September 1983 Vol. 108 No. 1290 The Analyst Inductively Coupled Spectrometry Using Plasma Source Mass Continuum Flow Ion Extraction Alan L. Gray* and Alan R. Date Department of Chemistry University of Surrey Guildford Surrey G U2 5XH Institute of Geological Sciences 64-78 Gray's Inn Road London WClX 8NG An inorganic mass spectrometry system is described that uses an atmospheric pressure inductively coupled plasma (ICP) as an ion source. Solution samples may be introduced directly by nebulisation and the analysis time is about 1 min. Ions are extracted from the bulk plasma without any inter-vening boundary layer so that the full advantages of the ICP as a high-temperature dissociation and ionisation medium are realised. A quadrupole mass analyser is used with a pulse counting ion detector followed by a multi-channel scaler data system.General background levels are low and a wide range of elements may be determined with detection limits below 1 ng ml-1. Spectra are very simple with few molecular analyte ions and only singly and doubly charged species are found. Mass resolution is adequate to avoid peak overlap and isotope ratio determinations may readily be made with precision below 0.5% with integration times of about 5 min. The operating characteristics and performance of the system are described and illustrated and the future development potential is discussed. Keywords Inductively coupled plasma mass spectrometry ; continuum pow ion extraction The use of an atmospheric pressure plasma as an ion source for atomic mass spectrometry is an attractive concept not least because of the ease and rapidity of sample introduction and exchange normally a difficult and restricting aspect of inorganic mass analysis.When an inductively coupled plasma (ICP) is used as the plasma it offers the additional possibility of high sensitivity and freedom from matrix and inter-element effects characteristics that have made the ICP such a useful source for atomic-emission spectrometry. It is not surprising, therefore that since the principle of using a plasma in this way was first demon~trated~-~ it has attracted the attention of a number of other worker^.^-^ The earlier publications in this field4,597 all referred to work with small sampling apertures, which in the flowing plasma are covered by an aerodynamic and thermal boundary layer.The resulting sample of the plasma ion population transmitted to the mass analyser was thus drawn from within this layer. The temperature in this region where the ions may reside for several microseconds was considerably lower than that in the bulk plasma. Although very low detection limits were reported the full advantages of the ICP particularly in respect of matrix tolerance and dynamic range were not achieved. One group has recently reported the results of sampling a microwave plasma6 with apertures large enough to induce continuum flow from the plasma which breaks through the inter-mediate boundary layer.8 The performance attained with this plasma however appears to be limited by the low gas temperature which is probably below that of the boundary layer formed over small apertures in an ICP.It would clearly be of great advantage if continuum flow sampling could be achieved from an ICP. Some preliminary steps towards this have also been described by Douglas et aL9 but the results so far published do not yet show all the advantages over the microwave induced plasma (MIP) that should be realised from the use of an ICP. * To whom correspondence should be addressed. 103 1034 Analyst Vol. 108 We have recently published an account of our work on the boundary layer mode of sampling from an ICPlO and this paper reports the outcome of the concurrent work on continuum flow sampling in which all the characteristics expected of the ICP as an ion source have been observed. These results have previously been reported in part on several 0ccasions.~1-1~ A standard commercial argon ICP system and torch are used and solution samples are introduced via a standard pneumatic nebuliser without desolvation.The torch and load coil mounting is modified to provide a horizontal torch position so that access to the torch mouth is simplified. Ions are extracted through an aperture typically of 0.5 mm diameter into the vacuum system and analysed in a quadrupole mass analyser. Data are recorded and displayed in a commercial multi-channel scaler (MCS) data system designed originally for X- and gamma-ray spectrometry. The ion extraction process and the ionisation characteristics of the ICP as a source are discussed and the performance obtained on solutions containing both trace and matrix elements is described.GRAY AND DATE ICP SOURCE MS USING Experimental Instrumentation The major components from which the system was assembled are listed in Table I. The complete system and its operation in the boundary layer mode have been described else-TABLE I EXPERIMENTAL SYSTEM Component Descriptions and supplier Inductively coupled plasma system . . Model ICP 2500 with APCS-1 automatic power supply and AMN-2600E automatic matching network. Output power up to 2 600 W. Frequency crystal controlled 27.12 MHz. Torch Type T1.O. Load coil 3 turns of +-in copper tubing, water cooled. Nebuliser . . . . Fixed geometry cross-flow type 09.790. Jarrel-Ash Co., Aerosol chamber local manufacture. Plasmatherm Inc. Kresson N J USA Waltham MA USA.No desolvation Vacuum system . . . . Vessel local supply. Vacuum pumps : Stage 1. Stage 2. ED 660 mechanical pump. E09 vapour pump, Stage 3. ED 100 mechanical pump. E04 vapour pump, Both vapour pumps fitted with water-cooled baffle valves. All supplied by Edwards High Vacuum Ltd. Crawley Sussex Mass analyser . . Quadupole analyser Type 12-12S mass range 0-800 8.m.u. Resolution >2.5 M. Fitted with Galileo channel electron multiplier ion detector Type 4870. VG Analytical Ltd., Altrincham Cheshire Pulse counting channel . . . . Pulse amplifier Type 9302. Ratemeter Type 449. Scaler timer Type 776. EGG Ortec Oak Ridge TN USA Data system . . . . Canberra Series 80 with TTY and X - Y plotter read-out and cassette tape storage of data.Canberra Instruments Meriden CT USA EDM 20A mechanical pump 342 1 min-'. 1200 1 s-1. 200 1 s-1. Sampling apertures . . . 0.4 and 0.5 mm diameter 0.5 mm long. Drilled direct into Skimmer . . Nickel stainless steel or copper of local manufacture. 1 .O mm Also 1.0 mm nickel skimmers copper nickel or nickel alloy cones aperture 56O external angle. by Beam Dynamics lnc. Minneapolis MN US September 1983 CONTINUUM FLOW ION EXTRACTION 1035 where.1° To sample through the boundary layer by establishing continuum flow at the gas temperatures reported for the ICP16Js requires sampling apertures of at least 0.2 mm diameter. The gas flow admitted by apertures of this diameter is just beyond the capacity of the vapour pump used in the first stage for boundary layer sampling if the pump is to operate efficiently (at a pressure below Torr).The need to increase the capacity of this stage is avoided by preceding it with a small vessel at an intermediate pressure from which the bulk of the entering gas is removed by a mechanical pump at a pressure of about 1 Torr. The operation of a molecular beam forming stage in this pressure range was first described by Camparguel' and later studies were reported by Greene et aL1* of operations between 1 and 100 Torr. A similar stage was used by Douglas and co-workers to sample from both an MIP and ICP.6pQ The pressure ratio across the sampling aperture in such a stage is still well below the critical value at which the flow becomes supersonic. A molecular beam is formed along the axis of the system the core of which is then admitted to the next stage through a skimmer aperture.No electric fields are used in this process. Once into the first vapour pumped stage the sample is focused through the differential aperture into the mass analyser just as in the boundary layer sampling system previously described. A diagram of the plasma sampling and analysis system is shown in Fig. 1 and the details of the expansion stage and sampling interface in Fig. 2. The only remaining differences from the original systemlo lie in the mounting of the plasma torch and sampling aperture on the same horizontal axis as the ion lens and mass spectrometer and the omission of the sample desolvation. 11 'P 6 i 8 Fig. 1. Arrangement of plasma and mass analyser.1 2 3 Successive vacuum stages; 4 torch housing; 5 radiofrequency matching unit; 6 7 coolant and plasma gas inlet; 8 sample (injector) gas flow; 9 ion lens; 10 quadrupole analyser; 11 ion detector; and 12 signal output. Sampling cones are machined from the solid and the aperture drilled direct in the tip. The cones have an internal angle of 120" and fit into a recess in the water cooled front plate of the expansion stage, The ICP torch is the standard type used for atomic-emission spectrometry. Its mouth projects just beyond the wall of the torch box so that the tip of the sampling cone may be positioned up to about 3 mm from the outer end of the load coil. The torch box is mounted on a conventional optical bench that provides motion for alignment in three directions.A fixed-geometry cross-flow pneumatic nebuliser is used for sample introduction. It is mounted immediately below the torch housing and sample mist is introduced direct to the torch through a short length of glass tube. Operating Conditions The plasma is normally operated at 1200 W (reflected power <5 W) with a coolant gas flow-rate of 12lmin-1 a zero plasma gas flow-rate and a sample injector flow-rate of 0.6 1 min-l. Sample uptake at this injector flow-rate is 1.8 ml min-l 1036 GRAY AND DATE ICP SOURCE MS USING m I ’ Analyst Vol. 108 Fig. 2. Ion extraction interface. 1 First (expansion) stage; 2 second vacuum stage; 3 load coil; 4 plasma torch; 6 quartz bonnet; 6 torch housing; 7 sampling cone; 8 water-cooled flange; 9 skimmer; and 10, extraction electrode.At the usual operating position of a 0.5 mm diameter aperture 10 mm from the load coil, the operating gas pressures in the system are as follows stage 1 (expansion stage) 1 Torr; stage 2 2 x lo-* Torr; and stage 3 (mass analyser) 2 x Torr. Sample Solutions For the performance studies described here sample solutions were prepared in the laboratory from analytical-reagent grade standard solutions at 1000 pg ml-l. Solutions were diluted with distilled water to the desired concentration and acidified to 1% with Aristar doubly distilled nitric acid. Operating Procedure In operating the system full advantage is taken of the facilities of the programmable multi-channel scaler. For normal spectrum display memory groups of 1023 channels are used to accumulate and store data from scans of up to 100 a.m.u.width. The mass analyser scan is triggered by the start of the MCS sweep and normally a dwell time per channel of 1 ms is used. The full scan therefore takes just over 1 s and when 60 scans are pre-set a total integrating time of approximately 1 min is obtained. The system may be programmed to make repeated integrations recording data on to tape while doing so for subsequent printout on teletype or graphics plotter for evaluation. Regions of interest may be set around any wanted peaks or background areas and total peak integrals printed out. Blank spectra may be stored before the sample run and then subtracted automatically before printing out peak integrals if desired. Operation in the scanning mode is usually most convenient for analytical work but some performance studies may be more easily carried out in the single ion mode.The mass analyser is set to the peak of the wanted analyte ion by observing the ratemeter response and the multiscaler then set to integrate the ion signal as a function of time by setting a convenient channel dwell time such as 1 s. In this way a plot of the response is obtained direct on the VDU the 1023 channels accommodating a run of about 17 min. This mode of operation is very suitable for determination of detection limits studying response time or memory effects noise levels stability and for other measurements where the response to a single ion is sufficient. Accumulated data may be printed or plotted out in the usual way. When a solution is presented to the nebuliser the response appears in a few seconds.A period of 20-30s is usually allowed for the system to equilibrate when a new solution is introduced before the start of an integration September 1983 CONTINUUM FLOW ION EXTRACTION Results and Discussion Sampling Interface Operation 1037 At the normal operating position of 10 mm from the load coil the gas temperature is reported as 7500 K15916 and this figure gives good agreement with the observed working pressures and pumping capacities. A t this temperature the aperture Knudsen number Kn, is 3 x 10-3, which is well below the value requirea to puncture the boundary layer. When the aperture is placed in the working position in the flame the boundary layer formed over the sides of the cone near the tip may be clearly seen.When water is nebulised this layer appears as a dark space but if a solution of 1000 pg ml-1 of an yttrium salt is introduced the boundary layer is made more visible by the red band emission from yttrium(I1) oxide reformed in the cooler layer. In the centre of the flame the blue emission from yttrium atoms and ions may be seen as they stream through the aperture.lg Within the expansion stage the incoming gas is formed into a molecular beam by the enhancement of the energy along the axis. A terminal Mach number of about 12 in front of the skimmer aperture is calculated from the parameters of this stage.s At a pressure of 1 Torr much higher than usual in molecular beam stages considerable scattering at the fringes of the beam is to be expected and it is not clear whether the normal theory of beam formation may be used at this pressure.However good spectra are produced at aperture -skimmer spacings greater than 5 mm with an optimum between 7 and 10 mm. At lower pressures also the gas temperature in the core of the beam should fall to about 75 K before reaching the skimmers but the skimmer tip operates well above room temperature so again this departure from theory is probably a consequence of the higher pressure. However a beam does appear to be formed as little further signal is gained as the skimmer aperture diameter is increased above 1 mm although up to that point the response is roughly pro-portional to aperture area. However in spite of the formation of a beam the efficiency of such a stage is low.By comparison with boundary layer sampling it is calculated that only 0.4% of the incoming ions pass into the next stage. This compares favourably with the figure of 0.1% reported by Campargue17 for a similar stage. Clean spectra are obtained with a simple structure similar to those obtained in boundary layer sampling and the random background count from all sources over the whole spectrum is between 10 and 20 counts s-1. An axial stop on one lens cylinder is used to obstruct direct photons from the plasma so that in spite of the in line torch and aperture these contribute a negligible amount to the background. The random count is thought to arise mainly from photon emission from decaying excited ions principally of argon within the ion lens. A similar but smaller random background has been observed in boundary layer sampling.The use of apertures large enough to induce continuum flow into a low pressure first stage, of the type used for boundary layer sampling has been found to produce an intense pinch discharge in the mouth of the aperture which contributes a very high photon count to the b a c k g r ~ u n d . ~ ~ ~ It was found in the initial exploration of continuum flow from the ICPll that this pinch discharge was more intense at low pressures behind the aperture and was completely quenched at pressures above about 4 Torr. It was also found to be more intense on the edges of the flame where the gas temperature was lower so that its intensity appears to be related to both gas density and rate of expansion through the aperture.At a working pressure in the expansion stage of about 1 Torr there is no observable pinch discharge in the aperture and no photon contribution to the background is detectable. A residual discharge may still be present as in the intense light from the plasma a weak discharge would not be visible. The temperature of the tip of the sampling cone is a function of the thermal conductivity and thickness of the cone walls. Nickel sampling cones have been dimensioned so that they operate at red heat as it was originally considered that this would minimise sample condensa-tion problems and memory effects. However in the presence of a high concentration of atomic oxygen from the nebulised water in the plasma flame nickel is found to oxidise on both internal and external surfaces where the cone is red hot.This does not interfere with the sampling process as far as can be observed but it does create a rough surface over the tip of the cone which tends to retain condensed material and cause slight memory effects. Nickel oxide is not readily removed without mechanical cleaning so some of the oxidation-resistant nickel alloys are being investigated. These unfortunately have considerably poore 1038 Analyst Vol. 108 thermal conductivity and require thicker sections. Copper cones operating below red heat have been found to work well with very low memory. Sample condensation does occur on the cooler parts of the cone at concentrations above about 10 pg rnl-l in an annulus around the aperture but does not obstruct the aperture itself.The response of the system to a 10pgml-l solution of lead is shown in Fig. 3. After a 4-min run on 208Pb+ the response takes 55 s to decay to of the peak signal after the last of the lead solution mist has reached the plasma. Metals such as lead cadmium and zinc show greater memory effects than the less volatile elements. For example the response to barium after solution ceased to reach the plasma fell to 3.2 x of the peak value in 60 s. Thus successive measure-ments with a 1-min blank run between them would show less than 0.1% of cross-contamina-tion. The lifetime of the present sampling cones is limited by sputtering in the region immediately behind the aperture. This removes metal and eventually weakens the lip of the aperture. Both copper and nickel are metals with high sputter yields so alternative materials of lower yield are being investigated.However the cones are not expensive and may be replaced in a few moments after the life of 30 h or more that is usually obtained. The principal draw-back of the sputtering is the presence of the metal peaks in the spectrum. There is also a small contribution to the spectrum from the skimmer tip although not enough to imply significant skimmer wear. If the same metal is used for both skimmer and cone no additional interference is caused. For this reason pure metals are preferred to alloys for these com-ponents. GRAY AND DATE ICP SOURCE MS USING I- 1 I I 0 1 2 3 4 5 6 Ti m e h i n 1 .o 0.8 U 0.6 d 4 0.4 2 0.2 0 i I I I I 2 4 6 8 1 0 Ion energylev Fig.3. Signal decay on 10 pg ml-l lead solution. Fig. 4. Distribution of ion energy Peak level 3.8 x 1 0 5 counts s-l on aosPb+. Arrow at entrance to quadrupole analyser at shows the 10-3 response point. 1200 w. Ion Optical Considerations The ion optical system and its operating potentials remain very similar to those used for boundary layer sampling one additional cylinder being mounted at the base of the skimmer to operate as an extraction electrode at a potential of about -200 V. In boundary layer sampling the low ion energy spread of about 2 eV appeared to be related to conditions in the boundary layer rather than in the plasma and enabled excellent resolution to be obtained from the mass analyser. Once ions had been drawn from the bulk plasma without the intervening collisional processes in the boundary layer it was expected that a much wider energy distribution would be found.However a retardation plot on cobalt ions at the entrance to the analyser produced the differential energy distribution shown in Fig. 4 which has a half width of about 7.5 eV. This is small enough to enable good mass resolution to be obtained without the use of an energy analyser. Choice of Plasma and Aperture Operating Conditions During the early stages of the development of continuum sampling based on the use of a standard ICP developed for emission analysis little variation on the normal plasma operating conditions has been explored partly because it was desired to build on atomic-emission experience and make as much use of already available data as possible.However withi September 1983 CONTINUUM FLOW ION EXTRACTION 1039 the variables available it was clearly important to determine the range of possible plasma operating powers and to establish the optimum sampling point in the flame. These studies were performed using indium as a convenient analyte. This has a low first ionisation energy (5.79 eV) and a second ionisation energy well above the first of argon so that it can be assumed to be fully ionised to singly charged ions. The l151n+ isotope abundance 95.72%, was monitored. A sampling aperture operating in the continuum flow regime disturbs the plasma upstream for a distance of 2-3 diameters. A 0.4 mm diameter aperture thus samples gas from a region of about 1 mm deep and 2.4 mm in diameter.The spatial resolution of an ion distribution plot obtained by moving the flame about over such an aperture is therefore about 2 mm. Transverse and longitudinal profiles obtained in this way using the single ion mode at a power level of 1200 W and at a solution concentration of 10 pg ml-l are shown in Figs. 5 and 6. The transverse profiles show a strong peak along the axis with a distribution that broadens with increasing aperture - load coil separation in accordance with the visible distribution of the blue emission in the flame seen when yttrium is introduced. In this particular plot the centre of the distribution may be seen to be slightly skew suggesting that the flame axis was not co-linear with the optical bench. The longitudinal profile shows a smoothly decreasing response from the torch mouth outwards.100 80 s ai 60 v) 0 a $ 40 a 20 0 5 4 3 2 1 0 1 2 3 4 5 Fig. 5. Profile of response across plasma flame a t various distances from load coil on ll5Inf. Power 1200 W. Load coil distance A 5 mm ; B 10mm; C 15mm; and D 20mm. Radial distance/mm 100 $ 60 40 al a 20 0 Axial distance from lead coil/mm Fig. 6. Variation of response along flame axis on l151n+. Power 1200 W. Clearly these profiles suggest that the greatest signals are obtained close to the torch before significant ion diffusion into the outer regions of the flame dilutes the analyte ions. This may not be so however where longer sample dwell times in the plasma are required to vaporise and dissociate samples containing a complex matrix or refractory molecules.It may be necessary to use greater aperture - load coil separations in instances where volatilisation interference appears to be present at high matrix concentrations or where incomplete dissociation of refractory oxides is observed. For normal operation a sampling position at 10 mm from the coil has been used at which matrix effects do not appear to be significant but this still requires further study for a range of matrices. Certainly the longitudinal profile shown in Fig. 7 of the metal and monoxide response from a uranium solution taken in the scanning mode suggests that there is no decrease in the degree of dissociation of this highly refractory oxide at close separations. The proportion of oxide response remains roughly constant out to 15 mm after which it increases sharply.This suggests that the majority of the oxide ions do not arise from undissociated uranium oxide molecules but are reformed in the boundary layer which i 1040 GRAY AND DATE ICP SOURCE MS USING Analyst Vol. 108 still present over the sides of the sampling cone up to the edges of the aperture from which they are then entrained into the extracted gas flow. In this instance a scan was performed at each power level over the mass range from 112 to 117 a.m.u. This enabled background channels either side of the l151n+ peak to be observed as well so that the signal to background (random) ratio could also be plotted. Plasma gas flow-rates were not changed during these runs. The l151n signal increases with power up to a plateau level at 1600 W but the signal to noise ratio peaks at 1400 W and decreases above that.As indium is expected to be fully ionised even at low power it is assumed that this increased response is a consequence of the contraction of the flame and central channel diameter which is observed close to the torch mouth as the power is increased. No problems with the plasma torch or sampling system were observed during short-term operation up to 1800 W but the increased heat transfer to the expansion stage for continuous operation at these powers would require better water cooling. There seemed to be little advantage in operating above 1400 W and a level of 1200 W was chosen for normal work. The variation of the response from indium with plasma power is shown in Fig.8. -\ 30 20,s 5 t-. 0 3 10 0 I 0 0 4 800 2 80 40___4400 0.8 1.0 1.2 1.4 1.6 1.8 Power/kW I I I I I Fig. 8. a Signal intensity I and 0 signal to background ratio I / B as. power for 1151n+ a t 10 mm from load coil. 0 5 10 15 20 Distance of aperature from load coil/mm Fig. 7. Variation of uranium metal and oxide response along flame axis at 1200 W showing uranium oxide to uranium ratio (yo). The effect of increased plasma power on the response to a uranium solution at 10 pg ml-l is shown in Fig. 9 again at 10 mm from the load coil. The scanning mode was again used to show both metal and oxide responses. The 238U+ signal increases just as l151n+ did up to 1400 W but beyond this the response falls again. Uranium has a low second ionisation energy of about 12 eV so the fall in response is probably associated with a shift in the ionisa-tion equilibria towards doubly charged ions at high plasma temperatures.The 254UO+ response however shows little change throughout which further supports an origin for the oxide ions in the boundary layer around the edge of the aperture. a I I I I I 0.8 1.0 1.2 1.4 1.6 1.8 Plasma power/kW Fig. 9. Response veYsws power for uranium and uranium oxide a t 10 mm from load coil Se$tember 1983 CONTINUUM FLOW ION EXTRACTION 1041 Background Response and Blank Spectra The experimental detection limits obtained by any analytical method depend critically on the background levels above which the response from the wanted species must be identified. The random background of the system across the whole spectrum has been found to be reasonably low between 10 and 20 counts s-l.However for samples introduced as nebulised solutions the response obtained to the blank normally 1% Aristar nitric acid in distilled water may for some parts of the spectrum be much more significant. Such a blank spectrum is shown in Fig. 10. The response is broadly similar to that obtained in boundary layer ~ a m p l i n g ~ t ~ ~ with two major groups of peaks one the oxygen and hydrogen (or water) group between 16 and 19 a.m.u. and the other at 40 and 41 a.m.u. due to 40Ar+ and 41ArH+. There are however some significant differences notably the great reduction in the third major group of peaks seen in boundary layer spectra at 30 32 and 33 a.m.u. due to the absence of the boundary layer.The important ion 30NO+ is only present as a small peak in continuum flow sampling. As this ion formed in the boundary layer has a low ionisation energy of 9.4 eV it greatly disturbs the ionisation equilibria if present at high concentration, and it can suppress the ionisation of species of higher ionisation energy in boundary layer sampling. Another ion in this group that is greatly reduced is 320,+ and 3302H+ is even smaller. However the presence in the plasma of such high concentrations of hydrogen and oxygen from the solvent does still result in substantial peaks at some inconvenient positions and while they are present some wanted ions will not be detected. Fortunately however, these are relatively few and while for some analyte ions there are small molecular background peaks resulting from ion molecule reactions between hydrogen nitrogen oxygen and argon, which increases the limits of detection for many ions only the random background is signifi-cant.Such small background peaks may be readily removed from the displayed spectrum by subtracting a blank in the data system which greatly contributes to the visibility of unexpected trace peaks. A final major difference from the boundary layer spectrum is the much smaller 80ArAr+. This residual peak is probably due to ions formed in the expansion by a polymerisation reaction but the much larger peak found in boundary layer spectra is more likely to be mostly the product of ion molecule reactions in the boundary layer. The proton attachment peak EIArArH+ found in the boundary layer spectra is absent in continuum spectra.Fig. 10. Blank spectrum on 1% nitric acid solution over the mass range 0-90 a.m.u. Integration time, 1 min; peak count rate “ArHf 9.8 x los s-l. Characteristic Spectra The spectra observed show the same characteristics of simplicity as those reported for boundary layer samplinglo However because the continuum sample is drawn from a higher temperature than exists in the boundary layer more doubly charged ions are seen and thus there are more peaks occurring at values of 9422. Apart from these doubly charged ions ion peaks occur effectively at unit mass intervals. Small oxide peaks less than 1% o 1042 GRAY AND DATE ICP SOURCE MS USING Analyst Vol. 108 the height of the main peaks are seen for elements with the most refractory oxides but apart from these simple elemental peaks are found for both metal and non-metal analytes.Typical spectra taken from the data system memory on to the graphics plotter are shown in Figs. 11-13. These were run on multi-element solutions each element being at a concentration of 10 pg ml-1 for an integration time of 1 min. Fig. 11 shows a spectrum at the bottom end of the mass range. The peaks of 1H+ and 2H+ are off-scale but a small 3H+ peak is seen and well isolated peaks of the lithium and boron isotopes are found. The boron isotopes are at about the correct isotope ratio but the apparently depleted state of the standard lithium solution with a low abundance for 6Li was a source of concern until the lithium was found to have been procured by the suppliers from a nuclear research establishment.The authors have speculated on the problems of un-suspecting users of a standard solution prepared on the basis of an incorrect relative atomic mass! Also to be seen in this spectrum are peaks of 12C+ and 13C+ probably arising from residual carbon dioxide in the argon. Even if it is present in the argon at levels sufficient to cause a significant background it may be simply removed from the response to a carbon containing analyte by blank subtraction. Direct measurements of carbon-12 to carbon-13 ratios may thus be made on solutions. This has been confirmed on 13C-enriched samples on which ratios have been measured with agreement better than 1%. Fig. 12 of about I 7Li+ I Fig.11. Spectrum of lithium and boron in solution a t 10 pg ml-l showing carbon peaks. Peak count rate 'Li+ 5.1 x lo4 s-l. shows part of a wider scan expanded in the data system to display a mass range 100-140 a.m.u. Again this was obtained on a mixed solution in this instance containing 15 elements each at 5 pg ml-1. In this part of the scan singly ionised peaks are seen for In Te I Cs and La with doubly ionised peaks of Pb and U. The resolution setting used for this run was not high enough to separate the doubly ionised Pb peaks. This plot is taken from a wide mass analyser scan from 0 to 260 a.m.u. which as an exception to normal practice was stored in 2048 memory channels in order to provide high print-out resolution for parts of the spectrum. A spectrum of a 10 pg ml-1 solution of heavy elements is shown in Fig.13 where singly ionised spectra of W Au Mg Pb Bi Th and U are seen. Only Pb Th and U doubly ionise I I Fig. 12. Expanded portion of wider scan showing Pb2+ In+, Ua+ Tef I+ Cs+ and La+ each a t a concentration of 5 pg ml-l. Peak count rate laaCs+ 6.2 x lo4 s-1 September 1983 CONTINUUM FLOW ION EXTRACTION 1043 Fig. 13. Spectrum of a mixture containing W Au Hg Pb Bi Th and U each a t a concentration of 10 pg ml-l. Peak count rate 20BBi+, 9.6 x 104 S-1. among these and Pb only slightly as its second ionisation energy is 15.03 eV. The lead response is thus almost as large in total as that for bismuth whereas the peaks for thorium and uranium which have second ionisation energies of about 12 eV are relatively small.Both Tho+ and UOf peaks are present. An expanded plot of the tungsten and mercury isotopes shown in Fig. 14 demonstrates the low background level. The clearly visible peaks of lS0W+ and lSsHg+ are each produced from a solution concentration of 14 ng ml-l. That for mercury is immediately adjacent to the lg7Au+ peak which is about lo3 times bigger but it is smaller than the tungsten peak of the same concentration as mercury is only about 50% ionised because of its higher ionisation energy. 182-4\(\1+ \ 19; Fig. 14. Expanded portion of Fig. 13 showing Isow+ and lS6Hg+ a t concentrations of 14 and 15 ng ml-l respectively. Peak count rate lsoW+ 250 s-l. Rare earth elements are usually difficult to analyse because of the complex spectra produced by multiple ionisation and by peaks from the refractory oxides.The continuum flow spectra from the ICP are remarkably simple however. All have low second ionisation energies and therefore show strong doubly charged ion peaks. First and second ionisation spectra are shown together in Fig. 15 for a mixture of 15 rare earth elements (supplied by Johnso 1044 GRAY AND DATE ICP SOURCE MS USING Analyst Vol. 108 (a) Ar.Ar 70 80 mlz 90 Tm Ho Lu HoO TmO LUO I 140 150 160 170 mlz 180 190 Fig. 15. Spectra of (a) singly and (b) doubly charged ions of a 15rare earth element mixture each at 10 pg ml-I. Upper spectrum M2+ ions mlz 69-90. Peak count rate 175L~2+ 2.4 x lo4 s-l. Lower spectrum M+ ions m/z 138-192. Peak count rate 175Lu+ 9.7 x lo4 s-l. Note *OY+ appears in (a).Matthey Chemicals) at a concentration of 10pgml-1 each. Although there are a number of coincidences between isotopes little difficulty exists identifying and quantifying each element. Although this selection of spectra include elements with ionisation energies ranging from Cs at 3.89 eV to I and Hg at over 10 eV the range of peak heights allowing for abundance, is small and illustrates well the wide element coverage. The only oxide peaks clear of the main spectrum can be seen to be very small September 1983 CONTINUUM FLOW ION EXTRACTION 1045 Quantitative Performance In contrast to other atomic ion sources the ICP source will accept fresh samples as fast as the chosen integration cycle will permit. It is there-fore possible to determine detection limits in terms of the concentration equivalent to the standard deviation of the background just as in other methods of atomic spectrometry.Detection limits determined in the single ion mode are shown in Table I1 expressed as 20 TABLE I1 COMPARISON OF DETECTION LIMITS (20 BLANK) A convenient rate is one every 2 min. Element Lithium . . Boron . . Magnesium Aluminium Titanium . . Vanadium Chromium Manganese Cobalt . . Zinc Germanium Arsenic . . Selenium . . Rubidium . . Silver . . Cadmium . . Indium . . Tellurium . . Caesium . Barium . . Lanthanum Cerium . . Tungsten . . Gold . . Mercury . . Lead Bismuth . . Thorium . . Uranium . . * .Detection limit/ng ml-l h r 3 ICP - SMS ICP - AES* FAAS? 3 1.9 2 1 3.2 1000 0.5 0.1 0.2 0.6 15 20 0.3 2.5 50 0.4 3.3 20 0.2 4.1 3 0.8 0.9 3 0.5 4 5 3 1.2 0.6 1 1 50 7 35 100 15 50 100 2 0.3 0.2 4.7 2 0.5 1.7 1 0.1 42 30 0.5 27 70 15 0.1 0.2 6.7 1600 0.2 32 0.5 20 500 0.2 11 10 0.4 17 200 0.3 28 20 0.2 23 40 0.2$ 43 0.4 170 7 000 --0.3 0.9 20 --* From reference 20. t From reference 21. 1 Doubly charged ion. values. The value of o was determined for each element by running a blank solution for ten integrations usually of 5 s each followed by an integration on a 10 pg ml-l solution of the element usually in a multi-element solution. Comparative values for ICP atomic-emission spectrometry (ICP-AES) and flame atomic-absorption spectrometry (FAAS) taken from recent literature are also shown.The ICP source mass spectrometry (ICPS-MS) values in the continuum mode can be seen to be generally higher than those previously reportedlo for the boundary layer mode but they are considerably more uniform over a wide range of elements. The isolated high values that remain are generally due to the use of an un-favourable isotope for the measurement due to a background coincidence or to the presence of a small background peak under the isotope peak used. These detection limits obtained at an early stage in the development of the continuum flow sampling technique are nevertheless very promising and compare very f avourably with other established techniques.It is known that ion collection and transmission in this system are very inefficient and offer considerable scope for improvement. It is therefore hoped that better detection limits will be achieved as the method is improved. The linear dynamic range of the plasma ion sampling interface and mass analyser is found to be greater than that of the ion detection system in its present form. The channel electro 1046 GRAY AND DATE ICP SOURCE MS USING Analyst Vol. 108 multiplier loses gain above pulse rates of 1 MHz so that ion pulses are increasingly lost below the discriminator threshold. To cover a range of more than 5 decades it is necessary to reduce the ion transmission for high sample concentrations by for example reducing the extraction electrode potential.By doing this the response has been found to be approxi-mately linear over 6 decades up to 1000 pg ml-l. The matrix effects found in boundary layer sampling were particularly severe and pre-cluded its use at total concentrations above about 10 pg ml-l. However in the continuum mode the matrix effects appear to be similar to those experienced in ICP-AES. Ion suppression by species of low ionisation energy is small as can be seen in Fig. 16 where the 10 20 50 100 200 500 1 000 Sodium concentration/pg ml-' Fig. 16. Effect of sodium addition on response to e 5sC~+; and 0 aosBi+ each at 10 pg ml-I. response of cobalt and bismuth falls only slowly with sodium concentrations up to 1000 pg ml-l. The effects of major matrix elements on the response to traces also appears to be small.Results of comparative analyses with and without a typical rock matrix show little effect on the trace responses at total solids contents over 500 pg ml-l with the matrix elements correctly recorded. However much more study is needed of this complex subject in order to confirm these initially promising results. The relatively uniform sensitivity shown by the detection limits in Table I1 over a wide range of ionisation energies suggests that the effective ionisation temperature Ti of the plasma at the sampling point is high. Measurements of the response to known concentrations of elements with a range of ionisation energies may be used to determine Ti by using the Saha equation to relate degree of ionisation to temperature and ionisation energy.The most direct method of doing this is to use the ratio of singly and doubly charged ions of the same atom but this requires an accurate knowledge of the ion transmission at values of m/z a factor of 2 apart. It was preferred therefore in this work to use two singly charged ions of similar m/z so that uniform ion transmission may be assumed. An ideal pair for this would appear to be caesium and iodine as they can be obtained in a fixed stoicheiometric ratio as caesium iodide and are close in mass at 133 and 127 a.m.u. The ionisation energy of caesium is so low (3.89eV) that it may be assumed to be fully ionised but the second ionisation energy is 25.08 eV. The 133Cs+ integral may therefore be used to represent the atomic concentration of both atoms before ionisation for comparison with the 12'1+ integral.A similar compound is rubidium bromide but this was not readily available and so a suitable solution of the two elements was used. Solutions of these two thermometric pairs were run at 10 pg ml-l at a power level of 1200 W together with blank solutions so that background could be subtracted. Two sampling positions were used and the values of Ti calculated from the ion integrals. Unfortunately values of partition coefficient for bromine were not available so a value of Zi/Z = 1 was assumed. In order to simplify the calculation for a multi-component plasma it is necessary to assume a value for the electron concentration H e . The values assumed at 1200 W were 7 x 1015 at 5 mm and 4.6 x 1015 at 10 mm.22923 In spite of the potential errors in the rubidium - bromine result the agreement shown for these two thermometric pairs in Table I11 is very reasonable although the values obtained for Ti are higher than usually reported from AES measurements.A plot of normalised response against ionisation energy for elements with a range of energies is shown in Fig. 17. For clarity only a few have been identified on the plot. The solid line shows the response calculated from the Saha equation for Ti = 9000 K Z i / Z = 1 and ne = 4.5 x 1015 cm-3. The experimental values fall close to this calculated response, which shows a high degree of ionisation at energies well above 1OeV. Even chlorine a September 1983 CONTINUUM FLOW ION EXTRACTION TABLE I11 IONISATION TEMPERATURE Ti AT 1200 W CALCULATED FROM Cs I Rb AND Br RESPONSE 1047 r Sampling position/mm Cs I Rb B; 5 9 090 8 920 10 9 060 8 930 13.01 eV is ionised to more than 5% and is readily detectable.Both halogens and non-metals appear to form positive ions normally and can be detected if the first ionisation energy is below that of argon (15.76 eV) and their response does not coincide with a major back-ground peak. As the ionisation equilibrium in the plasma is dominated by the high con-centration of argon species with ionisation energies above that of argon are not significantly ionised. However only three elements are thus excluded Ne F and He and of course argon cannot be detected if it is used as the plasma support gas. If a reasonable lower limit is set to the useful minimum degree of ionisation as lyo which is reached at 14 eV then this excludes only N (14.53 eV).The measured value of Ti of 9000 K thus appears to represent an acceptable compromise between an adequate degree of ionisation for as many elements as possible and the desire to avoid too many multiply charged ions. It excludes five elements at the penalty of double ionisation to varying degrees for 26 others. The lowest third ionisa-tion energy level is that for La at 19.17 eV so no ions with a charge state above 2 will be seen. The analyser can be set to resolve peaks only 0.5 a.m.u. apart (see Fig. 15) so that together with the low yield of molecular analyte ions the spectra remain simple and peaks are easy to identify. 1.0 $j 0.5 c x 0.2 ; 0.1 a 0.02 9) > 0) .- t;; 0.05 -0.01 ' ' ' ' ' 4 6 8 10 12 14 lo n isa t i o n en erg yleV Fig.17. Relative response zlersus ionisation energy at 10 mm and 1200 W. Graph shows response calculated from Saha equation for Ti = 9000 K and n = 4.5 x 1015 ~ m - ~ . Experimental values shown as points. Isotope Ratio Measurements The ability to resolve adjacent mass peaks sufficiently to reach background between them provides simple peak-area measurement in the data system. Random background from a blank run may be automatically subtracted and peak integral ratios printed out direct at the end of the run. The determination of isotope ratios on a routine basis on solution samples is thus straightforward. The results of such measurements with this equipment in the boundary layer mode have been reported previouslyll and recently the same technique has been described for the continuum mode.13 In both modes a precision of 0.5% or better has been demonstrated on natural galena samples and the results compared with thermal ionisatio 1048 GRAY AND DATE ICP SOURCE MS USING Analyst Vol.108 values on the same samples. The use of continuum sampling for this work offers the advantage of greater matrix tolerance but the higher random background and lower absolute count rate sensitivity were thought initially to be disadvantages. However in both modes the available precision certainly to the level of O.l% appears to be dependent only on counting statistics and continuum mode operation enables the sample concentration to be increased from the 1 pg ml-l normally used in boundary layer work to much higher levels.In practice the limit is set by counting loss in the data system. Losses of a few per cent. at high count rates caused by the finite pulse pair resolution of the system are usually ignored as insignificant but are intolerable when precise ratios between peaks of different sizes are required. Some correction is possible but loss of detector gain at high rates is not easy to determine. As a general rule therefore count rates for isotope ratio determinations are restricted to about 105 s-l and thus concentrations are limited to about 10 pg ml-l or less of the largest isotope. Even so peak integrals of well over lo6 counts may be accumulated in integration times of about 5 min provided that the scan width is limited to between 5 and 10 a.m.u.This is usually ample for the wanted isotopes. Although faster ion detectors and electronics are available they are costly and a better alternative to allow higher concentrations or better ion transmission to yield greater precision would be to use the electron multiplier in a d.c. mode and digitise the output to a lower rate that can be handled without loss. This and other d.c. techniques are being explored. However routine precisions of 0.1% are thought to be within the capability of the present system in the near future at concentrations from 1 to 10 pg ml-l. There seems to be little advantage in trying to better this unless it can be improved to the point where it approaches that achievable with thermal sources.The range l-O.l% adequately covers the needs of a wide range of analytical applications in stable tracer studies isotope dilution analysis and mineralogical prospecting. Although counting statistics may dominate precision in this range other more fundamental problems may make worthwhile further improvement much more difficult to achieve. Further Development Nebulisation of solutions is a very convenient method of sample introduction and a major step forward in mass spectrometry. It lends itself well to automatic operation by the use of a sample changer when samples blanks and calibration standards may be introduced in any desired pattern the resulting analytical data being handled by a dedicated computer. It is a very inefficient process only about 1% of the solution actually reaching the plasma, but even so in an automatic system of this type a volume of only about 3-10 ml is required for an analysis.When this is inconveniently large as for many biological samples a re-circulating nebuliser of the type described by Hulm~ton,~~ may be used with which analysis times of 30 min or more may be obtained from a 1-ml sample. This is particularly attractive for isotope ratio determinations on small samples. 0 2 4 6 8 ' Fig. 18. Single ion response to 20sPb+ for electrothermal vaporisation of a 5-111 drop of 10 pg ml-l solution (5 ng). Integral 2.4 x 106 counts. Vertical axis shows memory channel content. Ti me/s September 1983 CONTINUUM FLOW ION EXTRACTION 1049 Sample introduction to the ICP for use as an ion source remains similar to that for AES and any of the methods described for emission spectrometry may be used such as electro-thermal vaporisation laser ablation and hydride generation.The origin of most of the “permanent” background peaks in the high hydrogen and oxygen populations in the plasma, typically about 10s times the analyte concentration makes any method that avoids intro-ducing water to the plasma potentially attractive for improved detection limits. These alternative techniques are being studied. Early indications are that electrothermal vaporisa-tion of microsamples from a graphite rod offers some advantages. A system similar to that described by Gunn et aZ.25 has been used to introduce samples of 5-pl volume. The sample is desolvated in the usual way at low temperature and then flashed off at high temperature into the injector gas flow.As scan times of as little as 20 ms may be used this provides an ample number of scans over the changing signal to make isotope ratio measurements. On the other hand the scan width must be restricted to the minimum to avoid loss of signal while scanning beyond the important mass range. A typical single ion response from a 5-pl sample is shown in Fig. 18 and the spectrum obtained from 500 ng of cadmium in Fig. 19. A short pulse of ions lasting a few seconds is obtained. 114( Fig. 19. Spectrum of 500ng of cadmium by electrothermal vaporisation. Peak integral 114Cd+ 2.22 x lo5 counts. Excellent agreement has been obtained between isotope ratios determined on microsamples and nebulised solutions.Absolute detection limits ( 2 4 obtained from early runs on a few elements are shown in Table IV. The background spectrum is much simpler in the absence of water and hitherto unusable parts of the spectrum become available. The large 320,+ peak excludes the determination of sulphur by nebulisation but by thermal vaporisation the principal isotope ratios of sulphur have been determined on a single 200-ng sample to better than 1%. The promising results obtained even on very short duration samples introduced into the injector gas suggest that laser ablation from solids would be an attractive method of direct solids mass spectrometry with the very minimum of sample preparation. Direct solids introduction into a plasma ion source was always an aim of even the earliest work and it is intended to explore both laser ablation and arc aerosol generation with the ICP source.Both TABLE IV ABSOLUTE DETECTION LIMITS BY ELECTROTHERMAL VAPORISATION Measurements made on 5-pl samples. Element Detection limitlpg (20) As . . 7 Cd . . 2 Pb 1 Se . . 12 Zn . . * . 1050 GRAY AND DATE these techniques should be capable of greatly reducing the inefficient use of sample involved in preparing solutions of O.l-lyo of solid followed by the loss of 99% of the sample in the nebuliser. in sensitivity can be recovered the ultimate sensitivity of the method should greatly exceed that of any other laboratory method of comparable speed and convenience. If only part of this loss of at least This work is supported by the Institute of Geological Sciences and the R & D programme The authors thank Mr.P. J. The paper is published by on Uranium Exploration of the European Communities. Moore and Miss E. Waine for critically reading the manuscript. permission of The Director Institute of Geological Sciences (NERC). 1. 2. 3. 4. 6. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. References Gray A. L. Proc. SOC. Anal. Chem. 1974 11 182. Gray A. L. Analyst 1975 100 289. Gray A. L. “Dynamic Mass Spectrometry,” Volume 4 Heyden London 1976 pp. 153-162. Houk R. S. Fassel V. A. and Svec H. J. “Dynamic Mass Spectrometry,” Volume 6 Heyden, Houk R. S. Fassel V. A. Flesch G. D. Svec H. J. Gray A. L.and Taylor C. E. Anal. Chem., Douglas D. J. and French J. B. Anal. Chem. 1981 53 37. Gray A. L. and Date A. R. “Dynamic Mass Spectrometry,” Volume 6 Heyden London 1981, Steams C. A. Kohl F. J. Fryburg G. C. and Miller R. A. Nut. Bur. Stand. Spec. Publ. 1979, Douglas D. J. Quan E. S. K. and Smith R. G. Spectrochim. Acta Part B 1983 38 1/2 29. Date A. R. and Gray A. L. Analyst 1981 106 1255. Date A. R. and Gray A. L. Spectrochim. Acta Part B 1983 38 1/2 39. Gray A. L. and Date A. R. Int. J . Mass Spectrom. Ion Phys. 1983 46 7. Date A. R. and Gray A. L. Int. J . Mass Spectrom. Ion Phys. 1983 48 357. Date A. R. and Gray A. L. Analyst 1983 108 159. Fassel V. A, Pure Appl. Chem. 1977 49 1533. Human H. G. C. and Scott R. H. Spectrochim. Acta Part B 1976 31 459. Campargue R. Rev. Sci. Instrum. 1964 35 111. Greene F. T. Beachey J. E. and Milne T. A. Nut. Bur. Stand. Spec. Publ. 1979 No. 561 Volume Gray A. L. Eur. Spectrosc. News 1982 43 front cover. Boumans P. W. J. M. “Line Coincidence Tables for Inductively Coupled Plasma Atomic Emission Data Sheet 091/11/79 Instrumentation Laboratory Wilmington MA 1979. Alder J. F. Bombelke R. M. and Kirkbright G. F. Spectrochim. Ada Part B 1980 35 165. Kalnickey D. J. Fassel V. A and Knisely R. N. Appl. Spectrosc. 1977 31 137. Hulmston P. Analyst 1983 108 166. Gunn A. M. Millard D. L. and Kirkbright G. F. Analyst 1978 103 1066. London 1981 pp. 234-251. 1980 52 2283. pp. 252-266. No. 661 Volume 1 303. 1 431. Spectrometry,” Pergamon Press Oxford and New York 1980. Received Mavch 23rd. 1983 Accepted April Sth 198