S tomic pectroscopy. Special Issue New Dynamic Reaction Cell Technology. In This Issue: March/April 1999 Volume 20, No. 2

Transcription

1 A S tomic pectroscopy March/April 1999 Volume 20, No. 2 Special Issue New Dynamic Reaction Cell Technology In This Issue: Theory, Design, and Operation of a Dynamic Reaction Cell for ICP-MS Scott D. Tanner and Vladimir I. Baranov...45 The Analysis of High Purity Hydrogen Peroxide by Dynamic Reaction Cell ICP-MS Uwe Völlkopf, Klaus Klemm and Markus Pfluger...53 Analysis of High Purity Acids Using a Dynamic Reaction Cell ICP-MS David S. Bollinger and Anthony J. Schleisman...60 The Benefits of a Dynamic Reaction Cell to Remove Carbonand Chloride-Based Spectral Interferences by ICP-MS Kenneth Neubauer and Uwe Völlkopf...64 Effect of Collisional Damping in the Dynamic Reaction Cell on the Precision of Isotope Ratio Measurements Dmitry R. Bandura and Scott D. Tanner...69 Improving Sensitivity and Detection Limits in ICP-MS With a Novel High-Efficiency Sample Introduction System Ebenezer Debrah and Guy Légère...73 ASPND7 20(2) (1999) ISSN

2 A tomic S pectroscopy is printed in the United States and published six times a year by: The Perkin-Elmer Corporation 761 Main Avenue, Norwalk, CT USA Tel: Fax: Editor Anneliese Lust lustan@perkin-elmer.com Technical Editors Frank F. Fernandez, AAS Susan A. McIntosh, ICP-OES Eric R. Denoyer, ICP-MS SUBSCRIPTION INFORMATION: Atomic Spectroscopy P.O. Box 557 Florham Park, NJ USA Fax: Subscription Rates U.S. $49.00 includes third class mail delivery worldwide. U.S. $75.00 for foreign subscribers wishing airmail delivery. Check must be drawn on a U.S.bank in U.S. funds and made out to: Atomic Spectroscopy Back Issues/Claims Single back issues are available at $10.00 each. Subscriber claims for missing back issues will be honored at no charge within 90 days of issue mailing date. Address Changes to: Atomic Spectroscopy P.O. Box 557 Florham Park, NJ USA Copyright 1999 The Perkin-Elmer Corporation. All rights reserved. Microfilm of Atomic Spectroscopy issues are available from: University Microfilms International 300 N. Zeeb Road Ann Arbor, MI USA Tel: (within the U.S.) or: Dear Spectroscopist, Inductively coupled plasma mass spectrometry (ICP-MS) is entering a new era. Over the past 15 years, ICP-MS has achieved a wide acceptance in both research and routine high-throughput laboratories. It has proven to be a powerful yet rugged and reliable technique capable of withstanding the rigors of the most demanding laboratories. ICP-MS is improving the quality of our lives. As documented in these and other pages, its ability to determine isotopes has contributed to the understanding of sources of toxic blood poisoning in children of Europe and the Americas. Medical researchers have studied fundamental mechanisms and therapeutic treatments of many human illnesses including manic-depressive illness, arthritis, heart disease and cancer. Because of ICP-MS, we are exploring and discovering richer ore bodies easier. We are monitoring our nuclear power stations more precisely. We are cleaning and protecting our environment to improved standards. And we are breaking new technology barriers to developing smaller and faster microelectronic devices which promise better and better global communications. At the very least, these improved communications can help us better know and understand each other, and hopefully coexist more peacefully. ICP-MS has proven to be an invaluable problem-solving tool. Nonetheless, ICP-MS is not without its challenges. As performance has improved over the years, an increased understanding of interferences has identified some of the technological brick walls of the technique. Many approaches have been developed for hurdling these walls, and have in many cases improved capability significantly. However, an exciting technology utilizing a new Dynamic Reaction Cell (DRC) promises to resolve some of the most difficult interferences encountered in ICP-MS. Elements such as Fe, Ca, Se and As can be determined at significantly lower levels because of the ability of the DRC technology to eliminate competing spectral interferences. Though much has already been accomplished with this new DRC technology, there is still much to learn. This special issue of Atomic Spectroscopy is devoted to communicating the most recent applications of the DRC and documenting some of the most current theoretical understanding of its operation. It should assist ICP-MS users in developing an even broader base of new applications, and hopefully will help further extend the beneficial contributions of ICP-MS to our lives. Eric R. Denoyer ICP-MS Technical Editor Perkin-Elmer and PE Pure are registered trademarks and AutoLens, Dynamic Reaction Cell, DRC, and HESIS are trademarks of The Perkin-Elmer Corporation. SCIEX and ELAN are registered trademarks of MDS Inc., a division of MDS Inc. HP is a registered trademark of Hewlitt-Packard Corporation. Teflon is a registered trademark of E.I. dupont denemours & Co., Inc. Meinhard is a registered trademark of J.E. Meinhard Associates, Inc. Milli-Q is a trademark of Millipore Corporation. Registered names and trademarks, etc. used in this publication even without specific indication therof are not to be considered unprotected by law.

3 Theory, Design, and Operation of a Dynamic Reaction Cell for ICP-MS Scott D. Tanner* and Vladimir I. Baranov Perkin-Elmer Sciex Instruments 71 Four Valley Drive, Concord, Ontario, Canada L4K 4V8 THEORY Because an ion is charged, it can interact with the dipole moment of a polar molecule or induce a dipole in a non-polar molecule. Ion-dipole interactions are effective at long inter-atomic distances, and so the collision rate between ions and neutral molecules is high. The iondipole interaction is usually sufficiently strong to exceed the activation energy potential barrier to exothermic reactions, and so thermodynamically allowed ionmolecule reactions are usually fast. Because the activation energy barrier to reaction is usually insignificant, the probability of simple ion-molecule small particle (electron, hydrogen atom, and proton) transfer reactions is binary (it happens or does not) and depends on the reaction exothermicity (e.g., the difference in the ionization potential of the product and reactant neutrals for electron transfer reactions). Therefore, thermodynamically allowed ion-molecule reactions are usually fast and highly specific. For example, the reaction: Ar + + NH 3 NH Ar (1) is fast, having a rate constant of 1.7 X 10 9 cm 3 molecule -1 s 1 (1). Reaction (1) is an example of electron transfer (or charge exchange), and occurs because the ionization potential of NH 3 (10.2 ev) is less than that of Ar ( ev). The reaction is exothermic. On the other hand, the ionization potential of Ca (6.111 ev) is less than that of NH 3, and so the corresponding endothermic reaction: Ca + + NH 3 NH Ca (2) *Corresponding author. is not observed to proceed (the rate constant is less than cm 3 molecule -1 s 1 ). As will be seen later, the difference in the rate constants for the reactions of the isobars Ar + and Ca + with NH 3 allows the ion signal for Ar + at m/z=40 to be suppressed by nine of orders of magnitude while simultaneously the ion signal for Ca + is virtually unaffected. This specificity of reaction allows a dramatic improvement in detection limits when the inductively coupled plasma mass spectrometry (ICP-MS) instrument is configured to take advantage of it. The rf-only multipole has found great use in the study of ion-molecule reactions (2,3). A multipole is usually characterized by its order, which is the number of pairs of poles (a quadrupole is second order, a hexapole is third order, an octapole is fourth order, etc.). The order describes the shape of the effective potential well within the rod array. Hence, a higher order multipole has a wider potential well near the axis and a stronger field gradient closer to the rods. It is wellknown that a higher order multipole is preferred when it is used to study ion-molecule reactions, particularly when it is desired to monitor the product ions of the reaction over a wide range of masses. It will be shown that this latter characteristic, the efficient confinement of ions over a wide range of masses, is the major detriment for the use of a multipole for analytical purposes in ICP-MS. In fact, the preferred configuration is a quadrupole, which defies the conventional wisdom (3). An ion is said to be stable in a multipole if it can be confined and transmitted through the device. Stability is affected by the amplitude and frequency of the applied rf, the mass of the ion, the size of the multipole array, and the dc voltage applied between pole pairs (if any). It is convenient to describe the regions of stability and instability in terms of the Mathieu parameters, a and q: a e V dc n = 2n(n- 1) (1) mω 2 r 2 0 q e V rf n = n(n- 1) (2) mω 2 r 2 0 where n is the order of the multipole, e is the electronic charge, V dc and V rf are, respectively, the dc and the zero-to-peak rf amplitude applied between pole pairs [note that this definition follows that of Dawson (4)], m is the ion mass, ω is the rf angular frequency, and r 0 is the field radius of the multipole array. Because of its symmetry, a quadrupole is characterized by well-defined stability boundaries, as shown in Figure 1, and these are independent of the initial position of the ion. By contrast, the higher order multipole has more diffuse stability boundaries (there are wide regions of partial stability), and these are dependent on the initial position of the ion within the field. This is why the quadrupole has found unique application as a mass filter and, as will be seen, is necessary for optimal performance as an ion-molecule reactor for analytical ICP-MS. Referring to Figure 1, the quadrupole may be operated as an rf-only device along the a=0 axis. Ions having a value of q<0.908 have stable trajectories. Lower mass ions, having q>0.908, are unstable and are rapidly lost from the quadrupole. Therefore, the rf-only quadrupole is a high pass filter characterized by a well-defined low mass cut-off. The addition of a dc voltage between the pole pairs moves the 45 MS-114 D-6061

4 Fig. 1. The stability diagram (first region) for a linear quadrupole under collision-less conditions. The region bounded by the β y =0 and β x =1 curves provides stable ion motion in both x and y. operating point away from the a=0 axis, and introduces both a low mass (at high q) and high mass (at low q) stability cut-off. Operation at the apex of the stability region (near q=0.706, a=0.24) provides a narrow bandpass, and is the usual operating point for a quadrupole mass filter. The quadrupole can be operated as an ion-molecule reactor by enclosing it so that it can be pressurized with a reactive gas. Efficient elimination of plasma-based interference ions may be obtained at pressures in the vicinity of mtorr. Under these conditions, the input ion beam is thermalized through collisions, resulting in a near-thermal distribution of ion energies and the ions are focused towards the axis of the quadrupole (5,6). We have discussed the processes which lead to this state, and the implications that these have for analytical use in ICP-MS (7). We have shown that the contribution of the rf-field energy to the ion kinetic energy (and hence the reaction energy) increases with an increase in the amplitude of the rf, a decrease in the rf frequency, a decrease in the number of collisions per rf cycle, and particularly near the high-q stability boundary. For the ICP-MS application, it is desirable to resolve isobaric interferences on the basis of their ionmolecule reactivity. Operation of the reaction cell under non-thermal conditions (i.e., with a significant contribution of the rf-field energy to the reaction energy) can promote otherwise endothermic reaction processes. Hence, in order to retain the specificity of the chemistry it is desirable to minimize the contribution of the rf-field energy to the reaction energy. This can be assured, in large measure, by operation at low rf amplitude. Conventional operation of the quadrupole employs a fixed rf frequency, while the rf amplitude is increased with the ion mass. It is clear, from equations (1) and (2), that the device may also be frequencyscanned at fixed rf amplitude, which has the advantage that the rf amplitude can be maintained at a sufficiently low level to minimize the promotion of endothermic processes while retaining the efficient confinement of the ions of interest. For analytical purposes, for which it is desired to eliminate an isobaric interference in order to allow determination of an analyte at ultra-trace levels, it is evident that the efficiency of the reaction cell must be enormous. To enable the determination of Ca + at m/z=40 at subppt levels, the ion signal for 40 Ar + must be suppressed by about 9 orders of magnitude (from cps to 10 cps). Because of the distribution of the number of collisions that any ion may suffer, 9 orders of magnitude of reactive decay requires that, on average, each ion suffers approximately 20 reactive collisions (which may require many more collisions, depending on the reaction efficiency). This means that successive, sequential reaction chemistry is promoted. Through a series of reactions, new isobaric interferences may be produced at many masses. Consider, for example, the use of methane as the reactive gas. The dominant plasma ion, Ar +, is known to react according to: Ar + + CH 4 CH 2+ + Ar + H CH 3+ + Ar + H 2 CH 4+ + Ar Subsequently, these intermediate products react according to: CH 3+ + CH 4 C 2 H H 2 C 2 H H 2 and other similar reactions. Now, C 2 H 3+ and C 2 H 5+ are good proton donors. If, for example, there is a trace of acetone in the reaction cell, it may be protonated according to: C 2 H C 3 H 6 O C 3 H 7 O+ +C 2 H 2 C 2 H 5+ + C 3 H 6 O C 3 H 7 O+ +C 2 H 4 which produces an isobaric interference for Co + at m/z=59. More generally, if a hydrocarbon neutral is in the reaction cell, complex chemistry is promoted: 46

5 C 2 H 5+ + C n H x C n H x C 2 H 4 C n H x C 2 H 4 + H 2 C n H x C 2 H 4 +2H 2 which are examples of proton transfer and dissociative proton transfer reactions. Of course, all of the reactions shown here are simply representative of the types of reactions that may take place within the reaction cell when a relatively complex reaction gas is used. The reactions take place nearly simultaneously and continuously as long as the intermediate ions and the neutrals are within the cell. Because of the efficiency ( %) of the reaction cell, even a very trace amount of hydrocarbon will enable the production of a wide range of interferences. It is simply not possible to maintain the contamination level of the reaction cell low enough to completely suppress this chemistry. The special ion stability characteristics of a quadrupole reaction cell may be used to efficiently intercept this sequential chemistry and eliminate the interferences produced within the reaction cell. The representative chemistry discussed above is presented in Figure 2a in roughly chronological order. Alternatively, this may be reorganized in order of the mass of the intermediate product ions, as shown in Figure 2b. As long as the neutral species are present in the cell and all of the intermediate product ions are retained as well (i.e., by operation of the cell in the rf-only mode at low q), this chemistry proceeds nearly simultaneously and continuously. The efficiency of the reaction cell cannot be realized because of the production of new interferences at many masses. However, the quadrupole reaction cell is characterized by well-defined stability boundaries. Selection of the parameters a and q allow the definition of a mass bandpass; ions having m/z outside of the stability boundary are efficiently and rapidly ejected from the cell. Therefore, the sequential chemistry which leads to new interferences can be efficiently interrupted. Further, the bandpass can be swept in concert with the mass of the ion being determined by the downstream mass analyzer. Therefore, a dynamic bandpass can be defined for the reaction cell which allows efficient transfer of the analyte ion being measured downstream; this also allows for efficient reaction of the plasma-based isobaric interference ion and simultaneously suppresses the formation of new interferences at the analytical mass within the reaction cell. This is shown schematically in Figures 2c through 2f. In this case, a bandpass with both a high and low mass cutoff is shown, and is progressively shifted with the analytical mass (in the sequence of Figures 2c through 2f). The neutral reactant species are, of course, always present (they are unaffected by the operating point of the reation cell). However, only those reactions for which the reactant ion is simultaneously confined within the cell may proceed. These reactions are shown by solid arrows; the dashed arrows indicate reactions that are suppressed because the reactant (intermediate product) ion is unstable under the operating conditions. It is therefore evident that the reactions which remove an interference ion from the bandpass are allowed to proceed (with high efficiency), but those reactions that would otherwise produce an interference ion within the bandpass are suppressed. We therefore have the interesting situation where the bandpass is operated in order to suppress the formation of new interferences within the bandpass by destabilizing the intermediate product ions outside of the bandpass. At the same time, operation in this bandpass mode does not affect the efficiency of removal Vol. 20(2), March/April 1999 of plasma-based ions, and the unreacted analyte ion is transmitted efficiently without interference. The low mass boundary of the bandpass is determined by the selection of the parameter q at the analytical mass. Recall that this parameter specifies the frequency of the rf at the operating rf amplitude, and that therefore lower mass ions are simultaneously confined within the cell according to the value of q at their masses. The lower mass edge of the bandpass is defined by the mass below which the value of q is greater than the stability boundary limit (i.e., q>0.908 for a perfect quadrupole without collisions). In some instances it may also be desired to invoke a high mass cut-off for the bandpass, and this is achieved by specifying a non-zero value for a. Because the bandpass width is adjusted by changing the frequency of a low amplitude rf and a low dc voltage, these may be adjusted rapidly and on a per-element basis. Thus, the Dynamic Reaction Cell (DRC) is a low constant rf amplitude quadrupole which is characterized by a dynamic mass bandpass, which is dynamically moved in concert with the analyte mass and whose bandpass width may be adjusted dynamically. It is these dynamic characteristics that distinguish the Dynamic Reaction Cell from generic rf-only multipole reaction cells. Since the DRC efficiently suppresses interferences that would otherwise be produced through a complex sequence of reactions, the analyst has a wider selection of reaction gases at his disposal. Further, the cell may be operated at near thermal equilibrium conditions, for which the efficiency is maximized, since kinetic energy discrimination (the alternative for product ion discrimination with higher order multipoles) is not required. 47

6 Fig. 2. Sequential, secondary ion chemistry that has the potential to produce chemical interference within a reaction cell. (a) Chronology of a representative sequential chemistry. (b) The chemistry reorganized as a function of the mass of the intermediate products of reaction. As long as the reactant neutral is within the cell and the intermediate ions are stable (as in a conventional reaction cell), this chemistry will proceed nearly simultaneously and continuously. (c) through (f) A bandpass is introduced, defined by the stability boundaries of a quadrupole, which intercepts the chemistry. Only those reactions for which the reactant ion is within the stability bandpass may proceed; other reactions are suppressed. The bandpass is swept (in c through f) in concert with the analyzed mass, and the bandpass width may be adjusted on the fly as required for chemical resolution. DESIGN The Dynamic Reaction Cell (DRC) is an enclosed quadrupole which acts as the interface between the single lens ion optics chamber and the mass analyzer high vacuum chamber. Reaction gas is introduced through a clean gas manifold and controlled by one or both of two low flow (3 sccm) mass flow controllers. The ion beam is introduced to the DRC through an entrance aperture inlet. Ions exit the cell through an aperture which communicates with the AC-only prefilter, and hence are transferred to the quadrupole mass filter. Reaction gas exits the cell through both the entrance and exit apertures. For experiments without reaction gas, the gas flow is stopped and the DRC enclosure is remotely opened for venting into the high vacuum chamber. A schematic of the instrument is given in Figure 3. Fig. 3. Schematic of the DRC-ICP-MS. The DRC is a rf/dc quadrupole which may be enclosed and pressurized with a reactive gas. The DRC may be vented into the high vacuum chamber in order to emulate conventional ICP-MS. 48

7 Vol. 20(2), March/April 1999 A separate power supply provides the rf and dc voltages to the DRC. The rf drive frequency is determined by the mass of the ion currently being transmitted through the downstream mass analyzer and the value of q specified in the method for the DRC for this mass. This frequency is digitally synthesized and amplified to the specified rf amplitude. The dc voltage between pole pairs specified by the value of a is offset by the reaction cell rod offset and applied on top of the rf. Therefore, the DRC is frequency-scanned with the analyzer mass, and the frequency is chosen to provide the operating parameter, q, defined on a per-mass basis. For given values of a and q, the dc voltage is independent of ion mass. Accordingly, the mass bandpass window of the DRC is dynamically adjusted with the reference q and a defined for the mass being transmitted through the mass filter. The bandpass is adjusted on a perelement basis; that is, the value of q and of a is defined for each element in the peak hopping mode and the bandpass is adjusted prior to measurement of each signal. For spectral scanning, a constant value of q and of a is applied through a scan segment. In addition, the selection of reaction gas (with the twochannel option) and reaction cell pressure may be adjusted for a given spectral scan segment or on a perelement basis through specification of a reaction gas flow in the method. For the determination of elements that are not normally interfered, it may be desirable to emulate conventional ICP-MS operation with the DRC-ICP-MS. If reaction gas is not added to the DRC, plasma gas from the ion optics chamber is entrained into the reaction cell. If the reaction cell remains enclosed, the plasma gas pressurizes the cell and facilitates ion-molecule reactions which increase the spectral background. Suppression of this chemistry can be achieved by increasing the mean-free path within the reaction cell. This is most readily achieved by increasing its conductance into the high vacuum (mass analyzer) chamber. The DRC is equipped with an externally activated venting mechanism which opens the reaction cell. With the vent open and reaction gas flow stopped, the instrument is said to be operating in standard mode. Mode switching is performed automatically. OPERATION The DRC parameters are optimized independently of the remainder of the ICP-MS system. Typically, optimization of the plasma conditions and ion optic (AutoLens ) are performed as for a standard ELAN instrument. Since the DRC in the enhanced mode (with reaction gas) eliminates plasma-based interferences, the plasma is operated in the normal mode (high power and optimum nebulizer flow for robust analytical conditions. We have yet to find an analytical challenge that requires the use of cold plasma, although operation in this mode is also available. Insight into an appropriate reaction gas for use in the enhanced mode can be obtained by checking the database of ion-molecule reaction rate constants (1). A preferred reaction gas is one for which the rate constant for reaction of the analyte ion is considerably smaller than that for the isobaric interference ion. In this instance, the analyte ion is to be measured at its atomic m/z. Alternatively, a suitable gas is one for which the analyte ion produces a product ion having a different m/z than the isobaric interference produces (and it is additionally desirable that the mass ratio of the product ion/analyte ion is small, preferably less than two, and that the product ion does not interfere with another analyte ion of interest and is not interfered by another ion). It is further preferred that the reaction gas have a molecular weight that is less than the atomic mass of the analyte ion (to suppress scattering losses). Other considerations in the selection of the gas, such as its toxicity or reactivity with the gas manifold or vacuum system components, will be evident. Once a candidate reaction gas is selected, the appropriate bandpass for the DRC should be determined. Most often, this will be determined empirically by scanning the ion signal as a function of the parameter q for a blank solution and for a solution containing the analyte. An example of this type of optimization is given in Figure 4 for Fe + with NH 3 as the reaction gas. The analyte ion signal has a broad maximum in the vicinity of q=0.45. The blank signal shows a maximum near q=0.15. The signal in the blank is a m/z=56 ion (not Fe + ) which is formed in a series of reactions that include at least one intermediate product ion of lower mass. Operation at q>0.3 clearly suppresses the chemistry which creates this interference. Therefore, in this instance, it is convenient to optimize q for the maximum analyte ion signal (q ~ 0.45). In some instances, it may be necessary to add a high mass cut-off in order to suppress chemistry involving higher mass intermediate product ions. It might be noted that a lower than otherwise optimum value of q is required if a reaction chemistry is chosen which converts the analyte to a higher m/z polyatomic ion which is to be measured as indicative of the analyte. This is because the product ion will only be observed if both the analyte ion and its product ion are simultaneously stable. Since the product ion will be higher mass than the analyte ion (presumably it is a polyatomic of the analyte), the operating point q for the product ion must be sufficiently small as to provide a q-value 49

8 for the lower mass analyte (reactant) ion which remains within the stability bandpass. It might be further noted that the converse of this is also true: operation at a sufficiently high q that the product of a reaction of the analyte ion is not stable, allows suppression of, for example, cluster ions which could interfere with the determination of other analyte ions at higher m/z. Having chosen the reaction gas and the operating point for the DRC, it remains to optimize the pressure of the reaction cell. Generally, a higher pressure is required for slow reactions or for intense isobaric interferences (more collisions are required in either event). It is convenient to monitor the ion signal for the analyte as a function of the reaction gas for the same two solutions (blank and standard) as above. Typical results for an optimization of this type are given in Figure 5, again for Fe + with NH 3 as the reaction gas. The difference between the signal levels for these solutions provides the net sensitivity to the analyte as a function of flow (pressure) of the reaction gas. Figure 6 shows this result for data taken from Figure 5. The net Fe + signal initially increases due to collisional focusing, attains a maximum, and then decreases. The loss of ion signal at high flow may be due either to reaction of the analyte ion with the reaction gas or scattering by the gas. It is important to recognize, however, that the optimum reaction gas flow does not necessarily correspond to the flow which gives the maximum net sensitivity. This is because the detection limit is determined both by the sensitivity and by the noise on the background signal. It can be shown that, for ion signals below ca cps with the DRC pressurized, the standard deviation of the signal is well-approximated by counting statistics. Therefore, the detection limit can be estimated according to: EDL = 3 Β S where EDL is the estimated detection limit in ppt, B is the background signal (in cps, from Figure 5) and S is the net sensitivity in cps per ppt (from Figure 6). This estimated detection limit (3 σ for 1 second measurement period) can thus be determined as a function of the reaction gas flow, as shown in Figure 7 for Fe +. The estimated detection limit typically passes through a minimum (at ca. 1.2 ml/min for Fe + ). Measurements of the detection limit at this flow (3 σ, 1 second dwell, 10 replicates) are found to agree well with the estimated detection limit. Fig. 5. Optimization of the reaction gas (NH 3 ) flow for Fe + as the analyte. The solid line is the ion signal for a blank (Distilled Deionized Water) as a function of NH 3 flow; the dashed line is for a standard containing 100 ppt Fe. Fig. 4. Ion signals at m/z=56 for a blank (dashed line) and a standard containing 100 ppt Fe (solid line) as a function of the parameter q. The ion signal for the blank has been multiplied by a factor of 10 relative to the signal for the standard. The rf amplitude was fixed at 200 V peak-to-peak, and the frequency was scanned to adjust q. The peak near q=0.15 for the blank solution corresponds to an ion of m/z=56 which is formed through a series of reactions involving at least one significantly lower mass intermediate product ion. This interference ion is suppressed at q>0.3. Fig. 6. Net sensitivity for Fe as a function of reaction gas (NH 3 ) flow. The data are derived from the difference of the curves given in Figure 5. 50

9 Vol. 20(2), March/April 1999 An example of the specificity and efficiency of the DRC is given in the reaction profiles for Ar + /Ca + in Figure 8. The ion signal corresponding to 40 Ar + in the blank solution in the standard mode is ca cps. At the lowest flow of NH 3 shown in Figure 8, the 40 Ar + signal is already attenuated by an order of magnitude, and thereafter decreases semilogarithmic linearly to less than 100 cps (for the blank solution) at a flow of 1.4 ml/min. The curvature above this flow rate is due to the contribution of contaminant Ca + in the blank solution (corresponding to ca. 3 ppt). The corresponding reaction profile for 100 ppt Ca shows a plateau which decreases only slowly, indicating that Ca + is unreactive with NH 3 (the slow decay is due to scattering at high cell pressure). Allowing for the signal contributed by the contaminant Ca in the blank solution, Fig. 7. Estimated detection limit (3σ, 1 second dwell) for Fe as a function of reaction gas (NH 3 ) flow. The data are obtained by taking three times the square root of the blank data of Figure 5 and dividing this by the net sensitivity of Figure 6. The optimum flow corresponds to the minimum estimated detection limit. these data show some 9 orders of magnitude suppression for Ar + and only marginal scattering loss for Ca +. Clearly, the ion-molecule chemistry is highly specific (ca. 9 orders of magnitude discrimination) and efficient (again, 9 orders of magnitude, or %). Data emulating conventional ICP-MS are shown in Figure 9, where the spectra for a blank (DDIW) and 100 ppt mixed element standard are overlaid. These spectra were obtained in the standard mode, for which the reaction cell is evacuated to the mass analyzer chamber. Performance in this mode is at least equivalent to current generation ICP-MS systems. Simply because of the geometry of the device and the design of the optics, the continuum background signal is reduced to below 1 cps. As a result, elements that are not interfered by plasma ions may be determined at very low levels (sub 0.1 ppt). As an example, Figure 10 shows the mass spectra for 1 ppt U and the corresponding DDIW blank. The signal/noise for 1 ppt is of the order of 80, and the detection limit is below 0.03 ppt (for a 3-second measurement time in peak hopping mode). For elements which are interfered by polyatomic and argide ions, the Dynamic Reaction Cell offers an enormous improvement in detection limits. Figure 11 shows the spectra for the same solutions as Figure 9, but with the reaction cell optimized for Ca determination (i.e., high NH 3 reaction gas flow, which is greater than optimum for most elements). The chemical background signals are reduced to contamination levels. Many of the elements are unreactive (e.g., K +, Ca +, Cr +, and Mn + ) or slowly reactive (e.g., Fe + and Al + ) with NH 3, and these may be determined at exceedingly low levels because of Fig. 8. Reaction profiles for m/z=40 as a function of reaction gas (NH 3 ) flow. The solid line is the ion signal for a blank (Distilled Deionized Water), and the dashed line is for a standard containing 100 ppt Ca. Fig. 9. Spectra for blank (Distilled Deionized Water, lighter shaded) and a standard (darker shaded) containing 100 ppt (ng/l) each of Na, Mg, Al, K, Ca, Cr, Mn, Fe, Ni, Cu, Zn, and As. The data were obtained in the standard mode for which the reaction cell enclosure is opened and the cell is evacuated into the mass analyzer region. 51

10 Fig. 10. Spectra for a blank (solid line, Distilled Deionized Water) and a standard (dashed line) containing 1 ppt (ng/l) of U. The data were obtained in the standard mode. The blank signal is ca. 0.5 cps, and the signal/noise for 1 ppt U is ca. 80. Fig. 11. Spectra obtained for the same solutions as those of Figure 9, with the reaction cell in enhanced mode. For these data, the reaction cell was closed and pressurized with NH 3 at a flow of 1.5 ml/min; the resulting cell pressure is approximately optimum for the determination of Ca, but is somewhat higher than optimum for most other elements. the reduced interferences and the low continuum background signal. Of course, not all elements are readily determined using NH 3 as the reaction gas. A case in point is As +, which suffers interference from ArCl + (in solutions containing chloride). While ArCl + is highly reactive with NH 3, it is found that As + is as well [see Baranov and Tanner (7)]. In this instance, an alternate reaction gas, such as CH 4 or H 2, is prescribed, or As may be determined as the product of a condensation reaction [see Bollinger and Schleisman (8)]. Nonetheless, the impressive capability of the reaction cell is shown in the spectra of Figure 12 for which the signal for 1 ppt Fe is readily observed above the blank for DDIW. The continuum background signal remains below 1 cps, and permits the determination of the isotope ratios for 54 Fe + / 56 Fe + / 57 Fe + even at the level of 1 ppt. REFERENCES 1. V.G. Anicich, Ap. J. Supplement Series 84, 215 (1993). See also 2. K.M. Ervin and P.B. Armentrout, J. Chem. Phys. 83, 166 (1985). 3. D. Gerlich, in State-Selected and State-to-State Ion-Molecule Reaction Dynamics. Part 1. Experiment, C.-Y. Ng, M. Bear, Eds., Adv. Chem. Phys. 82, 1 (1992).. 4. P.H. Dawson (editor), Quadrupole Mass Spectrometry and its Applications, Elsevier, New York (1976). 5. D.J. Douglas, and J.B. French, J. Am. Soc. Mass. Spectrom. 3, 398 (1992). 6. D.J. Douglas, J. Am. Soc. Mass. Spectrom. 9, 101 (1998). 7. V.I. Baranov and S.D. Tanner, J. Anal. At. Spectrom. (1998) (submitted). 8. D.S. Bollinger and A.J. Schleisman, in Plasma Source Mass Spectrometry: Developments and Applications II, G. Holland and S.D. Tanner, (eds), The Royal Society of Chemistry, Cambridge 1999 (in press). Fig. 12. Spectra for blank (solid line, Distilled Deionized Water) and a standard (dashed line) containing 1 ppt (ng/l) of Fe, obtained in the enhanced mode. The reaction cell was pressurized with NH 3 at a flow rate of 1.2 ml/min. 52

11 The Analysis of High Purity Hydrogen Peroxide by Dynamic Reaction Cell ICP-MS Uwe Völlkopf* Perkin-Elmer Sciex Instruments, 71 Four Valley Drive, Concord, Ontario, L4K 4V8, Canada and Klaus Klemm and Markus Pfluger Merck KGaA, Frankfurter Strasse 250, D Darmstadt, Germany INTRODUCTION Outstanding figures of merit, especially high detection power and speed of multielemental analysis, have led to the very rapid growth of inductively coupled plasma mass spectrometry (ICP-MS) over the past decade. However, there is still room for improvement. Unfortunately, a number of polyatomic species can overlap with analyte masses of interest and reduce the applicability of ICP-MS to solve certain analytical problems. Different approaches have been suggested to reduce or overcome these spectral interferences. Among them are offand on-line matrix separation, mixed gas plasmas, electrothermal vaporization (ETV), laser sampling and magnetic sector high resolution ICP-MS (HR-ICP-MS). All these approaches have their advantages and disadvantages. Some of the common spectral interferences, e.g., 40 Ar on 40 Ca, 40 Ar and 38 Ar 1 H on 39 K, 40 Ar 12 C on 52 Cr, 38 Ar 16 O 1 H on 55 Mn, and 40 Ar 16 O on 56 Fe can be reduced or even eliminated applying reduced temperature (cool or cold) plasma techniques (1,2). Unfortunately, plasma stability suffers under these lower temperature plasma conditions if there are large amounts of dissolved solids or strong acids present in the sample. Most cool or cold plasma analyses therefore require the use of the method of standard additions to overcome matrix-induced signal suppression. This significantly reduces sample throughput and increases the risk of sample contamination. Further, an increased baseline noise has been reported *Corresponding author. ABSTRACT Over the past 10 years, many techniques have been developed to overcome spectral interference problems in ICP-MS. Most of them are only applicable to single elements or small groups of elements. Others, such as sector field high resolution mass spectrometry, are very expensive and suffer limitations with respect to detection power and speed of analysis when operated in high resolution mode. Perkin-Elmer SCIEX developed a new approach for eliminating spectral interferences, called chemical resolution, which is applicable to a conventional quadrupole mass spectrometer. This unique approach has been commercialized for the first time with the ELAN 6100 DRC. The concept of a bandpassed dynamic reaction cell permits extremely efficient removal of many argon-based interferences (up to 9 orders of magnitude) and has the potential of eliminating other polyatomic interferences as well. Instrument background is dramatically reduced to less than 1 count/second, while normal quadrupole ICP-MS sensitivity and scanning speed is maintained. This leads to detection limits that are typically an order of magnitude better than a conventional quadrupole ICP-MS, including the problematic elements like K, Ca, Cr, Fe, As, and Se. All analytes can be determined applying a single set of rugged, normal temperature plasma conditions, leading to improved long-term stability, more efficient sample throughput, and enhanced ease of use. for the direct analysis of neat H 2 O 2. Also, if many analytes must be determined, each sample must be analyzed twice, first with a reduced temperature plasma for analytes with a sufficiently low first ionization potential and a second time with a normal plasma for all other analytes. Due to this and the plasma stabilization times required when the plasma power is altered, the sample throughput is further reduced. The ELAN 6100 DRC (Perkin- Elmer SCIEX, Concord, Ontario, Canada) used for generation of all data presented in this work uses unique Dynamic Reaction Cell (DRC) (3 9) technology for eliminating spectral interferences based on different chemical reactions between analyte and interfering ions. The spectral background across the mass range is extremely low and detection limits of as low as 10 pg/l can be achieved in very clean laboratories. DRC-ICP-MS is of great interest to chemists performing elemental analysis of microelectronic materials. It combines extreme detection power, freedom from interferences for many important semiconductorrelated analytes, and analytical speed. All analytes can be determined using normal, high-temperature plasma conditions, so there is no need to determine elements such as K, Ca, and Fe using cool or cold plasma conditions. This leads to reduced matrix suppression effects and often allows direct determination of the analytes in reagents such as concentrated hydrogen peroxide (H 2 O 2 ) using aqueous calibration curves. Currently, Semiconductor Equip- 53 MS-115 D-6062

12 ment and Materials International (SEMI) discusses the specification of so-called Tier D guideline reagents levels where approximately 18 important analytes are specified to be present at less than 10 ng/l in these solutions (10). Analysis of concentrated reagents at these extreme ultratrace analyte concentration levels is a challenge for many laboratories. A number of them apply preconcentration techniques and reduce the reagent matrix almost to dryness. The residue is then re-dissolved in a weak HNO 3. While this classical sample preparation procedure is useful, it introduces the risk of contamination. DRC-ICP-MS is a technology that reduces or even eliminates many of the significant polyatomic interferences. This results in improved detection power for many analytes which, up to now, could only be determined using lower temperature plasmas or magnetic sector high resolution ICP-MS. EXPERIMENTAL Instrumentation The ICP-MS used for the work presented in this paper was an ELAN 6100 DRC. The instrument is controlled by new ELAN 6100 software, which permits full control of all dynamic reaction cell parameters such as gas type, gas flow, and power settings. The software also allows for the automatic grouping of analytes, which means that the instrument can be sequentially operated in standard ICP-MS mode (open cell, no reaction gas) and DRC mode (with reaction gas) in a fully automated manner. The ELAN 6100 DRC used to develop the method for H 2 O 2 analysis was installed in a class 1000 clean room. The sample introduction system of the instrument was protected further by a class 100 portable laminar flow bench. High purity (18.2 Meg-ohm) deionized water was produced using a Millipore Milli-Q water purification system. Only high purity acids were used for sample and wash solution preparation. Pre-cleaned PFA sample containers (Seastar) were used. All microliter pipette tips were leached in dilute acid solutions for several weeks. Most dilutions were carried out by weight using microbalances. Calibration standards were prepared from either single or multielement Merck stock standard solutions. All peristaltic pump tubes (tubing used to introduce sample solution) were pre-cleaned by soaking for several weeks in dilute nitric acid. Sample Introduction System A new all-quartz sample introduction system without O-rings, which are a potential source of contamination, was used for all experiments. This was developed for the ELAN 6100 DRC, based on experience in the analytical laboratories of Merck in Darmstadt, Germany. It included a quartz Meinhard -type nebulizer and a quartz cyclonic spray chamber, which was connected directly to a long quartz sample injector with a ball joint connection. A Teflon nebulizer fitting (AF Analysentechnik, Tübingen, Germany) was used to attach the nebulizer to the spray chamber. The sample introduction system was connected to a standard ELAN demountable torch. It was found that the all-quartz sample introduction system, free of O-rings, significantly reduced the base spectrometer contamination level. This high degree of cleanliness is a prerequisite for achieving detection limits below 1 ng/l for many analytes. Figure 1 shows Pb detection limits measured every eight minutes over almost three hours. The detection limit run was started immediately after calibration with a 1-µg/L standard. As a result, the detection limits improved over time as the system contamination was washed out. Finally, reproducible detection limits of less than 0.1 ng/l were achieved. Sporadic outliers, frequently observed with determinations at low ng/l levels with a conventional glass spray chambers and O-ring nebulizer fittings, were not observed. General Instrument Optimization Optimization of an ELAN 6100 DRC does not differ very much from that of a classical ELAN ICP- MS. The nebulizer gas flow and ion Fig. 1. Pb detection limits measured over 168 minutes using a quartz cyclonic spray chamber without O-ring fittings for nebulizer sealing. 54

13 Vol. 20(2), March/April 1999 lens parameters were optimized daily using the auto-optimization routines built into the software. Mass calibration and resolution setup on ELAN systems is typically checked only once a month. DRC Optimization Optimization of the dynamic reaction cell is fairly straightforward and, once a method has been developed, requires very little or no reoptimization on a daily basis. The reaction gas in DRC-ICP-MS should be selected such that the efficiency of the purifying reactions (ion chemistry) is high for the interfering species and as low as possible for the analyte(s) of interest. It was found that NH 3 is very effective for removing most argide interferences. However, CH 4 was found to be most effective for removing the Ar + 2 dimer for the determination of Se. Once optimum conditions for a certain application are established, no daily re-optimization is required. The DRC parameters are simply stored as part of the method optimization files. For method development, optimization of the reaction gas flow (cell pressure) is performed in two steps. First, a high-purity blank solution is aspirated and the gas flow ramped in small steps typically from 0 to 1 ml/min. The same experiment is then repeated aspirating a standard solution containing the analyte(s) of interest at a 1-µg/L (or less) concentration level. A software routine then overlays the two plots of intensity vs. cell gas flow and calculates, for each analyte selected, the estimated detection limit at each flow setting. These plots allow the user easy selection of the optimum gas flow(s). It was found that the optimum reaction gas flow varies very little between the different analytes. Because of this, all analyses discussed in this paper have been performed with the same gas flow for all analytes in the same method, even though it is possible to determine every single analyte with optimum flow rates. Fixed settings do not affect detection power to a large degree and permit faster multielemental analysis. Sample Preparation Calibration solutions were prepared from PE Pure single- and multielement solutions. Intermediate stock standard solutions of 10 µg/l were prepared with 0.1% HNO 3. From these intermediate solutions, final calibration and spike solutions were prepared in precleaned PFA bottles (weight/ weight). Two different batches of H 2 O 2 were used for method development and analysis. All method development work and the longterm stability tests were performed using 30% H 2 O 2, while quantitative analysis and spike recoveries were performed in a higher purity 31% H 2 O 2. RESULTS AND DISCUSSION Sensitivity and Instrument Noise Level The instrument sensitivity in both the standard and DRC modes is similar to that of a conventional ELAN 6100 ICP-MS. The noise level is uniformly low (typical <1 c/s) across the entire elemental mass spectrum and there is no significant difference, whether the instrument Fig. 2. Spectrum of 1 µg/l Tl, Pb, Bi and instrument noise level measured at mass 220 (standard ELAN mode). is operated in standard ELAN mode (reaction cell vented) or in DRC mode (reaction cell pressurized). A typical example of the signal-tonoise is shown in Figure 2 with the spectrum of 1 µg/l Tl, Pb, Bi (and the instrument background measured at mass 220). As can be seen, the noise level is in the order of 0.5 counts/sec with a standard deviation in the order of 0.3 counts/sec. Interferences Studied Spectral interferences important in the analysis of concentrated H 2 O 2 are listed in Table I. NH 3 is the reaction gas of choice for the elimination of all interferences. It was found that the same reaction TABLE I Polyatomic Interferences Studied by DRC-ICP-MS and Important in the Analysis of H 2 O 2 Polyatomic Produced Interfered Species from Analyte 12 C 15 N, C impurities 12 C 14 NH in plasma gas 27 Al 38 Ar 1 H plasma gas 39 K 40 Ar plasma gas 40 Ca 36 Ar 16 O plasma gas 52 Cr 40 Ar 12 C plasma gas 52 Cr 38 Ar 16 O 1 H plasma gas 55 Mn 40 Ar 16 O plasma gas 56 Fe 40 Ar 16 O 1 H plasma gas 57 Fe 55

14 gas flow could be used for the interference-free determination of Al, K, Ca, V, Cr, Mn, and Fe. The effect of using this single flow on detection power was less than 20% for all elements. The advantage of performing the entire DRC scan under one set of operating conditions is the enhancement of analytical speed. Detection Limits The overall high sensitivity of the ELAN 6100 DRC, the ultra-low instrumental background, and the very good short- and long-term precision lead to extremely low detection limits. The improvement in detection power is very significant, especially for analytes that suffer from Ar-related polyatomic overlap interferences in conventional quadrupole ICP-MS. Table II summarizes typical detection limits for 22 elements. The detection limits shown are not necessarily the best that have been obtained. They are the compiled mean values achieved from different instruments in a Perkin-Elmer SCIEX clean room laboratory over a period of several months and from a field test instrument installed in a Merck clean room laboratory in Darmstadt, Germany. TABLE II Detection Limits Achieved in a Class 100 Environment Using Microelectronic Grade DI Water Analyte DL Analyte DL (ng/l) (ng/l) 7 Li Fe a Be Co B 1 60 Ni Na Zn 1 24 Mg Se a 5 27 Al Rh K a In Ca a Sb V a Cs Cr a Pb Mn a U 0.01 DL = Detection Limit. a Indicates elements determined in DRC mode. In comparison, it is also possible to achieve low detection limits for analytes such as K, Ca, and Fe applying cool or cold plasma conditions. However, performing all analyses under only one set of operating conditions, with a more rugged high temperature plasma, is a significant advantage when analyzing samples containing a matrix or concentrated acids. Many of the cool or cold plasma methods require the method of standard additions to compensate for matrix-induced signal suppression. In fact, we have found that in the analysis of concentrated H 2 O 2 using the DRC, direct calibration is feasible. Another benefit of using normal operating conditions for all elements is that there are no stabilization delays associated with changing plasma conditions in the middle of an analytical run. This leads to shorter analysis times and ultimately higher sample throughput. Analysis of Hydrogen Peroxide Short- and Long-Term Stability The purpose of the stability tests was to determine if the ELAN 6100 DRC retains the good overall longterm stability seen with aqueous samples when continuously aspirating concentrated H 2 O 2 over a period of 20 hours. Calibration was performed with a blank and three acidified calibration solutions (100 ng/l, 500 ng/l, and 1000 ng/l). Table III summarizes the results of a typical multielement calibration. Unspiked 30% H 2 O 2 was continuously aspirated for more than 20 hours. Quantitative measurements were performed every two minutes. Results of one of many multielement long-term stability tests carried out on this system are shown in Figures 3, 4, and 5 for K, Ca, and Fe in DRC mode and in Figure 6 for Na in standard mode. Each point in the figures represents the mean value of a sample analysis (10 replicates, 1-second integration time). In addition to the concentration data, the RSD values of each sample run also are plotted. Very good long-term stability was achieved for all analytes determined. The RSD values are very low considering the low analyte concentrations. These results prove that the addi- TABLE III Results of a Multielement Calibration Using a Blank, 100 ng/l, 500 ng/l, and 1000 ng/l Calibration Solutions Analyte Mass Correlation Analyte Mass Correlation Coefficient Coefficient Li As Be Rb B Sr Na Ag Mg Cd Al In V Cs Cr Ba Mn Tl Co Pb Ni Bi Cu U Zn

15 Vol. 20(2), March/April 1999 tion of the dynamic reaction cell into the ion path of the ELAN mass spectrometer does not have any negative influence on long-term instrument stability at ppt levels. On the contrary, because all of the analytes in concentrated chemical reagents can be determined with just one set of normal plasma conditions, short- and long-term stability is improved compared to using cool or cold plasma conditions Quantitative Analysis and Recoveries High purity 31% H 2 O 2 samples (125 ml) were spiked with 100 µl concentrated semiconductor-grade HNO 3 prior to analysis. It was found that the addition of this very small amount of HNO 3 improved the recoveries for some of the analytes studied. The ELAN was calibrated using an acidified DI water blank and two calibration standards (100 ng/l, 500 ng/l). All analytes were determined applying just one set of normal plasma conditions (see Table I). The proposed SEMI Tier Guidelines (10 ng/l) require successful spike recoveries between 75% and 125%. For the study presented in this paper, we spiked the samples with ng/l multielement solutions (w/w). Table IV summarizes results of the determination of 56 Fe in one of the 31% H 2 O 2 samples. This particular sample was analyzed five times over two days. The table lists the mean values of all five individual runs, the results of the spike recoveries, and the measurement precision (RSD %) of five replicate analyses within each run. A mean Fe concentration of 3.3 ng/l with a relative standard deviation of 10.2% was measured. The mean spike recovery value of the ng/l spiked addition was 94.6%. This result is very satisfying because, not only is it consistent with the proposed SEMI Tier Guidelines, but also demonstrates that Fe, traditionally a very difficult element by ICP- MS, can be determined at the 3-ng/L Fig. 3. Long-term stability of 39 K in 30% H 2 O 2 using DRC mode (reaction cell pressurized). Fig. 4. Long-term stability of 40 Ca in 30% H 2 O 2 using DRC mode (reaction cell pressurized). Fig. 5. Long-term stability of 56 Fe in 30% H 2 O 2 using DRC mode (reaction cell pressurized). 57

16 Fig. 6. Long-term stability of 23 Na in 30% H 2 O 2 using standard mode (reaction cell vented). TABLE IV Quantitation, Spike Recovery, and Precision of Five Separate 56 Fe Determinations in 31% H 2 O 2 Over Two Days Run Concn in Concn with Concn. Spike RSD No. 31% H 2 O ng/l measured recovery (%) (unspiked) spike (%) Mean TABLE V Spike Recovery Study of ng/l Analyte in 31% H 2 O 2 Using DRC Mode Mean RSD Mean concn. a (%) spike (ng/l), recovery n=5 (%) 40 Ca V Cr Mn a Concentration = measured concn. concn. in original sample. TABLE VI Spike Recovery Study of ng/l Analyte in 31% H 2 O 2 Using Normal Mode Mean RSD Spike concn. a (%) recovery (ng/l), n=5 (%) Mg V Ni Cu As Ag Cd Sn Sb Pb U a Concentration = measured concn. concn. in original sample. level directly in neat 31% H 2 O 2, using external calibration, with good long-term reproducibility. Table V summarizes the results of a spike recovery experiment for four other analytes determined in DRC mode. Recoveries using direct calibration are well within the range required by SEMI, indicating that there are no matrix-induced signal suppression effects when rugged normal temperature plasma conditions are applied. The results of the analysis performed using the standard mode of the instrument are summarized in Table VI. With the exception of Mg, which is the most likely contamination, the results of the spike recovery study for 11 analytes are, once again, well within the range specified by the proposed SEMI Tier Guidelines. CONCLUSION Over the past decade, ICP-MS has become the instrumental tool of choice in many routine and research laboratories for analyzing a wide variety of sample types. Many laboratories switch to ICP-MS when other techniques cannot be used to solve the analytical problem. Unfortunately, like any other analytical technique, ICP-MS is not totally free from interferences. Over the years, many spectral interference problems were solved using either high resolution spectrometers, reduced temperature plasmas, alternative sample introduction devices, or applying on-line chemistry. However, in order to further improve sample throughput and continually achieve lower detection limits, alternative solutions were required to achieve these goals. This became the major driving force in the development of the DRC technology. Chemical resolution achieved through sophisticated ion chemistry inside a dynamic reaction cell reduces or even eliminates a num- 58

17 Vol. 20(2), March/April 1999 ber of the most troublesome spectral interferences in ICP-MS. Very little sensitivity is lost compared with conventional quadrupolebased ICP-MS, but the noise level is reduced by as much as two orders of magnitude. This leads to a significant improvement in detection capability. Coupled with the increased freedom from spectral interferences, DRC-ICP-MS shows great potential to solve difficult analytical problems, particularly those encountered in the microelectronics industry. The results presented in this study demonstrate the efficiency of Dynamic Reaction Cell technology. A series of analytes was successfully determined at very low ng/l levels using robust normal temperature plasma conditions. For the so-called cool or cold plasma elements, the advantage of chemical resolution has been clearly demonstrated. Iron was determined at the 3-ng/L level with a measurement precision of close to 10%. Spike recoveries for Fe and a suite of other analytes were well within the range required by SEMI Tier D Guidelines. NH 3 was the reaction gas of choice for all analytes determined in DRC mode. Further experiments have shown that NH 3 is also the most efficient gas for the determination of K, Ca, Cr, Mn, and Fe in high purity water, HNO 3, or HCl. For that reason, the method developed for H 2 O 2 can be easily transferred and applied to the analysis of other sample types found in the semiconductor industry. REFERENCES 1. S. J. Jiang, R.S. Houk, and M.A. Stevens, Anal. Chem. 60, 1217 (1988). 2. S.D. Tanner, M. Paul, S.A. Beres, and E.R. Denoyer, At. Spectrosc. 16, 1 (1995). 3. E.R. Denoyer, S.D. Tanner and U.Voellkopf, Spectroscopy, February 1999, in press. 4. S.D. Tanner and V.I. Baranov, At. Spectrosc., Vol 20, No. 2, 45 (1999). 5. V.I. Baranov and S.J. Tanner, in Plasma Source Mass Spectrometry: Developments and Applications II, G. Holland and S.D. Tanner, eds., The Royal Society of Chemistry, Cambridge, (1999), in press. 6. S.J. Tanner and V.I. Baranov, in Plasma Source Mass Spectrometry: Developments and Applications II, G. Holland and S.D. Tanner, eds., The Royal Society of Chemistry, Cambridge, (1999), in press. 7. U. Voellkopf, V.I. Baranov and S.J. Tanner, in Plasma Source Mass Spectrometry: Developments and Applications II, G. Holland and S.D. Tanner, eds., The Royal Society of Chemistry, Cambridge, (1999), in press. 8. D.S. Bollinger and A. Schleisman, in Plasma Source Mass Spectrometry: Developments and Applications II, G. Holland and S.D. Tanner, eds., The Royal Society of Chemistry, Cambridge, (1999), in press. 9. S.D. Tanner, V.I. Baranov, U. Voellkopf, and M. Werner, in Proceedings of the 17th Annual Semiconductor Pure Water and Chemicals Conference, Santa Clara, CA, March

18 Analysis of High Purity Acids Using a Dynamic Reaction Cell ICP-MS David S. Bollinger and Anthony J. Schleisman Air Liquide Electronics and Chemicals, Inc. Dallas, TX USA INTRODUCTION Matrix removal (1 3), plasma modifications (4,5), post-plasma reactions (6,7), high-resolution mass spectrometry (8,9), and mathematical corrections (10) have all been used in the reduction or elimination of polyatomic interferences in ICP-MS. A new approach to post-plasma reactions that eliminate polyatomic interference in standard quadrupole ICP-MS will be presented in this paper. Post-plasma reactions do not interfere with the plasma chemistry; therefore, normal hot plasma conditions can be used. Post-plasma chemistry was done in this study using a novel dynamic reaction cell (DRC) that was inserted between the ion lens and the analyzer of a conventional ICP-MS. Ion-molecule reactions were promoted by a reaction gas flowing through the DRC at rates up to about 1.5 ml/min. The ideal reaction gas will react with polyatomic isobaric interferences, atomic isobaric interferences, or neighboring m/z interferences, but not with the analyte ions. The reaction gas can also be chosen to react exclusively with the analyte ion to produce a polyatomic species at a new m/z that has no interferences. The direct determination of ultratrace metal contaminants in semiconductor-grade acids has been complicated by polyatomic interferences from the acid matrices (e.g., ClO, ClOH, and ArCl from HCl). These interferences can be overcome by many different approaches, but the ideal approach would be direct aspiration of the concentrated or diluted acid. The Polyatomic Interference Removal The plasma ions are continuously sampled through the cones and ion lens, and eventually pass through the Dynamic Reaction Cell where the post-plasma chemistry occurs. From there, they enter the quadrupole analyzer, where they are sorted conventionally by their m/z ratio. The DRC operates under dynamic flow conditions; therefore, the reaction gas must be continuously admitted under accurately controlled flow rates. The choice of reaction gas can be determined from the rate constant of the reaction, the number of collisions between the interference ion and the reaction gas, and the electrical conditions in the collision cell. The flow rate of reaction gas needed to eliminate the polyatomic interference(s) must be determined empirically. The optimum flow rate is found by monitoring the analyte m/z while varying the reaction gas from ml/min. Figure 1 is a plot of ammonia gas flow rate into the DRC versus pulse intensity of 37 Cl 16 O 16 O at m/z 69 and 35 Cl 18 O 18 O at m/z 71 for a 10% (v/v) hydrochloric acid solution. Both interference signals are proportionally reduced as the flow of ammonia is increased into the DRC; the higher the flow rate, the higher the pressure of the reaction gas in the DRC, thus prouse of the DRC-ICP-MS to eliminate these polyatomic interferences will be detailed in this paper. EXPERIMENTAL Instrumentation The experiment was carried out on a beta version of the ELAN 6100 DRC (Perkin-Elmer SCIEX, Concord, ON, Canada), equipped with a quartz spray chamber and a quartz nebulizer under the conditions listed in Table I. The instrument design is described by Tanner, Baranov, Völlkopf, and Werner in reference (11). Standards and Reagents The 1-ng/g metal standards were prepared by serial dilution from separate NIST (National Institute of Standards and Technology, Gaithersburg, MD USA) spectroscopy standards containing certified levels of near 10 mg/g of the individual elements. Semiconductor-grade deionized water (DIW) was used TABLE I Instrument Operating Parameters Beta-version of ELAN 6100 DRC ICP RF Power 1125 W Nebulizer gas 0.91 ml/min Auxiliary gas 1.5 ml/min Plasma gas 15 L/m DRC gas VLSI (Very Large Scale Integration)- grade NH 3,VLSI-grade N 2 DRC gas flow 0 to 1.4 ml/min Quartz cyclonic spray chamber Quartz Meinhard Nebulizer Peristaltic pump uptake 1 ml/min throughout the experiment for dilutions and blanks. The 10% (v/v) HNO 3 solution was prepared by diluting SEMI Tier B nitric acid with DIW. The 10% (v/v) HCl solution was prepared by diluting SEMI Tier C hydrochloric acid with DIW. RESULTS AND DISCUSSION *Corresponding author. 60 MS-116 D-6063

19 Vol. 20(2), March/April 1999 ducing more reactions per unit time. It can be seen that the interference signals at m/z 69 and m/z 71 are reduced by almost four orders of magnitude at the optimum ammonia flow. The reaction gas must ideally eliminate the interferences, but it must not react rapidly with the analyte. Figure 2 is a plot of ammonia flow rate versus pulse intensity at m/z 69 and 71 for a 10% HCl (v/v) solution containing 1 ng/g of gallium. Initially, the 69 Ga signal decreases rapidly because the predominant component of the signal is from the 37 Cl 16 O 16 O interference. The signal then increases from collisional focussing of the analyte mass in the DRC (11) and finally decreases slowly as the reaction gas flow is increased to 1.1 ml/min. The optimum gas flow rate of 1.0 ml/min was determined by the ICP- MS software from the data in Figures 1 and 2. The optimum flow rates are very reproducible over long periods; the authors have used the optimum flow rates for several months without any changes. Detection Limits for Semiconductor Chemicals An additional benefit of the DRC is that in the standard mode the background noise has been reduced to less than one count per second without any loss of sensitivity. The corresponding improvement in signal-to-noise results in very low detection limits for the standard mode. Tables II, III, and IV list the detection limits for DIW, HNO 3, and HCl, respectively. The detection limits were determined from twenty 10-second integrations using the instrument conditions stated in Table I. These detection limits were determined in the acid sample, diluted 1:10, in standard mode for elements with no interferences and in DRC mode for elements with interferences. The polyatomic interferences from hydrochloric acid on vanadium and arsenic were the only interferences that ammonia could not eliminate. A suitable DRC gas for the vanadium interferences is being investigated. The arsenic interferences have potentially been eliminated by using nitrogen as the DRC gas (12).The detection limits for potassium in all three chemicals is high because of a potassium background either coming from the introduction system or from residual contamination in the sample itself. The nickel detection limits are high because a nickel cone was used. The tin and zinc detection limits were also high because of either a contamination source in the sample introduction system or in the sample itself. The detection limits will improve when these sources are eliminated. CONCLUSION The DRC has been shown to eliminate many of the polyatomic interferences that have caused problems in the direct analysis of ultrapure acids. The detection limits for all but a small number of elements are sufficient to meet the current SEMI specifications listed in Tables II, III, and IV. However, small changes in the introduction system should significantly improve the detection limits for tin and zinc. In addition, a torch made from low potassium quartz might improve the potassium detection limit and Fig. 1. Signal intensity for 37 Cl 16 O 16 O at m/z 69 and 35 Cl 18 O 18 O at m/z 71 vs. DRC gas (NH 3 ) flow for a 10% (v/v) HCl solution. Fig. 2. Signal intensity of 1 ng/g of gallium at m/z and vs. DRC gas (NH 3 ) flow for a 10% v/v HCl solution. 61

20 TABLE II Detection Limits in (pg/g) for DIW Using Standard (STD) or Enhanced (DRC) Modes D.L. Background Element (3 σ) Equiv. conc. Mode Ag (107) STD Al (27) DRC As (91) * DRC Au (197) STD B (11) STD Ba (138) STD Be (9) STD Bi (209) STD Ca (40) DRC Cd (111) STD Co (59) STD Cr (52) DRC Cu (63) STD Fe (56) DRC Ga (69) STD Ge (74) STD In (115) STD K (39) DRC La (139) STD Li (7) STD Mg (24) STD Mn (55) STD Mo (98) STD Na (23) STD Nb (93) STD Ni (58) 1 8 DRC Pb (208) STD Pd (105) STD Pt (195) STD Rb (85) STD Sb (121) STD Sn (120) STD Sr (88) STD a (181) STD Ti (48) STD Tl (205) STD U (238) STD V (51) STD W (184) STD Zn (64) STD Zr (90) STD * Measured as AsO at Mass 91. TABLE III Detection Limits in (pg/g) for Concentrated Nitric Acid Using Standard (STD) or Enhanced (DRC) Modes Detection Limits (3 σ) Measured in Calculated for SEMI Spec Element HNO 3 diluted 1:10 original conc. acid Mode (TIER B) Ag (107) STD - Al (27) DRC 1000 As (91)* DRC 1000 Au (197) STD 1000 B (11) STD 1000 Ba (138) STD - Be (9) STD - Bi (209) STD - Ca (40) DRC 1000 Cd (111) STD - Co (59) STD - Cr (52) DRC 1000 Cu (63) STD 1000 Fe (56) DRC 1000 Ga (69) STD - Ge (74) STD - In (115) STD - K (39) DRC 1000 La (139) STD - Li (7) STD - Mg (24) STD 1000 Mn (55) STD 1000 Mo (98) STD - Na (23) STD 1000 Nb (93) STD - Ni (58) DRC 1000 Pb (208) STD 1000 Pd (105) STD - Pt (195) STD - Rb (85) STD - Sb (121) STD 1000 Sn (120) STD 1000 Sr (88) STD - Ta (181) STD - Ti (48) STD 1000 Tl (205) STD - U (238) STD - V (51) STD - W (184) STD - Zn (64) STD 1000 Zr (90) STD - * Measured as AsO at Mass

21 Vol. 20(2), March/April 1999 TABLE IV Detection Limits in (pg/g) for Concentrated Hydrochloric Acid Using Standard (STD) or Enhanced (DRC) Mode Detection Limits (3 σ) Measured in HCl Calculated for SEMI Spec Element acid diluted 1:10 original conc. acid Mode (TIER C) Ag (107) STD - Al (27) STD 100 As (91)* DRC 100 Au (197) STD 100 B (11) STD 100 Ba (138) STD - Be (9) STD - Bi (209) STD - Ca (40) DRC 100 Cd (111) STD - Co (59) STD - Cr (52) DRC 100 Cu (63) STD 100 Fe (56) DRC 100 Ga (69) DRC - Ge (74) DRC - In (115) STD - K (39) DRC 100 La (139) STD - Li (7) STD - Mg (24) STD 100 Mn (55) DRC 100 Mo (98) STD - Na (23) STD 100 Nb (93) STD - Ni (58) DRC 100 Pb (208) STD 100 Pd (105) STD - Pt (195) STD - Rb (85) STD - Sb (121) STD 100 Sn (120) DRC 100 Sr (88) STD - Ta (181) STD - Ti (48) DRC 100 Tl (205) STD - U (238) STD - V (51)** - - DRC - W (184) STD - Zn (64) STD 100 Zr (90) STD - * Measured as AsO at Mass 91. ** Investigations into effective DRC gas for polyatomic removal in progress. the use of a platinum cone will improve the nickel detection limit. These analyses were performed on 10% (v/v) dilutions because there was little plasma suppression or enhancement of the analyte species when compared to water-based solutions. It is anticipated that the acids could be analyzed at higher acid concentrations if matrix-matched standards are used for calibration. Because the DRC eliminates interferences through postplasma reactions, normal high power, high temperature plasma conditions can be used for the analysis. The use of post-plasma reactions as a routine solution to analytical problems looks to be very promising for ultratrace analysis of semiconductor-grade chemicals. REFERENCES 1. A. L. Gray, and A. R. Date, Analyst 108, 1033 (1983). 2. R. Tsukahara and M. Kubota, Spectrochim. Acta, Part B, 45B, 581 (1990). 3. M. R. Plantz, J. S. Fritz, F. G. Smith, and R. S. Houk, Anal. Chem. 61, 149 (1989). 4. E. H. Evans, and L. Ebdon, J. Anal. At. Spectrom. 4, 299 (1989). 5. J. W. H. Lam, and G. Horlick, Spectrochim. Acta, Part B, 5, 425 (1990). 6. J. T. Rowan and R. S. Houk, Appl. Spectrosc. 34, 38 (1989). 7. D. W. Koppenaal, C. J. Barinaga, M. R. Smith, J. Anal. At. Spectrom. 9, 1053 (1994). 8. N. Bradshaw, E. F. H. Hall, and N. E. Sanderson, J. Anal. At. Spectrom. 4, 801 (1989). 9. M. Morita, H. Ito, T. Uehiro, and K. Otsuka, Anal. Sci. 5, 609 (1989). 10. S. Munro, L. Ebdon, and D. J. McWeeny, J. Anal. At. Spectrom. 1, 211 (1986). 11. S. Tanner, V. Baranov, U. Völlkopf, and M. Werner, 17th Annual Semiconductor Pure Water and Chemicals Conference, 183 (1998). 12. A. Schleisman and D. Bollinger, 6th International Conference on Plasma Source Mass Spectrometry, Durham, U.K. (1998). 63

22 The Benefits of a Dynamic Reaction Cell to Remove Carbon- and Chloride-Based Spectral Interferences by ICP-MS Kenneth Neubauer* and Uwe Völlkopf The Perkin-Elmer Corporation 761 Main Avenue, Norwalk, CT USA INTRODUCTION As inductively coupled plasma mass spectrometry (ICP-MS) systems become more sensitive, spectral interferences become more noticeable and problematic. Three problem elements by ICP-MS are chromium, manganese, and arsenic which are plagued by interferences from carbon- and chloride-based polyatomic species. Attempts to eliminate or reduce the impact of these interferences using magnetic sector high resolution systems, low temperature plasmas, or interelement correction equations have been reported, but only with limited success (1 3) where carbon and chlorine are present at high concentrations. This work describes the use of a new Dynamic Reaction Cell (DRC) ICP-MS for reduction of the following polyatomic interferences, which inhibit determination of chromium, manganese, and arsenic: 40 Ar 12 C +, 35 Cl 16 OH +, 40 Ar 13 C +, 37 Cl 18 O +, and 40 Ar 35 Cl +. The DRC has been described in detail elsewhere (4) but basically consists of an enclosed cell pressurized with a reactive gas through which the ion beam is passed. Ion molecule reactions occur between the reactive gas and species in the ion beam; these reactions transform interfering species into innocuous byproducts that do not interfere with the analysis. Isotopic ratios and spike recovery studies are used to demonstrate the extent of interference reduction and to illustrate the potential of the DRC to determine elements in matrices which were previously considered too difficult to analyze by ICP-MS. *Corresponding author ABSTRACT Matrix-derived spectral interferences inhibit the determination of a variety of elements by ICP-MS. Carbon- and chloridebased matrices render low level chromium, manganese, and arsenic determination very difficult, if not impossible, due to the formation of interfering ions. A dynamic reaction cell (DRC) ICP-MS using chemical resolution has been developed to minimize this problem. This work demonstrates that major interferences on m/z 52, 53, 55, and 75 resulting from 40 Ar 12 C +, 35 Cl 16 OH +, 40 Ar 13 C +, 37 Cl 18 O +, and 40 Ar 35 Cl +, respectively, can be effectively reduced, so that ng/l levels of chromium, manganese, and arsenic can be determined in carbon- and chloride-based matrices. EXPERIMENTAL Instrumentation All work was performed in a class 1000 clean room using an ELAN 6100 DRC ICP mass spectrometer. The sample introduction system consisted of a quartz cyclonic spray chamber and a quartz Meinhard nebulizer, with an uptake rate 1 ml/min. A normal temperature plasma was used (power:1200 W, nebulizer argon flow:1.02 L/min), and the instrument s AutoLens feature was enabled (5). All data were acquired in the peak hopping mode, except for the spectra shown in this paper, which were acquired in the scanning mode for display purposes only. All calibrations were performed externally with aqueous standards; no internal standards were used. The reaction gases investigated were semiconductor-grade ammonia and hydrogen (Matheson Gas Products Inc., Montgomeryville, PA, USA). To reduce contaminants, the gases were passed through a scrubber prior to entering the instrument. The gas flow rates varied from 0.05 to 1.05 ml/min (calibrated for argon). All chemicals were reagent grade and used as received without further purification. Food-grade sucrose was used to prepare the carbon solutions. All solutions were made with ultrapure distilled deionized water. RESULTS AND DISCUSSION Chromium Determination in Carbon The determination of chromium in a carbon matrix is difficult due to the formation of 40 Ar 12 C + and 40 Ar 13 C + whose signals coincide with the two major chromium isotopes at m/z 52 and 53. These studies focus on reducing the ArC + interferences, so that low levels of chromium can be detected. To determine the feasibility of detecting chromium in a carbon matrix, the signals at m/z 52 and 53 were monitored for spiked (1 µg/l Cr) and unspiked 1200 mg/l carbon solutions while varying the reaction gas flow rate. The thermodynamically expected reaction of ArC + with ammonia is: ArC + + NH 3 NH 3+ + Ar + C. Figure 1 displays a plot of m/z 52 and 53 signals from these solutions as a function of ammonia flow rate. As seen in this figure, the signal from the unspiked solution decreases with increasing ammonia flow while the signal from 64 MS-117 D-6064

23 Vol. 20(2), March/April 1999 Fig. 1. Flow curves for chromium determination in 1200 mg/l carbon matrix. B = blank (matrix only); S = 1 µg/l Cr spike in matrix. the spiked solutions remains almost constant. These results demonstrate that the carbon-based interferences react much faster with ammonia than the chromium and will be preferentially reduced. As a result, low levels of chromium should be detectable in a carbon matrix. This experiment was repeated with hydrogen as the reaction gases, but ammonia proved most effective in reducing the 40 ArC + signals. Figures 2 and 3 confirm that the ArC + interferences are greatly reduced. Figure 2 shows two overlayed mass spectra of a 1200-mg/L carbon solution; the dashed spectrum was acquired with an ammonia flow rate of 0.6 ml/min, while the solid spectrum was obtained in a normal mode with no reaction gas. These spectra clearly show that the 40 Ar 12 C + signal is reduced by 3.5 orders of magnitude, and the 40 Ar 13 C + signal is decreased by about two orders of magnitude. In fact, background signal intensities at m/z 52 and m/z 53 have been reduced to less than 100 cps with the ammonia reaction gas present. Figure 3 displays overlayed mass spectra of 20 ng/l chromium spikes in 0.2% HNO 3 and 1200 mg/l carbon acquired with an ammonia flow rate of 0.6 ml/min. Equal peak intensities confirm that the ArC + interferences have been eliminated at both m/z 52 and m/z 53. Further verification of ArC + reduction is demonstrated through the measurement of isotope ratios. The natural 53 Cr/ 52 Cr isotope ratio is 0.113; Table I displays the ratios measured in solutions containing ,000 mg/l of carbon, both in the presence and absence of ammonia. As seen in the table, a measured chromium isotope ratio of 0.12 is determined in the presence of ammonia (no mass bias correction was made to calibrate for accuracy). The important point is that the ratio remains constant as the carbon content of the matrix increases up to 10,000 mg/l. By comparison, without ammonia, the isotope ratios vary with carbon content and are clearly inaccurate. Fig. 2. Mass spectra of 1200 mg/l organic carbon with (dashed) and without (solid) ammonia. Fig. 3. Mass spectra of 20 ng/l chromium in 1200 mg/l organic carbon (solid) and 0.2% nitric acid (dashed). Ammonia flow rate = 0.6 ml/min. TABLE I Measured Chromium Isotope Ratios in Carbon Matrices Matrix Measured 53 Cr/ 52 Cr Ratio (mg/l Carbon) Ammonia No Ammonia ,

24 TABLE II. Spike Recoveries Matrix Isotope Conc. Recovery Concentration (ng/l) (ng/l) (a) 1200 mg/l Carbon 52 Cr Cr (b) 1% HCl 52 Cr Mn The ability to quantitatively determine chromium in the presence of sugar was determined by spiking a 1200-mg/L sugar solution with 20 ng/l of chromium and then analyzing against an aqueous, external calibration curve. The results of this analysis appear in Table II and show that chromium consistently reads back between 22 and 23 ng/l. These data, combined with the spectra shown in Figure 2, indicate that the majority of the ArC + interferences are eliminated. The 2 ng/l discrepancy between the spiked and experimentally determined values may be due to chromium originally present in the sucrose or incomplete removal of the ArC + species. However, determination of the potential 2 ng/l chromium in the sucrose is difficult and challenges the ability of graphite furnace atomic absorption. Nevertheless, these results demonstrate that trace amounts of chromium can be quantitatively determined in the presence of carbon. Finally, estimated detection limits were determined based on counting statistics and are displayed in Table III. The calculations used for these measurements are available in the ELAN DRC software to allow users to determine the optimum gas flows for various analyses. Table III shows that, although the carbon content increases by about 50 times, the detection limits only increase three times. Chromium and Manganese Determination in Chloride The same approach used for the determination of chromium in carbon was applied to chloride matrices. Chloride-derived interferences include 35 Cl 16 OH + (m/z 52) and 37 Cl 16 O + (m/z 53) for chromium, and 37 Cl 18 O + (m/z 55) for manganese. The thermodynamically expected reactions of ClOH + and ClO + with ammonia are: ClOH + + NH 3 ClO + NH 4 + ClO + + NH 3 ClO + NH 3 + Figure 4 displays the m/z 52 and 53 intensities as a function of ammonia flow rate for a spiked (with 1 µg/l) and unspiked 1% HCl solution. The m/z 52 signal from the spiked solution decreases slightly with increasing ammonia flow, while the signal from the unspiked solution decreases more rapidly. This observation indicates that the 35 Cl 16 OH + is effectively reduced without adversely affecting 52 Cr +. However, the signals at m/z 53 decrease at about the same rate, which indicates that both 37 Cl 16 O + and 53 Cr + react at about the same rate with ammonia. Therefore, only 52 Cr + TABLE III Estimated Chromium Detection Limits in Varying Carbon Concentrations Matrix Estimated Detection Limits (ng/l) (mg/l Carbon) 52 Cr 53 Cr , Fig. 4. Flow curves for chromium determination in 1% HCl matrix. B = blank (matrix only); S = 1 µg/l Cr spike in matrix. Fig. 5. Flow curves for manganese determination in 1% HCl matrix. B = blank (matrix only); S = 1 µg/l Mn spike in matrix. can realistically be determined in a chloride matrix using ammonia as a reaction gas. Figure 5 shows the same experiment carried out for manganese. In this case, higher ammonia flow rates clearly reduce the 37 Cl 18 O + signal more rapidly than the 55 Mn + signal, which enables manganese to be determined at m/z 55 in a chloride matrix. 66

25 Vol. 20(2), March/April 1999 Spectral scans of Cr and Mn appear in Figures 6 and 7. Figure 6 displays overlayed spectra of a 1% HCl solution acquired with and without ammonia present in the reaction cell. Clearly, the presence of ammonia reduces the intensities of the signals from m/z 51 55, which indicates the reduction of 35 Cl 16 O +, 35 Cl 16 OH +, 37 Cl 16 O +, 37 Cl 16 OH +, and 37 Cl 18 O +. Figure 7 displays mass spectra of 20 ng/l spiked in 0.2% HNO 3 and 1% HCl solutions. The reduction of 35 Cl 16 OH + is evident in these spectra since the peaks at m/z 52 are of equal intensities. Likewise, the peak overlap at m/z 55 confirms that 37 Cl 18 O + is effectively reduced without eliminating the 55 Mn + signal. However, the signal produced from the HCl solution is more intense than that from HNO 3 at m/z 51 and 53, thus indicating that 35 Cl 16 O + and 37 Cl 16 O + are not reduced with ammonia gas. Quantitative determination of chromium and manganese in HCl were explored by analyzing a 20-ng/L spike in 1% HCl against an aqueous, external calibration curve. The results appear in Table II and show that 52 Cr consistently reads back between 23 and 25 ng/l, while the 55 Mn gives consistent values of 17 and 18 ng/l. Additionally, 53 Cr reads back high (330 ng/l) as expected, based on the spectra shown in Figure 8. The 4-ng/L discrepancy between the spike and experimental values for the 52 Cr may be due to chromium contamination in the acid or incomplete removal of ClOH +, but it is small compared to the 330-ng/L BEC without interference reduction. However, quantitative determination of such low levels of chromium is difficult. The cause of the 3-ng/L difference in manganese values is, as yet, undetermined. Despite the slight differences between the spiked and determined values, these results still show the degree to which interfering species can be eliminated using chemical resolution in the DRC. Arsenic Determination in Chloride For the determination of arsenic in a chloride matrix, hydrogen is the preferred reaction gas because it reacts faster with 40 Ar 35 Cl + than 75 As +, according to the following probable reaction scheme: ArCl + + H 2 ArH 2+ + HCl ArH 2+ + H 2 H 3+ + Ar Fig. 6. Mass spectra of 1% HCl with (dashed) and without (solid) ammonia. Fig ng/L multielement spike (dashed) (V, Cr, Mn) in 1% HCl (solid) and 0.2% nitric acid. Ammonia flow rate = 0.8. Evidence of the effectiveness can be seen in Figure 8. The spectral display on the left shows the overlap of signals from a 1-µg/L arsenic solution and a 1-µg/L arsenic spike in a 1000-mg/L sodium chloride solution. The equal intensities of both signals indicate that the 40 Ar 35 Cl + has been greatly reduced. The spectral display on the right shows an expanded portion of the baseline, which compares the signals from a 1000-mg/L NaCl solution and distilled deionized water (DDIW). As shown here, the Fig. 8. Mass spectra of DI water, 1 µg/l As in DI water, 1000 mg/l NaCl, and 1 µg/l As in 1000 mg/l NaCl. 67

26 Fig. 9. Two-hour stability from continuous aspiration of 100 ng/l As in 1000 mg/l NaCl. REFERENCES 1. M. Campbell, C. Demesmay, and M. Olle, J. Appl. Anal. Spectrosc. 9, 1379 (1994). 2. K. van den Broeck, C. Vandescasteele, and J. Geuns, J. Appl. Anal. Spectrosc. 12, 987 (1997). 3. J. Hedrick and A Guiterrez, Biological Applications of Cool Plasma ICP-MS, FACSS, 1997, Rhode Island, USA. 4. E. Denoyer, S. Tanner, and U. Voellkopf, Spectroscopy 14, 2 (1999). 5. E. Denoyer, D. Jacques, E. Debrah, and S. Tanner, At. Spectrosc. 16, 1, (1995). background counts from the NaCl solution are less than 20 cps, thus demonstrating that the 40 Ar 35 Cl + has been virtually eliminated. In order to confirm the reduction of the 40 Ar 35 Cl + interference on arsenic, spike recoveries over a two-hour time period were performed. An external, aqueous calibration curve was first established, and then a 1000-mg/L NaCl solution was analyzed ten times. Directly after the NaCl analysis, a solution containing 100 ng/l arsenic in 1000 mg/l NaCl was continuously aspirated for two hours. Readings were taken every three minutes, with the results being displayed in Figure 9. As evidenced in the figure, the continuous aspiration of NaCl did not lead to any longterm signal loss, and the As recoveries remained between 95% and 111% as shown in Table II(c). These results indicate excellent long-term stability in the presence of the sodium chloride matrix. CONCLUSION This work demonstrates the effectiveness of dynamic reaction cell ICP-MS in removing carbon- and chloride-based interferences which inhibit chromium, manganese, and arsenic determinations. In a 1200-mg/L organic carbon matrix, 20 ng/l chromium can be accurately detected at both the major isotopes. Accuracy of the measurement is confirmed through the measurement of isotope ratios. In HCl, both manganese and chromium can be measured; however, the 53 Cr isotope cannot be accurately determined since the 37 Cl 16 O + signal cannot be removed without simultaneously reducing 53 Cr +. Finally, the successful determination of arsenic, along with its long-term stability in a chloride matrix, is also demonstrated. Future work will focus on the determination of vanadium in chloride matrices and copper in sodium matrices. 68