- Email: [email protected]
Factors affecting analyte transport through the sampling orifice of an inductively coupled plasma mass spectrometer
Spectrochimica Acta Part B 56 Ž2001. 1687᎐1695
Factors affecting analyte transport through the sampling orifice of an inductively coupled plasma mass spectrometer 夽 Jeffrey H. Macedone, Dennis J. Gammon, Paul B. FarnsworthU Department of Chemistry and Biochemistry, Brigham Young Uni¨ ersity, Pro¨ o, UT 84602, USA Received 17 May 2001; accepted 5 June 2001
Abstract Laser-excited ionic fluorescence has been used to study the effects of sample matrix, operating conditions, and load coil shielding on analyte ion transport efficiency through the sampling orifice of an inductively coupled plasma mass spectrometer. Significant changes in ion transport efficiency result from changes in sample composition, RF forward power, nebulizer flow and torch shield configuration. The changes in ion transport efficiency correlate well with changes in the potential recorded on a single floating probe placed 1 mm upstream from the sampling orifice. 䊚 2001 Elsevier Science B.V. All rights reserved. Keywords: Inductively coupled plasma mass spectrometry; Matrix effects; Laser-induced fluorescence; Plasma potential; Efficiency
1. Introduction Inductively coupled plasma mass spectrometry ŽICP-MS. is a well-established technique for trace elemental analysis. However, degradation of ana夽 This article is published in a special honor issue dedicated to Walter Slavin, in recognition of his outstanding contributions to analytical atomic spectroscopy, in appreciation of all the time and energy spent in editing Spectrochimica Acta Part B. U Corresponding author. Fax: q1-801-378-6502. E-mail address: paul ᎏ farnsworth@ byu.edu Ž P.B. Farnsworth..
lytical performance due to matrix effects remains a weakness. Early studies of matrix effects in ICP-MS showed significant changes in instrument response with changing sample composition w1,2x. The matrix effects in ICP-MS are more severe than they are for ICP emission spectrometry, suggesting that they originate, at least in part, somewhere in the plasma᎐mass spectrometer interface, or in the mass spectrometer itself. Space charge effects have been identified as a possible source of matrix effects in the ICP-MS vacuum interface. Several research groups have conducted experiments designed to characterize w3᎐9x or reduce w10,11x space charge effects on ICP-MS
0584-8547r01r$ - see front matter 䊚 2001 Elsevier Science B.V. All rights reserved. PII: S 0 5 8 4 - 8 5 4 7 Ž 0 1 . 0 0 2 6 2 - 2
1688
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
performance. The working hypothesis in those studies was that changes in sample matrix compositions cause changes in the total ion density in the beam extracted from the atmospheric-pressure ICP. As the beam loses its charge neutrality, the altered positive charge density affects the efficiency with which ions can be guided through the vacuum interface into the mass analyzer of the instrument. The research reported in this paper has its origins in studies of space charge effects in ICPMS w12,13x. We began our experiments with the assumption that matrix effects unique to ICP-MS were occurring primarily in the second vacuum stage of the differentially pumped vacuum interface. We used laser-excited fluorescence to characterize the ion beam in the second vacuum stage of a mock-up instrument. While our studies in the second vacuum stage confirmed the existence of matrix-induced changes in the ion beam, they also pointed to changes farther upstream, in the first vacuum stage of the interface. Consequently, we shifted our attention to the first stage of the vacuum interface, and in particular, to the effi-
ciency with which ions are transported through the orifice in the sampling cone that is immersed in the plasma serving as the ion source for ICPMS. In our initial studies of the vacuum interface, we discovered that the addition of a grounded shield between the load coil and the plasma torch increased the ion signals in the second vacuum stage by more than a factor of two w14x. Because of this observation, it was natural to extend the study of transport efficiency through the sampling orifice to include not only matrix effects, but also the effects of the torch shield and other plasma operating parameters that could affect plasma potential.
2. Experimental 2.1. Instrumentation 2.1.1. Optics Much of the current instrumentation has been previously described w9,12᎐14x, and only a brief
Fig. 1. Schematic diagram of instrumentation used to measure transport efficiency.
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
description will be presented here. The arrangement used in the current set of experiments is shown in Fig. 1. The beam from an excimerpumped dye laser was split by a 50:50 beamsplitter. Each of the split beams was focused into a 0.4-mm fused silica optical fiber. One of the fibers passed through a vacuum feedthrough into the vacuum chamber that served as a mock-up of the first vacuum stage of an ICP-MS. Laser radiation emerging from the multimode fiber was collimated, turned 90⬚ with a folding prism, and then focused to a spot by a second lens. The spot was an image of the fiber end magnified by a factor of approximately two. A second set of optics, mounted in a plane at right angles to the plane of the excitation optics, served to collect the laser excited fluorescence and focus it into a 0.4-mm fiber that carried the radiation out of the vacuum chamber to a detector module. The detector consisted of a collimating lens, a narrow-band interference filter, a focusing lens, a pinhole, and a photomultiplier tube. The excitation and collection optics inside the vacuum chamber were mounted on a rigid aluminum frame, which was in turn mounted with four stainless steel posts to a computer-controlled x᎐y᎐z stage. The aluminum frame was watercooled to prevent heat damage to the optics. The frame and cooling-jacket assembly had a 2-inch hole, through which most of the hot gas in the supersonic expansion passed. Behind the hole, an aluminum deflection plate prevented the gases from reaching and heating the motorized stages. The volume probed was defined by the intersection of the excitation and collection optical paths. The probe volume, measured by techniques described in w13x, was approximately 0.4 mm3. For all of the measurements in this study, the probe volume was positioned 10 mm downstream from the sampling cone in the position normally occupied by the skimmer cone. The skimmer support plate and cone were removed for the experiments. The fiber containing the second half of the split laser beam was directed to a second set of excitation optics outside the vacuum chamber that focused the laser beam to a spot 1 mm upstream
1689
Table 1 Instrument operating conditions Incident power Outer gas flow Intermediate gas flow Nebulizer gas flow First stage pressure Sampling depth
1.25 kW 14 l miny1 0.4 l miny1 1.4 l miny1 1.0 torr 10 mm
from the sampling orifice. Collection optics on an axis at an angle of 45⬚ with respect to the excitation axis gathered fluorescence and passed it through a fiber to a second detection module, similar to the one described above. 2.1.2. Plasma generator and ¨ acuum chamber The plasma was maintained by a 27-MHz generator ŽICPr5500, Perkin Elmer, Norwalk, CT.. The three-turn load coil was grounded on the downstream end by a short, braided cable that was attached to the grounded vacuum interface. Except where noted, a grounded shield was placed between the load coil and the torch. The sampling cone with a 1-mm-diameter orifice was located 10 mm from the downstream end of the load coil. The vacuum chamber downstream from the sampling cone was maintained at a pressure of 1 torr. Samples were introduced into the plasma with an ultrasonic nebulizer ŽU 5000, Cetac, Omaha, NE.. Except where otherwise noted, the plasma operating conditions were as shown in Table 1. 2.1.3. Potential measurements A sketch of the apparatus constructed for the plasma potential experiments is shown in Fig. 2. A 1r16-inch-diameter tungsten rod served as the probe. A 1r8-inch alumina sleeve insulated all but a 1-mm section of the top of the tungsten rod. A setscrew made contact with the tungsten rod and held the probe within an aluminum solenoid adapter. A wire soldered to the setscrew provided an electrical connection to the probe. Potential measurements of the plasma represent the voltage difference between the probe and electrical ground at the interface, as collected by a 10X probe connected to a hand-held oscilloscope ŽModel THS 720A, Tektronix, Beaverton, OR..
1690
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
2.2. Solutions We used distilled deionized water ŽMilli-Q RG, Millipore, Bedford, MA. to prepare barium, lead, lithium, and magnesium solutions from their respective nitrates. Cesium and zinc solutions were prepared from chlorides. Scandium solution was prepared from Sc 2 O 3 . Lithium was chosen as the element for matrix concentration studies because it allowed for high molarities without clogging the sampling cone. 2.3. Analytes
Fig. 2. Schematic diagram of potential probe.
An RS-232 connection on the oscilloscope facilitated transfer of the data to a computer. An interface box controlled by a function generator ŽModel 191, Wavetek, San Diego, CA. actuated the solenoid. A single square-wave pulse width of 100 ms at the function generator resulted in a total actuation time of approximately 200 ms. Actuation of the solenoid directed the probe to the plasma center, 1 mm upstream of the sampling cone tip. A spring attached to the bottom of the solenoid plunger ensured probe retraction. Typical traces collected by the oscilloscope during probe immersion into the plasma are shown in Fig. 3. Although we were primarily interested in the DC component, both DC and RF components of the voltage are apparent. The RF component is badly aliased, so the frequency appears to be much lower than the 27-MHz generator frequency. The RF amplitude is accurately represented. The RF voltages recorded before and after the probe was actuated are indicative of the magnitude of RF pickup by the measuring system. The voltages reported in this paper are the DC components.
Barium and scandium were chosen as the test analytes, because both have efficient non-resonance fluorescence schemes that allow for effective discrimination against scattered light. The excitation and emission wavelengths for barium ions were 455.404 and 614.172 nm, respectively. For scandium, the respective values were 358.094 and 432.501 nm. All solutions used in the experiments were either 10 g mly1 in barium or scandium, with matrix species added as indicated. 2.4. Signal processing In our previous measurements of fluorescence inside the vacuum interface of an ICP-MS w13x, we compensated for changes in the source plasma by making sequential measurements of ion density inside the skimmer cone, then upstream from the sampling orifice. The measurements inside the skimmer cone were subsequently normalized to the upstream measurements. That approach
Fig. 3. Representative oscilloscope traces of potential on the voltage probe: ---- shield grounded; —— shield floating.
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
suffered from a severe drawback, in that it was susceptible to drift during the time required for the measurement. In the current experiments, the measurements were carried out simultaneously. The signals from the photomultiplier tubes on the external and internal channels were fed into identical boxcar averagers ŽModel 250, Stanford Research Systems, Sunnyvale, CA.. The outputs of the boxcar averagers were then digitized using two channels of a four-channel ArD card. Ratios of the two signals were then used as an indicator of the relative efficiency with which ions were transported through the sampling orifice. Finally, the relative efficiencies were normalized to the value recorded for a solution of pure analyte obtained under the conditions given in Table 1. Representative raw and ratioed data for one set of measurements are shown in Figs. 4 and 5. The measurements alternated between a solution containing only 10 g mly1 of Sc, and solutions containing the same concentration of Sc and increasing concentrations of Li. Several features of these data are noteworthy. The absolute magnitudes of the individual signals in Fig. 4 are not significant. The internal fluorescence signal is initially larger than the external fluorescence signal because of differences between the two channels in excitation volume, collection efficiency, and electronic gain. It is significant that the addition of Li suppresses fluorescence at both locations, but the suppression inside the vacuum chamber is larger than it is upstream from the sampling orifice. Large-scale fluctuations in the two signals are well correlated, but as shown by the residual noise in the ratios plotted in Fig. 5, significant uncorrelated noise remains as well. We presume that the major contributions to the uncorrelated noise are shot noise and inhomogeneities in the plasma on the scale defined by the separation of the two probes. The uncertainties in the ratios reported in subsequent figures and tables are based on the standard deviations of the ratios of the two signals. In practice, these reported values overestimate the short-term uncertainty of the ratios, which should be reflected in the standard deviations of the means of the values in each segment of the experiment, but give a good indication of the relative levels of noise in the mea-
1691
Fig. 4. Raw two-channel scandium fluorescence data: ----external optics; —— internal optics. Solutions of 10 g mly1 scandium were alternated with solutions containing 10 g mly1 scandium and increasing concentrations of lithium.
surements. Perhaps the best estimate of the uncertainty in the ratios can be obtained from the replicate measurements of pure analyte solutions carried out between each solution containing Li shown in Fig. 5, and from an analogous series of measurements carried out for Ba. The relative standard deviation for seven replicate ratio measurements of a 10-g mly1 solution of Sc was 1.7%. For the replicate measurements with a 10-g mly1 solution of Ba, the relative standard deviation was 4.3%. We should note that simultaneous detection at the two locations inside and outside the vacuum chamber was possible using a split laser beam,
Fig. 5. Ratio of internalrexternal scandium fluorescence signals from Fig. 4.
1692
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
despite the fact that the peak excitation wavelengths were not identical. The excitation inside the vacuum chamber was red-shifted due to motion of ions in the supersonic expansion toward the excitation optics. However, the shift was smaller than the laser linewidth. In practice, we tuned the laser to the excitation peak inside the vacuum chamber and accepted the loss in signal that resulted from being off-peak in the external optical channel. Fig. 6. Relative efficiency of analyte transmission through the sampling orifice of the ICP-MS as a function of lithium matrix concentration: ` barium; ^ scandium.
3. Results and discussion 3.1. Effect of matrix concentration on transmission efficiency
because the two elements differ in valence, mass and ionization potential. However, the small transport enhancement of the Ba efficiency observed for Zn compared to the suppressions for the other matrix species studied suggests that ionization potential of the matrix element is important.
The effect of the concentration of Li on the transmission efficiencies of Ba and Sc analytes is shown in Fig. 6. The two analytes respond similarly to increasing Li concentration. For a Li concentration of 7 mM Ž50 g mly1 ., the loss in transport efficiency through the sampling orifice is already significant.
3.3. Effect of plasma operating parameters
3.2. Effect of matrix mass and ionization potential on transmission efficiency
The effect of nebulizer flow on the transport efficiencies of the two analytes is shown in Fig. 7. No clear trend is apparent in the data, with the exception that efficiency drops for both analytes at flows above 1.4 l miny1 . The effect of RF forward power on analyte transport efficiency is shown in Fig. 8. The efficiency for both analytes drops dramatically as power is reduced. Transport efficiencies with the torch shield floated and grounded are listed in Table 3. Again, it is important to note that the uncertainties
The effects of mass and ionization potential on ion transmission efficiency were studied using pairs of matrix elements with extremes of mass ŽMg and Pb. and ionization potential ŽCs and Zn. ŽTable 2.. Mg and Pb are a particularly convenient pair for the study of matrix mass, because their ionization potentials are nearly the same. The Mg᎐Pb comparison indicates that the mass of the matrix is relatively unimportant. Interpretation of the Cs᎐Zn comparison is more difficult,
Table 2 Effect of matrix mass and ionization potential on transmission efficiency Matrix
None Pb Mg Zn Cs
Mass Žamu.
Ionization potential ŽeV.
Concentration ŽmM.
Relative transport efficiency Ba
Sc
᎐ 207.2 24.31 65.39 132.9
᎐ 7.42 7.64 9.93 3.89
᎐ 4.8 4.8 5.0 5.0
1 0.79" 0.02 0.85" 0.02 1.11" 0.05 0.79 " 0.02
1 0.92" 0.02 0.88" 0.02 0.96" 0.02 0.85" 0.02
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
Fig. 7. Relative efficiency of analyte transmission through the sampling orifice as a function of nebulizer gas flow: ` barium; ^ scandium.
Fig. 8. Relative efficiency of analyte transmission through the sampling orifice as a function of RF power: ` barium; ^ scandium.
quoted for the transport efficiency are short-term. Experience has shown that the changes in efficiency caused by changes in shield grounding can vary widely, depending on the condition of the torch, the cone and the shield. 3.4. Plasma potential and transport efficiencies As was mentioned in Section 1, the changes in Table 3 Effect of shield grounding on transport efficiency Shield connection
Relative transport efficiency Ba
Sc
Grounded Floating
1 0.08" 0.01
1 0.16" 0.01
1693
Fig. 9. Relative ion transmission efficiency vs. probe potential. The symbols indicate from which set of experiments the data were taken. Note that the cluster of points with an efficiency of 1.00 corresponds to the reference measurements against which efficiency values were determined.
transport efficiency associated with grounding and floating the torch shield led us to look for a relationship between plasma potential and transport efficiency. We repeated each of the experiments in the forgoing sections in which Ba was used as an analyte and recorded the potential of our probe instead of transport efficiency. The correlation between the efficiency and potential measurements is illustrated in Fig. 9. Clearly, the correlation is a strong one. However, interpretation of these results is not straightforward. We record our optimum efficiencies at probe potentials between y13 and y15 V. The measured potentials are not simply the plasma potential, but also include significant contributions from floating potentials that the probe assumes as a consequence of differences between the mobilities of ions and electrons, and differences between the temperature of the probe and that of the sampling cone w15x. The sign and magnitude of the floating voltage depends on conditions in the plasma and the configuration of the load coil. Gray et al. w16x reported positive potentials at the center of the plasma ranging between q6 and q18 V for a load coil grounded at the downstream end. Houk et al. w17x reported negative potentials at the plasma center ranging between 0.0 and y2.6 V for a center-tapped load coil. Our measurements differ from these earlier reports, in that the probe diameter is larger Ž1.6 vs. 1.0 mm. and the measurements were carried out
1694
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
closer to the tip of the sampler Ž1 vs. ; 3 mm.. With the torch shield in place and grounded, the absence of any secondary discharge, the long lifetime of the sampling cones, and the observation of peak transport efficiencies all lead us to believe that the negative values we have recorded reflect the floating probe voltage rather than the plasma potential. If such is the case, then changes in the measured voltage to less negative values are indicative of increasingly positive plasma potentials. We should note, however, that the floating probe potential should be affected by changes in the plasma conditions, and that it is impossible with this type of single-electrode arrangement to measure a true plasma potential.
4. Conclusions Despite the ambiguities in the potential measurements, they underscore the importance of controlling plasma potential in ICP-MS measurements. Previous discussions of the negative effects of uncontrolled plasma potential have focused on orifice erosion, formation of doubly charged ions, and increased spread in ion kinetic energies w18x. Existing models of the ICP-MS vacuum interface do not provide a ready explanation for changes in ion transport efficiency with changes in sample composition or plasma potential. They are based on neutral flow through the sampling cone, and do not consider the effects of charged species w19x. The correlations between efficiency and potential observed in this study, combined with the lack of any changes in first-stage vacuum pressure associated with changes in transport efficiency, lead to the conclusion that a chargerelated mechanism is affecting flow through the sampling orifice. Additional experiments are in progress in our laboratory that are designed to gain a better understanding of the changes in transport efficiency reported in this study. They include studies of different load coil and grounding geometries and detailed mapping of flow and temperature fields for neutral and charged species in the first vacuum stage of the ICP-MS interface.
Acknowledgements The authors gratefully acknowledge support for this work received from the Brigham Young University College of Physical and Mathematical Sciences and the Office of Research and Creative Activities. Equipment used in the work was developed with support from the National Science Foundation under grant 9415384. References w1x J.A. Olivares, R.S. Houk, Suppression of analyte signals by various concomitant salts in inductively coupled plasma mass spectrometry, Anal. Chem. 58 Ž1986. 20᎐25. w2x S.H. Tan, G. Horlick, Matrix-effect observations in inductively coupled plasma mass spectrometry, J. Anal. At. Spectrom. 2 Ž1987. 745᎐763. w3x G.R. Gillson, D.J. Douglas, J.E. Fulford, K.W. Haligan, S.D. Tanner, Non-spectroscopic interelement interferences in inductively coupled plasma mass spectrometry, Anal. Chem. 60 Ž1988. 1472᎐1474. w4x S.D. Tanner, Space charge in ICP-MS: calculations and implications, Spectrochim. Acta Part B 47 Ž1992. 809᎐823. w5x G. Li, X. Duan, G.M. Hieftje, Space charge effects and ion distributions in plasma source mass spectrometry, J. Mass Spectrom. 30 Ž1995. 841᎐848. w6x X. Chen, R.S. Houk, Spatially resolved measurements of ion density behind the skimmer of an inductively coupled plasma mass spectrometer, Spectrochim. Acta Part B 51 Ž1996. 41᎐54. w7x L.A. Allen, J.J. Leach, R.S. Houk, Spatial location of the space charge effect in individual ion clouds using monodisperse dried microparticulate injection with a twin quadrupole inductively coupled plasma mass spectrometer, Anal. Chem. 69 Ž1997. 2384᎐2391. w8x J.W. Olesik, M.P. Dziewatkoski, Time-resolved measurements of individual ions cloud signals to investigate space charge effects in plasma mass spectrometry, J. Am. Soc. Mass Spectrom. 7 Ž1996. 362᎐367. w9x Y. Chen, P.B. Farnsworth, Ion deposition experiments as a tool for the study of the spatial distribution of analyte ions in the second vacuum stage of an inductively coupled plasma mass spectrometer, Spectrochim. Acta Part B 52 Ž1997. 231᎐239. w10x N. Praphairaksit, R.S. Houk, Attenuation of matrix effects in inductively coupled plasma mass spectrometry with a supplemental electron source inside the skimmer, Anal. Chem. 72 Ž2000. 2351᎐2355. w11x N. Praphairaksit, R.S. Houk, Reduction of space charge effects in inductively coupled plasma mass spectrometry using a supplemental electron source inside the skimmer: ion transmission and mass spectral characteristics, Anal. Chem. 72 Ž2000. 2356᎐2361.
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695 w12x B.S. Duersch, Y. Chen, A. Ciocan, P.B. Farnsworth, Optical measurements of ion density in the second vacuum stage of an inductively coupled plasma mass spectrometer, Spectrochim. Acta Part B 53 Ž1998. 569᎐579. w13x B.S. Duersch, P.B. Farnsworth, Characterization of the ion beam inside the skimmer cone of an inductively coupled plasma mass spectrometer by laser-excited atomic and ionic fluorescence, Spectrochim. Acta Part B 54 Ž1999. 545᎐555. w14x B.S. Duersch, J.E. Patterson, P.B. Farnsworth, The effects of a torch shield on performance of the vacuum interface of an inductively coupled plasma mass spectrometer, J. Anal. At. Spectrom. 14 Ž1999. 615᎐619. w15x H. Niu, S. Luan, H. Pang, R.S. Houk, Langmuir probe measurements of the ion extraction process in inductively coupled plasma-mass spectrometry. Part 2. Mea-
w16x
w17x
w18x w19x
1695
surements of floating voltage and radiofrequency voltage, Spectrochim. Acta Part B 50 Ž1995. 1247᎐1261. A.L. Gray, R.S. Houk, J.G. Williams, Langmuir probe potential measurements in the plasma and their correlation with mass spectral characteristics in inductively coupled plasma mass spectrometry, J. Anal. At. Spectrom. 2 Ž1987. 13᎐20. R.S. Houk, J.K. Schoer, J.S. Crain, Plasma potential measurements for inductively coupled plasma mass spectrometry with a centre-tapped load coil, J. Anal. At. Spectrom. 2 Ž1987. 283᎐286. D.J. Douglas, J.B. French, An improved interface for inductively coupled plasma-mass spectrometry ŽICP-MS., Spectrochim. Acta Part B 41 Ž1986. 197᎐204. D.J. Douglas, J.B. French, Gas dynamics of the inductively coupled plasma mass spectrometry interface, J. Anal. At. Spectrom. 3 Ž1988. 743᎐747.
Factors affecting analyte transport through the sampling orifice of an inductively coupled plasma mass spectrometer 夽 Jeffrey H. Macedone, Dennis J. Gammon, Paul B. FarnsworthU Department of Chemistry and Biochemistry, Brigham Young Uni¨ ersity, Pro¨ o, UT 84602, USA Received 17 May 2001; accepted 5 June 2001
Abstract Laser-excited ionic fluorescence has been used to study the effects of sample matrix, operating conditions, and load coil shielding on analyte ion transport efficiency through the sampling orifice of an inductively coupled plasma mass spectrometer. Significant changes in ion transport efficiency result from changes in sample composition, RF forward power, nebulizer flow and torch shield configuration. The changes in ion transport efficiency correlate well with changes in the potential recorded on a single floating probe placed 1 mm upstream from the sampling orifice. 䊚 2001 Elsevier Science B.V. All rights reserved. Keywords: Inductively coupled plasma mass spectrometry; Matrix effects; Laser-induced fluorescence; Plasma potential; Efficiency
1. Introduction Inductively coupled plasma mass spectrometry ŽICP-MS. is a well-established technique for trace elemental analysis. However, degradation of ana夽 This article is published in a special honor issue dedicated to Walter Slavin, in recognition of his outstanding contributions to analytical atomic spectroscopy, in appreciation of all the time and energy spent in editing Spectrochimica Acta Part B. U Corresponding author. Fax: q1-801-378-6502. E-mail address: paul ᎏ farnsworth@ byu.edu Ž P.B. Farnsworth..
lytical performance due to matrix effects remains a weakness. Early studies of matrix effects in ICP-MS showed significant changes in instrument response with changing sample composition w1,2x. The matrix effects in ICP-MS are more severe than they are for ICP emission spectrometry, suggesting that they originate, at least in part, somewhere in the plasma᎐mass spectrometer interface, or in the mass spectrometer itself. Space charge effects have been identified as a possible source of matrix effects in the ICP-MS vacuum interface. Several research groups have conducted experiments designed to characterize w3᎐9x or reduce w10,11x space charge effects on ICP-MS
0584-8547r01r$ - see front matter 䊚 2001 Elsevier Science B.V. All rights reserved. PII: S 0 5 8 4 - 8 5 4 7 Ž 0 1 . 0 0 2 6 2 - 2
1688
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
performance. The working hypothesis in those studies was that changes in sample matrix compositions cause changes in the total ion density in the beam extracted from the atmospheric-pressure ICP. As the beam loses its charge neutrality, the altered positive charge density affects the efficiency with which ions can be guided through the vacuum interface into the mass analyzer of the instrument. The research reported in this paper has its origins in studies of space charge effects in ICPMS w12,13x. We began our experiments with the assumption that matrix effects unique to ICP-MS were occurring primarily in the second vacuum stage of the differentially pumped vacuum interface. We used laser-excited fluorescence to characterize the ion beam in the second vacuum stage of a mock-up instrument. While our studies in the second vacuum stage confirmed the existence of matrix-induced changes in the ion beam, they also pointed to changes farther upstream, in the first vacuum stage of the interface. Consequently, we shifted our attention to the first stage of the vacuum interface, and in particular, to the effi-
ciency with which ions are transported through the orifice in the sampling cone that is immersed in the plasma serving as the ion source for ICPMS. In our initial studies of the vacuum interface, we discovered that the addition of a grounded shield between the load coil and the plasma torch increased the ion signals in the second vacuum stage by more than a factor of two w14x. Because of this observation, it was natural to extend the study of transport efficiency through the sampling orifice to include not only matrix effects, but also the effects of the torch shield and other plasma operating parameters that could affect plasma potential.
2. Experimental 2.1. Instrumentation 2.1.1. Optics Much of the current instrumentation has been previously described w9,12᎐14x, and only a brief
Fig. 1. Schematic diagram of instrumentation used to measure transport efficiency.
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
description will be presented here. The arrangement used in the current set of experiments is shown in Fig. 1. The beam from an excimerpumped dye laser was split by a 50:50 beamsplitter. Each of the split beams was focused into a 0.4-mm fused silica optical fiber. One of the fibers passed through a vacuum feedthrough into the vacuum chamber that served as a mock-up of the first vacuum stage of an ICP-MS. Laser radiation emerging from the multimode fiber was collimated, turned 90⬚ with a folding prism, and then focused to a spot by a second lens. The spot was an image of the fiber end magnified by a factor of approximately two. A second set of optics, mounted in a plane at right angles to the plane of the excitation optics, served to collect the laser excited fluorescence and focus it into a 0.4-mm fiber that carried the radiation out of the vacuum chamber to a detector module. The detector consisted of a collimating lens, a narrow-band interference filter, a focusing lens, a pinhole, and a photomultiplier tube. The excitation and collection optics inside the vacuum chamber were mounted on a rigid aluminum frame, which was in turn mounted with four stainless steel posts to a computer-controlled x᎐y᎐z stage. The aluminum frame was watercooled to prevent heat damage to the optics. The frame and cooling-jacket assembly had a 2-inch hole, through which most of the hot gas in the supersonic expansion passed. Behind the hole, an aluminum deflection plate prevented the gases from reaching and heating the motorized stages. The volume probed was defined by the intersection of the excitation and collection optical paths. The probe volume, measured by techniques described in w13x, was approximately 0.4 mm3. For all of the measurements in this study, the probe volume was positioned 10 mm downstream from the sampling cone in the position normally occupied by the skimmer cone. The skimmer support plate and cone were removed for the experiments. The fiber containing the second half of the split laser beam was directed to a second set of excitation optics outside the vacuum chamber that focused the laser beam to a spot 1 mm upstream
1689
Table 1 Instrument operating conditions Incident power Outer gas flow Intermediate gas flow Nebulizer gas flow First stage pressure Sampling depth
1.25 kW 14 l miny1 0.4 l miny1 1.4 l miny1 1.0 torr 10 mm
from the sampling orifice. Collection optics on an axis at an angle of 45⬚ with respect to the excitation axis gathered fluorescence and passed it through a fiber to a second detection module, similar to the one described above. 2.1.2. Plasma generator and ¨ acuum chamber The plasma was maintained by a 27-MHz generator ŽICPr5500, Perkin Elmer, Norwalk, CT.. The three-turn load coil was grounded on the downstream end by a short, braided cable that was attached to the grounded vacuum interface. Except where noted, a grounded shield was placed between the load coil and the torch. The sampling cone with a 1-mm-diameter orifice was located 10 mm from the downstream end of the load coil. The vacuum chamber downstream from the sampling cone was maintained at a pressure of 1 torr. Samples were introduced into the plasma with an ultrasonic nebulizer ŽU 5000, Cetac, Omaha, NE.. Except where otherwise noted, the plasma operating conditions were as shown in Table 1. 2.1.3. Potential measurements A sketch of the apparatus constructed for the plasma potential experiments is shown in Fig. 2. A 1r16-inch-diameter tungsten rod served as the probe. A 1r8-inch alumina sleeve insulated all but a 1-mm section of the top of the tungsten rod. A setscrew made contact with the tungsten rod and held the probe within an aluminum solenoid adapter. A wire soldered to the setscrew provided an electrical connection to the probe. Potential measurements of the plasma represent the voltage difference between the probe and electrical ground at the interface, as collected by a 10X probe connected to a hand-held oscilloscope ŽModel THS 720A, Tektronix, Beaverton, OR..
1690
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
2.2. Solutions We used distilled deionized water ŽMilli-Q RG, Millipore, Bedford, MA. to prepare barium, lead, lithium, and magnesium solutions from their respective nitrates. Cesium and zinc solutions were prepared from chlorides. Scandium solution was prepared from Sc 2 O 3 . Lithium was chosen as the element for matrix concentration studies because it allowed for high molarities without clogging the sampling cone. 2.3. Analytes
Fig. 2. Schematic diagram of potential probe.
An RS-232 connection on the oscilloscope facilitated transfer of the data to a computer. An interface box controlled by a function generator ŽModel 191, Wavetek, San Diego, CA. actuated the solenoid. A single square-wave pulse width of 100 ms at the function generator resulted in a total actuation time of approximately 200 ms. Actuation of the solenoid directed the probe to the plasma center, 1 mm upstream of the sampling cone tip. A spring attached to the bottom of the solenoid plunger ensured probe retraction. Typical traces collected by the oscilloscope during probe immersion into the plasma are shown in Fig. 3. Although we were primarily interested in the DC component, both DC and RF components of the voltage are apparent. The RF component is badly aliased, so the frequency appears to be much lower than the 27-MHz generator frequency. The RF amplitude is accurately represented. The RF voltages recorded before and after the probe was actuated are indicative of the magnitude of RF pickup by the measuring system. The voltages reported in this paper are the DC components.
Barium and scandium were chosen as the test analytes, because both have efficient non-resonance fluorescence schemes that allow for effective discrimination against scattered light. The excitation and emission wavelengths for barium ions were 455.404 and 614.172 nm, respectively. For scandium, the respective values were 358.094 and 432.501 nm. All solutions used in the experiments were either 10 g mly1 in barium or scandium, with matrix species added as indicated. 2.4. Signal processing In our previous measurements of fluorescence inside the vacuum interface of an ICP-MS w13x, we compensated for changes in the source plasma by making sequential measurements of ion density inside the skimmer cone, then upstream from the sampling orifice. The measurements inside the skimmer cone were subsequently normalized to the upstream measurements. That approach
Fig. 3. Representative oscilloscope traces of potential on the voltage probe: ---- shield grounded; —— shield floating.
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
suffered from a severe drawback, in that it was susceptible to drift during the time required for the measurement. In the current experiments, the measurements were carried out simultaneously. The signals from the photomultiplier tubes on the external and internal channels were fed into identical boxcar averagers ŽModel 250, Stanford Research Systems, Sunnyvale, CA.. The outputs of the boxcar averagers were then digitized using two channels of a four-channel ArD card. Ratios of the two signals were then used as an indicator of the relative efficiency with which ions were transported through the sampling orifice. Finally, the relative efficiencies were normalized to the value recorded for a solution of pure analyte obtained under the conditions given in Table 1. Representative raw and ratioed data for one set of measurements are shown in Figs. 4 and 5. The measurements alternated between a solution containing only 10 g mly1 of Sc, and solutions containing the same concentration of Sc and increasing concentrations of Li. Several features of these data are noteworthy. The absolute magnitudes of the individual signals in Fig. 4 are not significant. The internal fluorescence signal is initially larger than the external fluorescence signal because of differences between the two channels in excitation volume, collection efficiency, and electronic gain. It is significant that the addition of Li suppresses fluorescence at both locations, but the suppression inside the vacuum chamber is larger than it is upstream from the sampling orifice. Large-scale fluctuations in the two signals are well correlated, but as shown by the residual noise in the ratios plotted in Fig. 5, significant uncorrelated noise remains as well. We presume that the major contributions to the uncorrelated noise are shot noise and inhomogeneities in the plasma on the scale defined by the separation of the two probes. The uncertainties in the ratios reported in subsequent figures and tables are based on the standard deviations of the ratios of the two signals. In practice, these reported values overestimate the short-term uncertainty of the ratios, which should be reflected in the standard deviations of the means of the values in each segment of the experiment, but give a good indication of the relative levels of noise in the mea-
1691
Fig. 4. Raw two-channel scandium fluorescence data: ----external optics; —— internal optics. Solutions of 10 g mly1 scandium were alternated with solutions containing 10 g mly1 scandium and increasing concentrations of lithium.
surements. Perhaps the best estimate of the uncertainty in the ratios can be obtained from the replicate measurements of pure analyte solutions carried out between each solution containing Li shown in Fig. 5, and from an analogous series of measurements carried out for Ba. The relative standard deviation for seven replicate ratio measurements of a 10-g mly1 solution of Sc was 1.7%. For the replicate measurements with a 10-g mly1 solution of Ba, the relative standard deviation was 4.3%. We should note that simultaneous detection at the two locations inside and outside the vacuum chamber was possible using a split laser beam,
Fig. 5. Ratio of internalrexternal scandium fluorescence signals from Fig. 4.
1692
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
despite the fact that the peak excitation wavelengths were not identical. The excitation inside the vacuum chamber was red-shifted due to motion of ions in the supersonic expansion toward the excitation optics. However, the shift was smaller than the laser linewidth. In practice, we tuned the laser to the excitation peak inside the vacuum chamber and accepted the loss in signal that resulted from being off-peak in the external optical channel. Fig. 6. Relative efficiency of analyte transmission through the sampling orifice of the ICP-MS as a function of lithium matrix concentration: ` barium; ^ scandium.
3. Results and discussion 3.1. Effect of matrix concentration on transmission efficiency
because the two elements differ in valence, mass and ionization potential. However, the small transport enhancement of the Ba efficiency observed for Zn compared to the suppressions for the other matrix species studied suggests that ionization potential of the matrix element is important.
The effect of the concentration of Li on the transmission efficiencies of Ba and Sc analytes is shown in Fig. 6. The two analytes respond similarly to increasing Li concentration. For a Li concentration of 7 mM Ž50 g mly1 ., the loss in transport efficiency through the sampling orifice is already significant.
3.3. Effect of plasma operating parameters
3.2. Effect of matrix mass and ionization potential on transmission efficiency
The effect of nebulizer flow on the transport efficiencies of the two analytes is shown in Fig. 7. No clear trend is apparent in the data, with the exception that efficiency drops for both analytes at flows above 1.4 l miny1 . The effect of RF forward power on analyte transport efficiency is shown in Fig. 8. The efficiency for both analytes drops dramatically as power is reduced. Transport efficiencies with the torch shield floated and grounded are listed in Table 3. Again, it is important to note that the uncertainties
The effects of mass and ionization potential on ion transmission efficiency were studied using pairs of matrix elements with extremes of mass ŽMg and Pb. and ionization potential ŽCs and Zn. ŽTable 2.. Mg and Pb are a particularly convenient pair for the study of matrix mass, because their ionization potentials are nearly the same. The Mg᎐Pb comparison indicates that the mass of the matrix is relatively unimportant. Interpretation of the Cs᎐Zn comparison is more difficult,
Table 2 Effect of matrix mass and ionization potential on transmission efficiency Matrix
None Pb Mg Zn Cs
Mass Žamu.
Ionization potential ŽeV.
Concentration ŽmM.
Relative transport efficiency Ba
Sc
᎐ 207.2 24.31 65.39 132.9
᎐ 7.42 7.64 9.93 3.89
᎐ 4.8 4.8 5.0 5.0
1 0.79" 0.02 0.85" 0.02 1.11" 0.05 0.79 " 0.02
1 0.92" 0.02 0.88" 0.02 0.96" 0.02 0.85" 0.02
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
Fig. 7. Relative efficiency of analyte transmission through the sampling orifice as a function of nebulizer gas flow: ` barium; ^ scandium.
Fig. 8. Relative efficiency of analyte transmission through the sampling orifice as a function of RF power: ` barium; ^ scandium.
quoted for the transport efficiency are short-term. Experience has shown that the changes in efficiency caused by changes in shield grounding can vary widely, depending on the condition of the torch, the cone and the shield. 3.4. Plasma potential and transport efficiencies As was mentioned in Section 1, the changes in Table 3 Effect of shield grounding on transport efficiency Shield connection
Relative transport efficiency Ba
Sc
Grounded Floating
1 0.08" 0.01
1 0.16" 0.01
1693
Fig. 9. Relative ion transmission efficiency vs. probe potential. The symbols indicate from which set of experiments the data were taken. Note that the cluster of points with an efficiency of 1.00 corresponds to the reference measurements against which efficiency values were determined.
transport efficiency associated with grounding and floating the torch shield led us to look for a relationship between plasma potential and transport efficiency. We repeated each of the experiments in the forgoing sections in which Ba was used as an analyte and recorded the potential of our probe instead of transport efficiency. The correlation between the efficiency and potential measurements is illustrated in Fig. 9. Clearly, the correlation is a strong one. However, interpretation of these results is not straightforward. We record our optimum efficiencies at probe potentials between y13 and y15 V. The measured potentials are not simply the plasma potential, but also include significant contributions from floating potentials that the probe assumes as a consequence of differences between the mobilities of ions and electrons, and differences between the temperature of the probe and that of the sampling cone w15x. The sign and magnitude of the floating voltage depends on conditions in the plasma and the configuration of the load coil. Gray et al. w16x reported positive potentials at the center of the plasma ranging between q6 and q18 V for a load coil grounded at the downstream end. Houk et al. w17x reported negative potentials at the plasma center ranging between 0.0 and y2.6 V for a center-tapped load coil. Our measurements differ from these earlier reports, in that the probe diameter is larger Ž1.6 vs. 1.0 mm. and the measurements were carried out
1694
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695
closer to the tip of the sampler Ž1 vs. ; 3 mm.. With the torch shield in place and grounded, the absence of any secondary discharge, the long lifetime of the sampling cones, and the observation of peak transport efficiencies all lead us to believe that the negative values we have recorded reflect the floating probe voltage rather than the plasma potential. If such is the case, then changes in the measured voltage to less negative values are indicative of increasingly positive plasma potentials. We should note, however, that the floating probe potential should be affected by changes in the plasma conditions, and that it is impossible with this type of single-electrode arrangement to measure a true plasma potential.
4. Conclusions Despite the ambiguities in the potential measurements, they underscore the importance of controlling plasma potential in ICP-MS measurements. Previous discussions of the negative effects of uncontrolled plasma potential have focused on orifice erosion, formation of doubly charged ions, and increased spread in ion kinetic energies w18x. Existing models of the ICP-MS vacuum interface do not provide a ready explanation for changes in ion transport efficiency with changes in sample composition or plasma potential. They are based on neutral flow through the sampling cone, and do not consider the effects of charged species w19x. The correlations between efficiency and potential observed in this study, combined with the lack of any changes in first-stage vacuum pressure associated with changes in transport efficiency, lead to the conclusion that a chargerelated mechanism is affecting flow through the sampling orifice. Additional experiments are in progress in our laboratory that are designed to gain a better understanding of the changes in transport efficiency reported in this study. They include studies of different load coil and grounding geometries and detailed mapping of flow and temperature fields for neutral and charged species in the first vacuum stage of the ICP-MS interface.
Acknowledgements The authors gratefully acknowledge support for this work received from the Brigham Young University College of Physical and Mathematical Sciences and the Office of Research and Creative Activities. Equipment used in the work was developed with support from the National Science Foundation under grant 9415384. References w1x J.A. Olivares, R.S. Houk, Suppression of analyte signals by various concomitant salts in inductively coupled plasma mass spectrometry, Anal. Chem. 58 Ž1986. 20᎐25. w2x S.H. Tan, G. Horlick, Matrix-effect observations in inductively coupled plasma mass spectrometry, J. Anal. At. Spectrom. 2 Ž1987. 745᎐763. w3x G.R. Gillson, D.J. Douglas, J.E. Fulford, K.W. Haligan, S.D. Tanner, Non-spectroscopic interelement interferences in inductively coupled plasma mass spectrometry, Anal. Chem. 60 Ž1988. 1472᎐1474. w4x S.D. Tanner, Space charge in ICP-MS: calculations and implications, Spectrochim. Acta Part B 47 Ž1992. 809᎐823. w5x G. Li, X. Duan, G.M. Hieftje, Space charge effects and ion distributions in plasma source mass spectrometry, J. Mass Spectrom. 30 Ž1995. 841᎐848. w6x X. Chen, R.S. Houk, Spatially resolved measurements of ion density behind the skimmer of an inductively coupled plasma mass spectrometer, Spectrochim. Acta Part B 51 Ž1996. 41᎐54. w7x L.A. Allen, J.J. Leach, R.S. Houk, Spatial location of the space charge effect in individual ion clouds using monodisperse dried microparticulate injection with a twin quadrupole inductively coupled plasma mass spectrometer, Anal. Chem. 69 Ž1997. 2384᎐2391. w8x J.W. Olesik, M.P. Dziewatkoski, Time-resolved measurements of individual ions cloud signals to investigate space charge effects in plasma mass spectrometry, J. Am. Soc. Mass Spectrom. 7 Ž1996. 362᎐367. w9x Y. Chen, P.B. Farnsworth, Ion deposition experiments as a tool for the study of the spatial distribution of analyte ions in the second vacuum stage of an inductively coupled plasma mass spectrometer, Spectrochim. Acta Part B 52 Ž1997. 231᎐239. w10x N. Praphairaksit, R.S. Houk, Attenuation of matrix effects in inductively coupled plasma mass spectrometry with a supplemental electron source inside the skimmer, Anal. Chem. 72 Ž2000. 2351᎐2355. w11x N. Praphairaksit, R.S. Houk, Reduction of space charge effects in inductively coupled plasma mass spectrometry using a supplemental electron source inside the skimmer: ion transmission and mass spectral characteristics, Anal. Chem. 72 Ž2000. 2356᎐2361.
J.H. Macedone et al. r Spectrochimica Acta Part B: Atomic Spectroscopy 56 (2001) 1687᎐1695 w12x B.S. Duersch, Y. Chen, A. Ciocan, P.B. Farnsworth, Optical measurements of ion density in the second vacuum stage of an inductively coupled plasma mass spectrometer, Spectrochim. Acta Part B 53 Ž1998. 569᎐579. w13x B.S. Duersch, P.B. Farnsworth, Characterization of the ion beam inside the skimmer cone of an inductively coupled plasma mass spectrometer by laser-excited atomic and ionic fluorescence, Spectrochim. Acta Part B 54 Ž1999. 545᎐555. w14x B.S. Duersch, J.E. Patterson, P.B. Farnsworth, The effects of a torch shield on performance of the vacuum interface of an inductively coupled plasma mass spectrometer, J. Anal. At. Spectrom. 14 Ž1999. 615᎐619. w15x H. Niu, S. Luan, H. Pang, R.S. Houk, Langmuir probe measurements of the ion extraction process in inductively coupled plasma-mass spectrometry. Part 2. Mea-
w16x
w17x
w18x w19x
1695
surements of floating voltage and radiofrequency voltage, Spectrochim. Acta Part B 50 Ž1995. 1247᎐1261. A.L. Gray, R.S. Houk, J.G. Williams, Langmuir probe potential measurements in the plasma and their correlation with mass spectral characteristics in inductively coupled plasma mass spectrometry, J. Anal. At. Spectrom. 2 Ž1987. 13᎐20. R.S. Houk, J.K. Schoer, J.S. Crain, Plasma potential measurements for inductively coupled plasma mass spectrometry with a centre-tapped load coil, J. Anal. At. Spectrom. 2 Ž1987. 283᎐286. D.J. Douglas, J.B. French, An improved interface for inductively coupled plasma-mass spectrometry ŽICP-MS., Spectrochim. Acta Part B 41 Ž1986. 197᎐204. D.J. Douglas, J.B. French, Gas dynamics of the inductively coupled plasma mass spectrometry interface, J. Anal. At. Spectrom. 3 Ž1988. 743᎐747.