from the skimmer tip to the detector of about Note however that the transmission efficiency for analyte ions is better than this. A fully ionized element of mass 56 u at a concentration in the original nebulized solution of 1 ppm contributes about 1.4 x 10" ions cm-3 to the plasma (assuming a nebulization efficiency of 2% a sample uptake rate of 1 ml min-' a plasma temperature of 5300 K and uniform distribution throughout the central channel at an aerosol gas flow rate of 0.771rnin-').This sample should then yield a flow of about 10" analyte ions s-' through the skimmer. With a sensitivity of about 25 x lo6 counts s-' per ppm (for Fe') this suggests a trans- mission efficiency for Fe' ions of about 3 x and in turn suggests an improvement in transmission efficiency for analyte ions of about 300 relative to that for the background ions (i.e. a corresponding loss mechanism for the background ions). This factor of 300 is consistent with the ion current measure- ments of Gillson et aL2* downstream of the skimmer for which the measured differential U+ current for 0.04 mol U 1-' was the same as the calculated U+ current through the skimmer aperture whereas the total ion current was reduced to 6 pA from the calculated 1500 PA.If the 1 ppm Fe solution is nebulized with the same efficiency into the 'cold plasma' (1.1 1 min-' aerosol gas flow rate and plasma gas kinetic temperature of 1450K) and were 15% ionized (as calculated for an electron density of 4 x 10'O cmV3 and local thermal equilibrium of the ionization and electron temperatures at 3500 K) it would contribute about 6 x lo9 ions cm-3 to the plasma. The sensitivity observed for Fe was 1.8 x lo6 counts s-' per ppm. This corresponds to about 5% of the total ion current measured (4 x lo7 counts s-').If the transmission efficiency for Fe' is the same as for the plasma ions (i.e. if there is no preferential loss of background ions) then the total ion density in the plasma should be about (4 x 107/1.8 x lo6) x 6 x lo9 z 10" ~ m - ~ which is close to the assumed electron density. For the 'cold plasma' the increase in the operating pressure suggests that the flow through the skimmer is doubled to about 2 x loi9 s-'. For an average gas kinetic temperature of 1450K this corresponds to a sample of about 4cm3 s-l from the ICP. For an electron (and ion) density in the ICP of 4 x 1O1O cmP3 this corresponds to an electron (and ion) flow through the skimmer of about 1.6 x 10" s-'. The over-all transmission efficiency is then about 3 x approximately 300 times improved over the normal plasma.This improvement in over-all transmission efficiency may result from reduced space charge effects because the ion current is r e d ~ c e d . ~ ' . ~ ~ The transport efficiency of the analyte Fe' under 'cold plasma' conditions is then approxi- mately the same as in the normal plasma and the ratio of sensitivities (1.8 x lo6 versus 25 x lo6 counts s-' per ppm) is primarily accounted for by the reduced degree of ionization in the 'cold plasma'. These crude estimates of transmission efficiency for the cold plasma clearly depend on the assumed electron density (both for the direct calculations and for the DO1 used). It is worth noting that if the electron density and temperature are this low then the Debye radius at the skimmer is compar- able to the skimmer aperture diameter which may have important ramifications for the dynamics of ion transport through the interface.20 Parametric Study of Plasma Power and Injector Flow The variation of dominant plasma ions and trace analyte ions as a function of injector flow (with a constant 0.45 1 min-' introduced through the nebulizer) at a plasma power or 1200 W is shown in Fig.5. The Ar' signal decreases as the injector flow increases presumably reflecting the decrease in the plasma temperature owing to the translation of the normal analytical zone towards (and past) the sampling orifice. This decrease in plasma temperature is further evidenced by the increase in polyatomic argides (Ar2' ArH' and ArO') with injector flow.At very high injector flows an increase in NO+ and 02' is observed. As noted above the relative intensities of NOf and 02+ are a function of the nitric acid concentration suggesting that the dominant source of the NO' is the acid introduced with the sample. Other data presented here suggest that the change in dominant ion is a result of the cooling of the plasma with the concomitant survival (or production) of neutral molecular species including NO and 02. The various analyte ions optimize at more-or-less the same injector flow nearly independent of ionization potential or mass. There may be a trend favouring higher injector flows for lower mass ions which might reflect mass-dependent radial diffusion within the plasma.30 This effect would result in a higher density of low mass ions slightly earlier in the plasma at higher injector flow.The optimum injector flow observed for analyte ions in Fig. 5(b) is higher than the optimum indicated for 'normal' plasma conditions in Table 2. This is a direct result of having fixed the lens voltage at + 3 V for this experiment as discussed under Experimental. The uniformity of analyte ion optimiz- ation with injector flow shown in Fig. 5(b) is in contrast to the element-dependent optimization observed by Sakata and Kawabata (cJ Fig. 5 of ref. 11 obtained without the ShieldTorch). Comparison of these results indicates that the occurrence of a secondary discharge perturbs the distribution 91 0 Journal of Analytical Atomic Spectrometry November 1995 Vol. 101.2 1 .o 0.8 0.6 0.4 - 2 0.2 5 0.0 0 v) .- .- 0.0 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 Injector fiow/i min-' Fig.5 (a) Variation of plasma background ions with injector (carrier or nebulizer) gas flow at 1200 W plasma power. The injector flow includes 0.45 1 min-' passed through the nebulizer and an adjustable make-up flow added downstream of the spray chamber. (b) Variation of trace elemental ions under the same conditions of ions either within the plasma or during extraction; a low plasma potential has the advantage of a more uniform ioniz- ation region hence more uniform optimization. Corresponding data obtained at 600 W are given in Fig. 6. At this low plasma power there is a very clear separation of Ar-derived plasma ions (at lower injector flows) and NO+ and 02+ (at higher injector flows).The optimization of analyte species now also shows a strong dependence on injector flow with elements having higher ionization potentials or heats of vaporization appearing at lower injector flows and more easily vaporized and ionized elements appearing at higher injector flows. The correlation with plasma temperature is easy to draw but there may be an over-riding dependence for the low ionization potential elements on the concomitant appearance of NO' and 02+ as dominant plasma ions. It is possible then that the separation on the basis of ionization potential may be due to plasma temperature (thermal ionization or electron impact ionization) or it may be due to a change towards dominance of ion-molecule reaction ionization (charge transfer) of the lower ionization potential elements with NO + and 02+ as those ions become important at lower temperature.The W' ion optimizes only at lower injector flows despite having a moderate ionization potential. The heat of vaporiz- ation for W is not available. However the melting-point of W is very high (3650 K3'). It is probable that the hotter plasma conditions at lower injector flows are required to vaporize and atomize this element. Some of the elements show local maxima at both low and high injector flows. The increase in response at low injector flows for Rh+ and Co' species may reflect their relatively high heats of vaporization (the highest of those known for the suite of elements investigated). The Mg' ion also shows a minor local maximum at low injector flow.This might suggest that Mg can be vaporized as either the atom (which has a low heat of vaporization and may appear at .N 1.2 Q - E g 1.0 0.8 0.6 0.4 0.2 0.0 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 Injector fbwA min-' Fig. 6 (a) Variation of plasma background ions with injector (carrier or nebulizer) gas flow at 600 W plasma power. The injector flow includes 0.45 1 min-l passed through the nebulizer and an adjustable make-up flow added downstream of the spray chamber. (b) Variation of trace elemental ions under the same conditions higher injector flow rates) or as the oxide MgO (which has a high melting-point and may therefore be vaporized only at lower injector flow rates). The appearance of two local maxima (with respect to injector flow) of the same atomic analyte ion may indicate either a change of source of ionization or a change in the source of vaporization and atomization.The intermediate behaviour of Be having a high ionization poten- tial is difficult to explain. These data are comparable to the ShieldTorch data presented in Fig. 6 of ref. 11 with the import- ant addition here of the rise to dominance of NO+ and 02+ at high injector flow. Nonose et ~ 1 . ' ~ also observed the distinc- tion of the argide species from NO' and 02+ on the basis of injector (carrier) gas flow under conditions where the secondary discharge was minimized although that work was performed at higher plasma power. The variation of ion signals with plasma power at an injector flow of 1.10 1 min-' (corresponding to the 'cold plasma'injector flow) is given in Fig.7. The argide ions are important at higher plasma powers and the oxides and molecular ions at lower powers. The Ar2+ ion shows a bimodal optimization appearing at both high and low plasma power perhaps reflecting its source (Ar') at higher power but its persistence owing to less fragmentation at lower power and temperature. Again the more easily vaporized and ionized elements optimize at lower plasma powers corresponding either to lower plasma tempera- tures or to the appearance of NO' as a reactant ion. Elements having a higher ionization potential or vaporization energy appear at higher plasma powers. The behaviour of the analyte ions shown here notably the appearance of W' only at higher power and the optimization of Be+ at intermediate power correlates with their responses to injector flow in Fig.6(b). The decrease in the Ar+ signal as the injector flow is increased or the plasma power is decreased may be due to the cooling of the plasma. The increase of the polyatomic argide ions (ArO' and Ar2') at intermediate injector flow or power Journal of Analytical Atomic Spectrometry November 1995 Vol. 10 91 1500 600 700 800 900 1000 1100 1200 1300 1400 1500 Plasma powerMl Fig.7 (a) Variation of plasma background ions with plasma power at injector (carrier or nebulizer) gas flow of 1.10 I min-l. The entire injector flow passed through the nebulizer and no make-up gas was added. (b) Variation of trace elemental ions under the same conditions may reflect the combined effects of the reduced source ions (0' and Ar') and the persistence of the polyatomic ions resulting from their reaction with Ar at the lower plasma temperature.These effects have opposite temperature depen- dence and could give rise to the appearance of the ionic species at intermediate plasma temperatures. It could be postu- lated that the appearance of NO+ and 02+ at even cooler plasma conditions is a result of the production or persistence of neutral H20 NO or 02 and their subsequent ionization by electron impact. Primary ionization of NO or 0 by electron impact may lead directly to NO+ and 02+. Primary ionization of atoms (H or 0) may be followed by ion-molecule reactions leading to the prominence of NO+ and 0,' accord- ing to H++02+02++H (k=l.l7x 10-9)32 H+ +H20-+H20+ +H (k=8.2 x O+ +H20-+H20+ +O H20+ +02+02+ +H20 H20+ + H20+H30+ +OH (k=2.3 x 10-9)33 ( k z 2 x 10-10)34 (k= 1.7 x 10-9)33 H30+ +NO +X+NO+ + H20 +XH 0,++N0-+NO++O2 ( k = 4 .5 ~ 1 0 - ~ ' ) ~ ~ 0,' +N+NO+ +O (k= 1.2 x where the known rate constants k are given in units of cm3 molecule- s- at room temperature. The reaction forming NO+ from H30+ is a two-step process involving the formation of an intermediate adduct ion and may involve a radical (X = OH 0 or H).37 These reactions have been postulated to be important in flame^^'.^^ for which the flame temperature is of the order of the 'cold plasma' temperature inferred above. The intermediates H,O+ and H30+ are observed as major ions in the 'cold plasma' mass spectrum of Fig. 3(b). Whether NO+ and 0,' are formed by primary electron impact ionization or through subsequent ion-molecule reaction chemistry the very large signal for H30+ which can only be produced by proton transfer indicates that ion-molecule chemistry does play an important role in determining the ionic composition of the 'cold plasma'.Charge transfer ionization of metal atoms by NO+ or 02+ is exothermic for atoms having ionization potentials less than those of NO or O2 (9.26436 and 12.071 eV re~pectively~~). Because these ions are molecular they have multiple internal degrees of freedom (rotation and vibration) that relax the requirement for electronic energy level resonance that charac- terizes atomic ion-atom charge transfer reactions. Therefore NO + and 0,' could behave as relatively indiscriminate charge transfer reactant ions for metal atoms.In flames however the primary source of alkali atomic ions is dissociative charge transfer with H30+ according to4' H30+ +A+A+ +H,O + H This reaction is exothermic for atoms (A) having ionization potentials less than 6.4 eV (being the difference between the ionization potential of H 13.595 eV,31 and the proton affinity of H20 7.22eV41). The importance of this reaction in flames is due in part to the preponderance of H 3 0 + . For most flames NO+ and O,+ are much less abundant and their role in elemental ionization is less well known. By direct analogy with flames it is to be anticipated that ion-molecule chemistry principally charge transfer involving H,O+ NO+ and 0,' can provide an important source of ionization for elemental atoms in the 'cold plasma'.Sensitivity Under normal plasma conditions sensitivity is a reasonably smooth function of analyte mass increasing with mass as shown in Fig. 8 where the sensitivity has been corrected for natural abundance of the isotopes. The ion density in the plasma is expected to be a function of the molarity of the solution although in ICP-MS the sensitivity is often quoted in units of mass/volume. The decrease in sensitivity at low mass is partially due to the enhanced radial diffusion of low mass analyte elements in the plasma,3o but is probably more significantly influenced by space charge effects downstream of the skimmer aperture.19i42 As can be seen in Fig. 8 sensitivity under 'cold plasma' conditions does not appear to show a strong mass dependence.The data of Fig. 8 are presented in Fig. 9 as a function of the ionization potential of the analyte elements. Significant scatter is observed for data obtained under normal plasma conditions reflecting the over-riding 0 50 100 150 200 d z Fig. 8 Sensitivity (corrected to 100% abundance) for trace elements as a function of atomic ion mass-to-charge ratio. Open circles were obtained under normal plasma conditions and filled circles were obtained under 'cold plasma' conditions 91 2 Journal of Analytical Atomic Spectrometry November 1995 Vol. 10(4 x lo1' cmd3) and ionization temperature (3500 K). In fact there is a relatively restricted range of ionization temperature and electron density that is consistent with the observed response curve assuming equilibrium thermal ionization.Assuming that elements having ionization potentials below 6 eV are completely ionized the data of Fig. 9 suggest that an element having an ionization potential of 8 eV is between 5 and 40% ionized and an element with an ionization potential of 9.5 eV is 0.02-0.5% ionized. The ionization temperatures and electron densities that would provide these degrees of ionization can be calculated. These results are presented as the four curves of Fig. 10 where the partition functions of the atom and ion have been assumed to be equal in order to simplify the calculation. The region bounded by these curves indicates the ranges of ionization temperature and electron density which could result in the observed response curve. These ranges can be further restricted by recognizing that the maximum electron density cannot exceed approximately 4 x 10l2 cm-3 (being the ratio of the total ion signals measured for the cold and normal plasmas multiplied by the known electron density of about 10'' cmT3 under normal plasma conditions).Furthermore the electron density cannot be less than about 108cm-3 being the total ion signal measured under cold plasma conditions (4 x lo7 s-l) multiplied by 10 (assuming that the transmission efficiency of the quadrupole is and that this ion flow also represents the minimum number of electrons per second extracted through the skimmer) divided by the neutral gas flow extracted from the plasma through the skimmer (about 4 cm3 s-l as derived earlier). The ionization temperature characterizing the cold plasma response curve is then in the range 2900-4400K (indicated by the shaded region of Fig.10). Again this ionization temperature range is consistent with the dissociation equilibrium tempera- ture for ArX' ions determined by Nonose et all3 It could therefore be argued either that trace element thermal ionization is approximately equilibrated at low charge density or that the mechanism of trace element ionization changes to ion-molecule charge transfer under the cooler plasma conditions. The latter mechanism can explain the bimodal optimization of several of the elements [Figs. 6(b) and 7(b)] on the basis of the change of mode of ionization. 4 5 6 7 8 9 10 Ionization potentiaVeV Fig. 9 Sensitivity (corrected to 100% abundance) for trace elements as a function of ionization potential of the trace element.Open circles were obtained under normal plasma conditions and filled circles were obtained under 'cold plasma' conditions effect of ion mass. Little explicit dependence on ionization potential is observed for normal conditions at least for ioniz- ation potentials <10eV. On the other hand the sensitivity obtained under 'cold plasma' conditions is clearly a function of ionization potential with a significant decrease in sensitivity for ionization potentials above about 8eV. It is notable that the sensitivities for T1 and Bi (high mass) are comparable to those of Li Na and K (low mass) at least within a factor of about four on a molar basis; these elements all have ionization potentials <7.5 eV. The much reduced mass bias (relative to normal conditions) supports the expectation that space charge effects should be less significant for the lower ion current condition. The improved mass bias may also derive in part from reduced radial expansion of the central channel and mass-dependent diffusion in the ICP because the plasma is cooler and sampled earlier relative to the initial radiation zone. It was observed that the sensitivity to trace elements (e.g.Fe Ca K Na and Li) under 'cold plasma' conditions was insensi- tive to the concentration of the acid at least for acid concen- trations below 1 YO. Since the relative proportions of NO' and 02' are a function of the nitric acid concentration this observation suggests that chemical ionization by charge trans- fer with NO+ is not the determinant for sensitivity or might suggest that NO' and 02+ behave equivalently as charge transfer reactant ions.Cooler plasma conditions promote the formation of polya- tomic ions notably the oxide ions of refractory elements. The sensitivity shown for Sc (45 u ionization potential = 6.54 eV) is lower by several orders of magnitude than that expected on the basis of its ionization potential partially because it appears almost exclusively as ScO'. The data given in Figs. 8 and 9 report measurements only for the atomic ion and the elements determined were chosen as being relatively poor oxide-formers. The anomalously low sensitivity to W is probably due to inefficient vaporization as discussed above and noted by others under similar conditions." Oxide or other polyatomic ions of W were not observed at significant intensity.Data given in Figs. 6(b) and 7(b) show that W is efficiently determined only under conditions yielding higher plasma temperatures (higher power and lower injector flow). For elements having ionization potentials below about 6 eV the sensitivities obtained under 'cold plasma' conditions are comparable to those obtained under normal plasma conditions. For elements having ionization potentials between 7 and 8 eV the cooler plasma conditions yield sensitivities that are up to two orders of magnitude lower and this discrepancy increases to 3-4 orders of magnitude above 9 eV. The general form of this response curve was simulated assuming equilibrium ioniz- ation in Fig. 4 for conditions of rather low electron density 4000 c C 3500 5 3000 .- w i Maximum [e-] from ratio of total ion signals I I I I *-*I __- = mi *I- ___- I ' 2500 Electron density/cm* Fig.10 Calculated ionization temperature To, and electron density [e-1 ranges that are consistent with the observed 'cold plasma' response curve assuming thermal equilibrium ionization. The curves indicate the Ton and [e-] which would result in 5 and 40% ionization of an element having an ionization potential (IP) of 8.0 eV and 0.02 and 0.5% ionization of an element having IP = 9.5 eV. The calculations assume that the electron and ionization temperatures are equilibrated. These degrees of ionization bound the observed responses shown in Fig. 9. The minimum and maximum electron densities have been derived as discussed in the text.The shaded range is consistent with the observed response curve. Because the elements having these ionization potentials are generic the electronic partition functions of the atom and ion have been assumed to be equal Journal of Analytical Atomic Spectrometry November 1995 Vol. 10 91 3There does not appear to be a discontinuous change in sensitivity near 6.4 eV and so charge transfer with H30+ does not appear to be a dominant ionization mechanism (at the least it will not apply to elements having ionization potentials greater than 6.4eV). Charge transfer with NO+ and 0,' remains a potential source of ionization. However this ioniz- ation mechanism does not explain the change in sensitivity that occurs near 8 eV which is consistent with thermal ionization.Load Coil Ground Reference and Secondary Discharge The earlier reports on low power/high injector flow oper- ation7*11*13 have all indicated the requirement for reduction or elimination of the secondary discharge to facilitate reduction of the Ar-related background. For the instrument used in this work the plasma potential is maintained at a low level to minimize the secondary discharge under all operating con- ditions by appropriate ground-referencing of the load coil. This ground reference is determined in part by the position of the capacitor plate CP1 relative to CP2 and CP3 in the Colpitts oscillator (Fig. 2). As CP1 is moved towards CP2 the ground reference is moved along the load coil away from the front of the load coil (adjacent to the sampler orifice).Adjusting the capacitor in this manner is expected to increase the plasma potential thereby enhancing the formation of a secondary discharge. The data in Fig. 11 were obtained by adjusting the position of CP1; the position denoted '0' is that used in normal operation to minimize the secondary discharge and provide optimum sensitivity. The plasma dc potential for this position of CP1 inferred from the intercept of ion kinetic energy versus ion mass for a similar ICP-MS system (Elan 5000) is approxi- mately + 3 V.22 For that instrument which shows less discrimi- nation against high energy ions the plasma dc potential was observed to increase non-linearly with a negative displacement of CP1. At a position thought to yield a strong secondary discharge the plasma dc potential was +22 V.In the present work moving the ground reference away from the sampler (in the negative direction) results in an increase in sensitivity for Co but a larger increase in the background signals for Ar' and ArO ' . The position for optimum determination for Fe ' defined as the maximum in the ratio of Fe+ ArO' (approxi- mated as Co' ArO') is close to the '0' position. The relatively monotonic increase in signals with the change in CP1 position to negative values is abruptly discontinuous near the position marked - 10 mm. At this point all the background ion signals dramatically increase with the exception of NO' which decreases markedly in concert with the analyte ion Co' (which is typical for that of atomic elemental ions derived from the -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 Capacitor positiodmm Fig.11 Analyte ion (Co') and plasma background ion (Ar+ ArOf NO+ and 02+) signals as a function of the position of the Colpitts oscillator capacitor plate CP1 (see Fig. 2) which determines the position of the ground reference along the plasma load coil sample). The data are interpreted to indicate a gradual increase in the intensity of a secondary discharge (or an equivalent effect) as the capacitor position is adjusted to move the ground reference away from the sampler with a sudden onset of a strong discharge near - 10 mm. The dramatic reduction in the signals for NO+ and Co+ at this point is probably due to a sudden change in ion kinetic energy resulting from the second- ary discharge.13 It was noted above that ion transmission through the ion optics is a strong function of ion kinetic energy with optimum transport for ions having kinetic energies comparable to the applied lens voltage.The reduction in the NO' and Co' signals probably reflects an increase in ion energies above the bandpass of the optics rather than a relative decrease in the ion densities in the plasma. At each position of CP1 an attempt was made to optimize the lens voltage for ion transmission. The optimum voltage was found to remain constant for Co' throughout the range of this experiment while that for Ar' was found to increase (from +3 to about + 6 V at the -9.5 mm position although the results reported were obtained at a constant - 3 V). The optimum lens voltage appears to be approximately equivalent to the ion kinetic energy,18 suggesting that the ion energy for Co' did not change significantly through the range of this experiment while that for Ar' did change somewhat.However the range of voltages available was limited to less than + 17 V. If the ion energy distribution was bimodal (corresponding to ions created in the source plasma and to those created in the secondary discharge) and the ion energies for ions created in the secondary discharge were above + 17 eV then these latter ions would not be effectively transmitted through the ion optics. That is the ion optics employed here discriminate against high energy ions created in a secondary discharge. Therefore the increase in the background ions near the capaci- tor position -10mm together with the decrease at that position for NO+ and Co' suggests that there was a very substantial change in the ion distribution within the sampled plasma strongly favouring argide and other background ions at this position. This effect would probably be more apparent using ion optics that did not discriminate against the higher energy ions.Concomitant Element Matrix Effect The initial publication on cooler plasma operation7 indicated a self-induced matrix suppression effect for K beginning at a concentration of 10mg 1-l (ppm). Since the total ion signal measured under 'cold plasma' conditions is significantly lower than that obtained under normal conditions it can be implied that the ion current through the skimmer tip is also less (perhaps less than the detected ion ratio as indicated above).This suggests that matrix effects resulting from space charge in the ion beam within the ion optics should be less evident. Therefore the appearance of a self-induced matrix effect at rather low concentration might provide some fundamental information regarding the ion dynamics within the plasma source. Reported here are the results of a study of matrix effects under 'cold plasma' conditions intended to provide insight into the mechanism of ionization and ion interaction within the plasma. Mass-dependent matrix effects (depending on the mass of both the analyte and the matrix elements) are expected to arise from space charge effects within the ion beam in the ion optics. Matrix effects that are a function of the ionization potentials of the analyte and matrix elements may be ascribed to ioniz- ation suppression in the plasma.When the sample is introduced as an aerosol the droplets dry to a particle before vaporization and ionization and the size of the dried particle is a function of the salt content of the sample. A matrix effect could then arise owing to incomplete vaporization of the dried particle if 91 4 Journal of Analytical Atomic Spectrometry November 1995 Vol. 10the heat transfer from the plasma is limited and this effect would depend on both the concentration and the heat of vaporization of the matrix element. There remains also the possibility of fractionation in the vaporization process whereby some elements are more readily vaporized than are others even from the same dried particle. Therefore a suite of analyte and matrix elements were selected which covered a range of mass ionization potential and heat of vaporization.Fig. 12 presents the measured ion signal (corrected for isotopic abundance) as a function of the concentration of the elements treated as concomitant elements. The data include results for the 60 different matrix solutions each including only one element at 'high' concentration (covering the range from 1 to 300 mg 1-l) plus the 'clean' 10 ppb standard solution. Since Sc was observed primarily as its monoxide ion the response for ScO' as a function of Sc concentration is given. The ion signal observed is more-or-less a function of the ionization potential of the matrix element. The response curves show two substantially linear regions.In all instances the signal is linear up to an ion signal of about lo7 counts s-'. For matrix elements having ionization potentials below 6.0 eV a second linear region having lower slope appears above about 5 x lo7 counts s-'. The intersection of these linear sections occurs near 3 x lo7 counts s-l which is approximately the ion signal observed for NO' in the blank solution (1% HNO for which NO' is the dominant ion being approximately a factor of four more intense than 02+). For matrix elements having ionization potentials above 6.0eV the slope of the second response region appears to approach zero. Fig. 13 shows normalized ion signals for trace elements as a function of the Rh' signal for solutions containing Rh as the matrix element.No clear evidence of analyte mass-dependence is apparent. In fact the trace elemental ions appear to be suppressed in a rather uniform manner. Nonetheless it is apparent that the use of a single internal standard to correct for the matrix suppression would result in significant error (of the order of a factor of three at best). All the matrices studied showed similar behaviour. The extent of suppression for a given analyte ion is a function of the concomitant matrix element. This is shown in Figs. 14 and 15 where the Co' signal is plotted against matrix concentration and matrix ion signal respectively. If the vaporization of the particle plays an important role in the matrix effect the effect should be a Molarity of matii Fig. 12 Matrix ion signals (corrected to 100% abundance) as a function of concentration of the matrix element.The data presented were obtained for 61 separate solutions comprised of six solutions each of 10 matrix elements plus the 'clean' 10 ppb standard solution. Response curves are indicated by drawing lines through the signals for the seven solutions for each of the matrix elements corresponding to concentrations of 0.01 1,3 10 30,100 and 300 ppm. The data were obtained for the most abundant atomic ion for each matrix element (with the exception of Sc for which ScO+ was measured) and were corrected for 100% natural abundance 100 - P) C .- .Q 10'' .- E E b z 10-2 10 10' Rh matrix ion signakounts s-' Fig. 13 Ion signals for trace elements (10 ppb each) as a function of Rh ion signal for solutions of different Rh concentration.The analyte ion signals were normalized to their intensities in the lowest Rh concentration 0.01 ppm 10" 1b -5 Matrix molarity 10 -2 i Fig. 14 Ion signal for Co' (added as a trace element at 10 ppb) measured for various matrix solutions as a function of matrix concen- tration. Lines are drawn through the signals obtained for the seven solutions for each of the matrix elements corresponding to concen- trations of 0.01 1 3 10 30 100 and 300 ppm l o 4 10 u lo6 10 10 * Matrix ion signallcounts s-' Fig. 15 Ion signal for Co' (added as a trace element at 10 ppb) measured for various matrix solutions as a function of matrix ion signal (corrected to 100% abundance). The data are the same as those used in Fig. 14 but are plotted here against matrix ion signal function of the matrix concentration (Fig.14) with a more severe suppression for matrix elements with high heats of vaporization (e.g. Rh and Sc). No strong correlation is observed. There is no evidence of a change in suppression for matrix elements having ionization potentials above or below that of the analyte element Co (ionization potential = 7.86 eV). The suppression of the analyte signal does not appear to correlate with the absolute ion signal measured for the various Journal of Analytical Atomic Spectrometry November 1995 Vol. 10 91 5matrix ions (Fig. 15). The best correlation appears to be with ionization potential of the matrix element with lower ioniz- ation potential matrix elements causing more severe suppres- sion.The greater suppressions observed for Bi T1 and Rh appear to be inconsistent and may reflect a mass bias as well (although it is noted that Rh also has a high heat of vaporiz- ation which may also play a role). It remains a possibility that all of these effects are simultaneously important so that the correlation with any one is not good. As with the data of Fig. 13 the trends (or lack thereof) observed for analyte Co' with matrix ion signal were also observed for the other analyte ions. The ion signals measured for Bi' and the major background plasma ions as a function of Bi concentration are given in Fig. 16(u). The matrix ion signal increases linearly with matrix concentration then becomes non-linear and finally shows a plateau at high matrix signal. It is apparent that the matrix ion signal is anti-correlated with the background ion signals and that the non-linearity in the matrix ion response curve appears where the matrix ion signal exceeds the blank NO' signal.This cross-over might imply that NO' acts as the ionization source for Bi (ie. by charge transfer) which in turn suggests that the 'cold plasma' is dominated by ion-molecule chemistry. The similar suppression of the 02' and H20+ signals could indicate that charge transfer from these ions is also important. The H,O+ signal is expected to follow that for H20' since the latter ion is probably the primary source for proton transfer leading to H30+ (charge transfer of H,O+ with Bi is not thermodynamically allowed). Alternatively the common suppression of the background ions may indicate an approach to equilibrium with Bi' being the thermodynamically favoured terminal ion.However chemical equilibrium should result in a large difference in suppression of analyte ion signals 10 10 10 c v) 3 10 a 9 2 109 Molarity of Bi matrix element cn C .- - 108 10 7 106 105 10 -4 lo9 10" Molarity of Na matrix element 10" Fig. 16 (a) Bi' signal and dominant plasma background ion signals as a function of the concentration of Bi as a concomitant matrix element. Bi acts as a 'high ionization potential' matrix element (see text). (b) Na+ signal and dominant plasma background ion signals as a function of the concentration of Na as a concomitant matrix element. Na acts as a 'low ionization potential' matrix element for analyte elements having ionization potentials above or below that of the matrix element; this was not observed in the data of Figs.13-15. Finally the common suppression may indicate enhanced ion-electron recombination if the matrix element induces an increase in the electron density. The data shown in Fig. 16(a) were similar to those obtained for all matrix elements having ionization potentials greater than 6.0 eV. However lower ionization potential elements showed behaviour similar to that shown in Fig. 16(b) for Na as the matrix element. Here the NO' signal behaves as it did for the higher ionization potential elements but the other back- ground ions show first a modest attenuation at moderate matrix concentration (about low3 mol l-') then a recovery as the matrix concentration was increased further. The latter results argue against equilibrium amongst the background ions. For matrix elements having ionization potentials above 6.0 eV the argide ions (Ar' ArH' and ArO') behaved much as did the other trace and background ions. Typical results (obtained for Bi matrix) are given in Fig.17(u) which is analogous to Fig. 16(u). The background argide ions are sup- pressed by the concomitant matrix element although appar- ently not as severely as are NO' and the analyte ions. However for matrix elements having ionization potentials < 6.0 eV the background argide ions are significantly enhanced at high matrix concentration. The data given in Fig. 17(b) are typical of those observed for the low ionization potential matrix elements K Na and Li.For these matrix elements the argide ions are not suppressed at moderate matrix concen- tration; the initial suppression of the ArO' signal is almost certainly due to the suppression of the isobaric Fe+ (Fe was added as an analyte element at 10 ppb). Even for a K S ! Molarity of Bi matrix element .- w 107 u) E - 104 103 los5 10'~ 1 0 ' ~ lo-* Molarity of Na matrix element Fig. 17 Argide ion signals obtained as a function of matrix element concentration obtained in the same experiment as the data of Fig. 16. (a) Bi as a concomitant element. (b) Na as a concomitant element. In both instances the ArO+ signal is dominated by Fe+ added as a trace element at 10 ppb. The initial decay of the ArOf/Fe+ signal in (b) is almost certainly due to the suppression of the Fe' signal and the recovery of the signal is almost certainly due to the enhancement of the ArO' component 91 6 Journal of Analytical Atomic Spectrometry November 1995 Vol.10concentration as low as 30ppm the background signals are significantly enhanced. As noted earlier the lower ionization potential matrix elements also appear to show a second linear response region with a smaller (but non-zero) response factor beyond the concentration yielding a matrix ion signal comparable to the blank NO' signal. It is likely that this second region is a result of an additional or enhanced mechanism of ionization peculiar to matrix elements having low ionization potentials. The transition from 'high ionization potential' behaviour (sup- pression of all background ions) to 'low ionization potential' behaviour (modest suppression and then recovery of back- ground ions other than NO') occurred abruptly near an ionization potential of 6.0 eV; where A1 (ionization potential = 5.984 eV) showed substantially 'low ionization potential' behaviour TI (ionization potential = 6.1 eV) showed strictly 'high ionization potential' behaviour.The dependence has been ascribed to ionization potential rather than mass since 39K behaved strongly as a low ionization potential element and lo3Rh behaved strongly as a high ionization potential element although they differ only by a factor of about 2 in mass. C4%c (ionization potential = 6.54 eV) and 66Zn (ionization poten- tial = 9.391 eV) also appeared to show 'high ionization poten- tial' behaviour although the extent of suppression (about a factor of 3 at the highest concentrations studied) was not sufficient to be certain.] It is significant also that the ion signal obtained at high matrix concentration for some of the elements substantially exceeded the total ion signal for the blank solution.Thermal ionization of high concentrations of low ionization potential elements may substantially increase the electron density in the plasma. For example complete ionization of 30 ppm of K corresponds to an increment of 1.5 x 1014 ions and electrons per second in the plasma (assuming 2% nebuliz- ation efficiency at a sample uptake rate of 1 ml min-'). If these ions and electrons are uniformly distributed within the central channel flow the electron density would be increased by about 10l2 C M - ~ (assuming 1.1 1 min-l and a plasma temperature of 1450K).This is probably a significant enhancement in the electron density. The local plasma temperature could be increased owing to collisional heating involving the electrons accelerated in the rf field and to improved radiative transport. Miller et aE.43 discuss these effects for the dc plasma. The inversion of local electron densities by easily ionized elements (EIE) resonance absorption raises the optical absorption cross- sections for both EIE and Ar. This improves the rate of energy transport and enhances local ohmic heating. Enhancement of the electron density on-axis by the presence of 0.5 mol 1-1 Cs in the nebulized sample has been reported by Caughlin and Blades!4 Of particular interest is the report of Hanselman et u I .~ ~ specifically the data relating to the relatively cool conditions of 1.25 kW plasma power 1.2 1 rnin-l central gas flow with measurements made 7mm above the load coil. An increase in electron density on-axis was observed with the addition of 0.1 mol I-' Cs Li or Ag to the sample while the electron and gas kinetic temperatures were enhanced only for Cs (suppression was observed for Ag and no change for Li). If under 'cold plasma' conditions the plasma temperature is increased by the presence of an EIE this could then enhance thermal ionization and possibly further increase the electron density. This mechanism might explain the additional or enhanced mechanism of ionization in the presence of high concentrations of low ionization potential matrix elements.It does not however explain why most of the background ion signals increase under these conditions but NO' does not. It therefore appears that the ionization mechanism in the 'cold plasma' may involve ion-molecule chemistry in addition to thermal ionization but that there may also be an additional or enhanced mechanism of ionization at high concentrations of low ionization potential elements. It is this enhanced ioniz- ation that is responsible for the second region of linearity in the K-matrix data seen both in this work and in the original work of Jiang et aL7 The concern for analytical use is that the increase in the background signals under these conditions could confound the determination of trace elements (e.g. the determination of Fe in a sample containing a high concen- tration of a low ionization potential concomitant element for which the ArO' background signal is increased).The preceding results show that matrix suppressions are rather severe (occur at low matrix concentration) under 'cold plasma' conditions. Furthermore if the matrix element has a low ionization potential and is at sufficient concentration there is a possibility of increased interference from polyatomic background ions. Finally refractory elements form polyatomic ions (e.g. ScO' from Sc) readily under these plasma conditions and these can also interfere with trace determinations. While the application of 'cold plasma' conditions for the determi- nation of certain elements (e.g. K Ca and Fe) in ultrapure samples is clear it remains to be determined whether there is an analytical protocol that can be established to extend its application to moderate (e.g. 300ppm or less) salt content samples.As was seen in Figs. 14 and 15 an analyte ion signal (Co' in that instance) is suppressed to different extents for different matrix elements and concentrations. However it was noted that the phenomenology of Fig. 13 was duplicated for various matrix elements. As shown now in Fig. 18 the relative suppres- sion for a given analyte element in various matrices is relatively invariant. Normalization of the analyte signal (e.g. Co') to an internal standard (e.g. Rh') minimizes the dependence of the matrix effect on the identity of the matrix element although a matrix effect which is a function of the concentration of the matrix element remains.This is unlike operation under normal plasma conditions where the extent of analyte suppression and the ratio of responses for analyte ions of different masses is strongly dependent on the mass of the matrix element. Although the analyte (Co') and internal standard (Rh+) used in Fig. 18 differ in mass by only a factor of two and have similar ionization potentials the same correlation was observed for any analyte ion using any other as internal standard (provided that there was no contamination of the analyte or internal standard in the matrix solution and there was no isobaric interference). More significantly there appears to be no need to add an internal standard. Fig. 19 gives data comparable to those of Fig. 18 but using the background ion NO' as the internal standard.Clearly NO' is suppressed by the matrix in much the same manner as the analyte ions. This is to be expected if the ionization mechanism is charge transfer 1 0 ' 1 0 2 10 Rh ion SignaVcounts s-' Fig. 18 Correlation of Co' and Rh+ ion signals (both present as trace elements at 10 ppb) in different matrix element solutions at different matrix element concentrations Journal of Analytical Atomic Spectrometry November 1995 Vol. 10 91 7J 105 c 'Y) 104 v) 3 c $ 103 cn v) C .- .- * 102 0 10' l o 4 10' l o 6 10' NO ion signallcounts s-' Fig. 19 Correlation of Co+ signal (Co was added as a trace element at 10 ppb) with the background NO' signal in different matrix element solutions at different matrix element concentrations from NO+ and the matrix effect results from suppression of the reactant ion.This conclusion requires also that charge transfer from the matrix ion is not effective at ionizing trace elements. The latter requirement is likely to hold except for resonant energy levels for which the rate constant for charge transfer between atomic species increases dramatically. The correlations of Fig. 18 and 19 are also consistent with ioniz- ation suppression resulting from enhanced ion-electron recom- bination provided that the matrix element significantly enhances the electron density. It is not clear why the data for the Sc matrix in Fig. 19 are not consistent with the data for the other matrices (it appears that the NO' signal is insufficiently suppressed since in Fig 18 the Co -t signal appears to be appropriately suppressed relative to Rh ). Under normal plasma conditions the matrix effect can be compensated by using multiple internal standards because the matrix suppression is predominantly determined by the masses of the matrix and analyte ions (although the mass-dependence is not linear).Under 'cold plasma' conditions the matrix effect is not strongly correlated with mass. However the analyte ion signal is correlated with the background NO' signal and a plot of one against the other is linear over a wide range of matrix concentrations. This is shown in Fig. 20 for a number of analyte elements in K matrix solutions of various K concentrations. The slopes of the curves differ for the different analyte elements and suggest differing rates of ion production or loss (e.g.ion-electron recombination) relative to those for NO'. The largest slopes are for Co+ Rh' and Pd' which are the elements having the largest heats of vaporization. As implied by the data of Fig. 19 the slopes are independent of the matrix element and are reasonably linear (excepting instances where the matrix standard is contaminated with the trace element). This then allows for determination of the matrix correction factor by measuring the analyte ion and NO+ responses for any acid-matched matrix solution (including an unrelated synthetic acid-matched matrix) and also its acid- matched blank. The concentration of the analyte element [MI is then given by where RFclean is the response factor (e.g.counts s-l per ppb) in a clean standard solution S(M'),ample is the ion signal for the analyte in the sample solution AS(NO+) is the difference in the NO' signals between the sample and the acid-matched - 'v) 2 ' .- cn .Q v) 3 CI 10 10' l o 5 NO ion signallcounts s-' la - E 2 10 10 10 lo-* 10" Normalized NO' signal Fig. 20 Analyte ion signals as a function of K concentration for a series of K matrix solutions. Raw signals are given in (a). The data are replotted in (b) after normalizing the signals to their intensities in the 10 ppb (clean) solution. The data of (b) allow better observation of the variation of slope with analyte element blank and is the ratio of the differences of the analyte ion and NO' signals obtained for any acid- matched matrix and blank solutions.Since the matrix suppres- sion is linear when normalized to the NO' signal the concen- tration of the matrix solution is not important but should be such as to provide substantial signal suppression (e.g. 100 ppm). From a mechanistic point of view it is notable that the trace element signals appear to be suppressed in concert with the NO' signal regardless of the ionization potential of the matrix element. In particular the trace analyte signals do not appear to be enhanced by the presence of a high concentration of a low ionization potential matrix element (cJ the linearity of analyte signals with NO' suppression in Figs. 19 and 20 combined with the apparent insensitivity of the NO + suppres- sion to conditions under which enhanced ionization is observed as shown in Figs.16 and 17). The trace element ion signals appear to follow the NO' signal rather than the other background ions or the EIE matrix signal. It is recommended that NO+ be the reference ion used for matrix correction and not some other background ion. The correlation of analyte ion signal with NO' as shown in Fig. 19 was reproducible for all the elements studied as either trace analyte or concomitant matrix. However correlation with OZ' or another background ion is not as straightfor- ward. For matrix elements having ionization potentials above 6.0eV the slope of the analyte ion versus 0,' response is relatively independent of the matrix element. However for matrix elements having ionization potentials less than 6.0 eV the correlation is not good as shown in Fig.21 in which considerable scatter in the correlation of Co' with 02' is observed for the matrix elements K Na and Li and the A1 91 8 Journal of Analytical Atomic Spectrometry November 1995 Vol. 1010' l o 5 l o 6 10' 0 ion signavcounts s-' Fig.21 Correlation of Co' signal with background 02+ signal for different matrix elements at various concentrations. The scatter observed for K Na and Li matrix solutions is a result of the initial suppression and then enhancement of the 02+ signal as the concen- tration of the matrix elements was increased. The A1 matrix data are distinctly curved reflecting the fact that the 02+ signal was not fully suppressed in these solutions data are significantly non-linear (note that these are the matrix elements having ionization potentials less than 6.0 eV).In fact this scatter arises because the 0,' is first suppressed and then recovers as the matrix concentration is increased. The distinct behaviour for A1 results because the 0,' signal did not recover at high A1 concentration but it also was not as fully suppressed as with higher ionization potential matrix elements. This difference in behaviour between low and high ionization poten- tial matrix elements is exactly analogous to the data presented in Figs. 16 and 17 above and for the same reason. As noted above it appears that there is an additional or enhanced mechanism of ionization for high concentrations of matrix elements having low ionization potentials. This enhanced ionization does not appear to affect the sensitivity to trace level analyte ions but it does affect the background plasma ion signals.Where there is a potential isobaric interference from a background plasma ion the analyst must have a means by which to identify a possible increase in the background signal. One way to do this is to determine the signals for 0,' and NO+ in the sample. Both of these ions are background ions having high signal intensity and low mass and hence are unlikely to suffer serious isobaric interference from the sample (especially since S which might interfere with 02+ has a high ionization potential and is therefore inefficiently ionized). The correlation of these ion signals as a function of matrix concen- tration for various matrix elements is shown in Fig. 22 (this is analogous to Fig.19). For matrix elements having ionization potentials greater than 6.0 eV the O,+ and NO' signals are suppressed more-or-less in unison (slope of 1). However for low ionization potential matrix elements the 0,' signal is not as suppressed as that of NO+. Other background ions includ- ing H20+ H30+ and 0' and the argide ions Ar' ArH' ArO' and AT,+ behave in a manner similar to that of 0,'. It is particularly the response of the argide ions that stimulates our concern since an increase in the ArO' background could be confused with a response for analyte Fe. The analytical protocol allowing recognition of this potential background interference is to measure the 0,' :NO+ ratio in the blank solution and in the sample (these must have the same nitric acid concentration as the NO+ signal is a function of the acid concentration).If the ratio in the sample is similar to that in the blank there is unlikely to be an enhanced background interference. Great accuracy in the measurement of this ratio is probably not required as the background increase is not significant until the ratio changes by a factor of 3 or more. As noted above the matrix element may significantly 107 .- C .- 1 0 5 lo6 10' NO ion signaVcounts s-' i ! l o B Fig. 22 Correlation of the background signals for NO+ and 0,' for different matrix elements at various concentrations. For the low ionization potential matrix elements the 02+ signal suppression does not correlate with the NO+ signal suppression. It is this behaviour for 02+ that is responsible for the scatter shown for the low ionization potential elements in Fig.21 and is taken as an indication of an additional source of ionization under these conditions increase the electron density in the plasma. If the plasma temperature (specifically the ionization and electron tempera- tures) is not concomitantly increased this will result in enhanced ion-electron recombination. Fig. 23 presents the results of calculations of ion densities in the plasma assuming that the matrix effect is one of suppressed ionization owing to enhanced ion-electron recombination. For these calculations the electron density in the absence of a matrix element was assumed to be 4 x 10" cm-3 and the ionization and electron temperatures were assumed to be equilibrated at 3500 K. For each matrix element and for each concentration the electron density was calculated assuming equilibrium thermal ionization of the matrix element (as before assuming 2% nebulization efficiency and uniform dispersal in the central channel at a gas kinetic temperature of 1450 K).The electron density is then a strong function of the concentration and ionization potential of the matrix element. Fig. 23(u) is the analogue of the experimental data given in Fig. 12. The curvature of the response curve at high matrix concentration is predicted although the ion density at which curvature is calculated to occur should be a function of the ionization potential of the matrix element rather than near the NO' signal level for the blank solution. Fig. 23(b) is the analogue of Fig.14 and predicts rather well the order of suppression of the Co' signal with matrix element. The model underestimates the suppression for T1 Bi and Rh perhaps indicating an underlying dependence of the suppression on the mass of the concomitant element which is not accounted for in the model. This mass dependence may point towards the neglect of space charge in the model or may indicate the importance of mass-dependent radial diffusion in the plasma. Fig. 23(c) shows the calculated corre- lation of analyte ion densities with NO+ density as a function of the concentration of K matrix element [cf Fig. 20(b)]. Although the model predicts a non-linear correlation for low ionization potential analyte elements the relative order of the slopes agrees with the experimental data (except for the high ionization potential analyte elements notably Cd and Zn).It is evident that the model also predicts that the relative suppres- sion (that is after normalization to either an internal standard or to NO') is independent of the identity of the matrix element as was observed experimentally in Figs. 18 and 19. The model does not account for the 'enhanced ionization' effect (increase of background ion signals and their lack of correlation with NO+ ) for the low ionization potential matrix elements. The model might be refined by including the effects of radial dispersion and non-uniformity in the plasma (particu- Journal of Analytical Atomic Spectrometry November 1995 Vol. 10 91 910 % .% C U 10 10s 10'~ loe2 Matrix molarity 10 -' 10 -' 10" l o o Calculated normalized NO' density Fig.23 Calculated ion densities as a function of matrix element concentration assuming equiliorium thermal ionization at 3500 K and an electron density in the absence of a matrix element of 4 x 1O1O cmW3. The electron density is increased by an amount given by thermal ionization at 3500 K of the amount of matrix element introduced into the plasma and assuming uniform distribution of the electrons in the central gas flow (1.1 1 min-l) at a gas kinetic temperature of 1450 IS. (a) (b) and (c) are the calculated analogues of the experimental data given in Figs. 12 14 and 20(b) respectively larly in the vicinity of vaporizing particles) and ohmic heating resulting from the increased electron density. Nonetheless the model suggests that thermal ionization may be dominant in the plasma.CONCLUSIONS 'Cold plasma' conditions permit the determination of elements normally interfered with by Ar+ and ArO+ because the conditions suppress the appearance of Ar' and 0'. Without the source ions the polyatomic argide ions are also suppressed. The resultant plasma is dominated by NO+ and 02+ and if no attempt at desolvation is made also the water-derived ions H20+ and H30+. Under the conditions used in this work the relative proportions of NO' and 0,' are determined principally by the nitric acid concentration of the sample. Since the NO' derives from the sample it behaves as a sample species and this has importance for its use as an internal standard. Some polyatomic argide ions persist notably ArH + and Ar2+.The approach is useful only if alternate sources of excitation such as a secondary discharge are eliminated. Sensitivity under 'cold plasma' conditions appears to be primarily a function of ionization potential. The sensitivity decreases markedly above 8 eV and so the technique is most useful for elements having lower ionization potentials. Heat transfer from the plasma to the sample appears to be insufficient to vaporize elements having high heats of' vaporization. Because the cooler plasma conditions favour the formation of molecular ions (notably the oxides of refractory elements) it is important for analytical purposes also to monitor ions at 16 u below the mass of interest to account for isobaric oxide interferences. For example under the conditions used in this work 40CaO+ presents an isobaric interference for 56Fe+ which is approxi- mately 30% of the signal obtained for 40Ca+.Matrix effects under cold plasma conditions are more severe than those obtained under normal analytical conditions. The matrix effect could be consistent with a model of the plasma dominated by ion-molecule chemistry as the reactant ions notably NO+ and 02+ are suppressed by charge transfer to the matrix ion. Alternatively the matrix suppression effect is also consistent with thermal ionization as the ion signals are suppressed by ion-electron recombination enhanced by the increased electron density contributed by the matrix element. However chemical equilibrium is not achieved in the plasma as the suppression of analyte ions is not strongly correlated with ionization potentials of the elements above and below that of the matrix element.Because of the severity of the matrix effect even at low matrix concentrations (in the low ppm range) the approach is most applicable to clean waters and acids with low salt content. However the technique may also be extended to less pristine samples since much of the matrix effect can be corrected for by correlating the analyte ion response with the NO+ ion signal (or to any other sample- derived internal standard). An additional or enhanced mechan- ism of ionization appears for high concentrations of matrix elements having ionization potentials below 6.0 eV. This enhanced ionization is responsible for the change to a smaller but non-zero response factor for the matrix element when the matrix ion signal exceeds the acid-matched blank NO + signal.The additional ionization does not appear to affect the response factor for trace elements. The appearance of enhanced ioniz- ation is significant because it is accompanied by an increase in the background signals of Ar' and polyatomic ions that may interfere with the elements of interest. However the appearance of this interference is indicated by a change in the ratio of NO" to 0,'. An equivalent indicator is the ratio of NO' (or any other sample-derived internal standard) to any background ion other than NO+ (such as H20+ H,O+ or O+ or even Ar' ArH+ or Ar2+). The author is grateful to Professor R. S. Houk (Iowa State University) and Dr. John Olesik (Ohio State University) for very helpful discussions and Peter Muellerchen (SCIEX) for assistance with the Colpitts oscillator experiment.REFERENCES 1 Houk R. S. Fassel V. A. Flesch G. D. Svec H. J. Gray A. L. and Taylor C. E. Anal. Chem. 1980 52 2283. 2 Date A. R. and Gray A. L. Analyst 1981 106 1255. 3 Douglas D. J. and French J. B. J. Anal. At. Spectrom. 1988 3 743. 4 Lam J. W. and McLaren J. W. J. Anal. At. Spectrom. 1990 5 419. 920 Journal of Analytical Atomic Spectrometry November 1995 Vol. 105 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 Bradshaw N. Hall E. F. H. and Sanderson N. E. J. Anal. At. Spectrom. 1989 4 801. Moens L. Verrept P. Dams R. Greb U. Jung G. and Laser B. J. Anal. At. Spectrom. 1994 9 1075. Jiang S.-J. Houk R. S. and Stevens M.A. Anal. Chem. 1988 60 1217. Farnsworth P. B. paper presented at the 1994 Winter Conference on Plasma Spectrochemistry San Diego CA January 10-15 1994 paper IL 12. Ogilvie C. M. and Farnsworth P. B. Spectrochim. Acta Part B 1992 47 1389. Farnsworth P. B. and Omenetto N. Spectrochim. Acta Part B Sakata K. and Kawabata K. Spectrochim. Acta Part B 1994 49 1027. Gray A. L. J. Anal. At. Spectrom. 1986 1 247. Nonose N. S. Matsuda N. Fudagawa N. and Kubota M. Spectrochim. Acta Part B 1994 49 955. Uchida H. and Ito T. J. Anal. At. Spectrom. 1994 9 1001. Douglas D. and French J. B. Spectrochim. Acta Part B 1986 41 197. Douglas D. in Inductively Coupled Plasma in Analytical Atomic Spectrometry eds. Montaser A. and Golightly D. W. VCH New York 2nd edn. 1992 ch. 13. Fulford J. E. and Douglas D. J. Appl. Spectrosc. 1986 40 971. Denoyer E. R. Jacques D. Debrah E. and Tanner S. D. At. Spectrosc. 1995 16 1. Tanner S. D. Spectrochim. Acta Part B 1992 47 809. Tanner S. D. Douglas D. J. and French J. B. Appl. Spectrosc. 1994,48 1373. Barnes R. M. Crit. Rev. Anal. Chem. 1978 7 203. Tanner S. D. J. Anal. At. Spectrom. 1993 8 891. Houk R. S. Anal. Chem. 1986 58 97A. de Galan L. Smith R. and Winefordner J. D. Spectrochim. Acta Part B 1968 23 521. Hanselman D. S. Sesi N. N. Huang M. and Hieftje G. M. Spectrochim. Acta Electron. Part B 1994 49 495. van der Mullen J. A. M. Nowak S. van Lammeren A. C . A. P. Schram D. C. and van der Sijde B. Spectrochim. Acta Part B 1988 43 317. Hobbs S. J. and Olesik J. W. Anal. Chem. 1992 64 274. i993,4a 809. 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 Gillson G. R. Douglas D. J. Fulford J. E. Halligan K. W. and Tanner S . D. Anal. Chem. 1988 60 1472. Tanner S. D. Cousins L. M. and Douglas D. J. Appl. Spectrosc. 1994 48 1367. Olesik J. W. Dziewatkoski M. P. McGowan G. J. and Thaxton C. paper presented at the 1995 European Winter Conference on Plasma Spectrochemistry Cambridge UK January 8-13 1995 paper 15. Handbook of Chemistry and Physics ed. Weast R. C. CRC Press Cleveland OH 51st edn. 1970 E-74 and D-56-D60. Huntress W. T. Jr. Astrophys. J. (Suppl. Ser.) 1977 33 495. Ferguson E. E. 1973 Atomic Data and Nuclear Data Tables Academic Press New York 1973 vol. 12 p. 159. Fehsenfeld F. C. Schmeltekopf A. L. and Ferguson E. E. J. Chem. Phys. 1967,46 2802. Lindinger W. Fehsenfeld F. C. Schmeltekopf A. L. and Ferguson E. E. J Geophys. Res. 1974,79 4753. Fehsenfeld F. C. Planet. Space Sci. 1977 25 195. Debrou G. B. Goodings J. M. and Bohme D. K. Combust. Flame 1980 39 1. Goodings J. M. Bohme D. K. and Ng C.-W. Combust. Flame 1979 36 27. Huber K. P. and Herzberg G. Molecular Spectra and Molecular Structure I V Constants of Diatomic Molecules Van Nostrand Reinhold New York 1979. Hayhurst A. N. and Telford N. R. Trans. Faraday SOC. 1970 66 2784. Lias S. G. Bartmess J. E. Liebman J. F. Holmes J. L. Levin R. D. and Mallard W. G. ‘Gas-Phase Ion and Neutral Thermochemistry’ J. Phys. Chem. Reference Data 1988 17 (Suppl. l) p. 622. Turner P. J. in Applications of Plasma Source Mass Spectrometry eds. Holland G. and Eaton A. N. Royal Society of Chemistry Cambridge 1991 p. 71. Miller M. H. Eastwood D. and Hendrick M. S. Spectrochim. Acta Part B 1984 39 13. Caughlin B. L. and Blades M. W. Spectrochim. Acta Part B 1985 40 987. Paper 5/02599K Received April 24 1995 Accepted July 12 1995 Journal of Analytical Atomic Spectrometry November 1995 Vol. 10 921