Analytical performance of axial inductively coupled plasma time of Øight mass spectrometry (ICP-TOFMS) Xiaodan Tian, Ha kan Emteborg and Freddy C. Adams* Micro and Trace Analysis Center, MiTAC, Department of Chemistry, University of Antwerp (UIA), Universiteitsplein 1, B-2610 Antwerp, Belgium Received 19th August 1999, Accepted 11th October 1999 Analytical performance of an axial ICP-TOFMS, in terms of accuracy, precision, resolution, signal stability, repeatability, reproducibility, precision of isotope ratios and detection limits is reported. Cool plasma conditions (low power and high central gas Øow) allow the determination of K, Ca and Fe at trace levels.Detection limits for 64 elements are reported with typical values from 0.5±20 pg ml21 (3 s criterion). The long-term stability over 4 h for the raw signal ranges between 1.0% RSD for 7Li to 2.1% RSD for 40Ca operating under cool plasma conditions. Under normal plasma conditions the corresponding values are 6.9% RSD for 208Pb to 12.8% RSD for 59Co. Accuracy for the determination of 16 different elements is assessed by measuring NIST 1643d (trace elements in water), yielding results in agreement with certiÆed values.Results for isotopic ratio determinations demonstrate that a precision of 0.07±0.7% RSD is obtained within a short data acquisition period. Introduction Since inductively coupled plasma mass spectrometry (ICP-MS) was introduced in the early '80s, it has become a widespread multi-elemental technique for elemental analysis in many sample types.1,2 It offers almost complete elemental coverage combined with a high sensitivity for most elements, detection limits being in the 1±10 pg ml21 range.Another important advantage of ICPMSover other analytical techniques, such as inductively coupled plasma atomic emission spectrometry (ICP-AES) and atomic absorption spectrometry (AAS), is the possibility of quantiÆcation by isotope dilution. Most ICP-MS instruments employ quadrupole mass Ælters, resulting in high sensitivity, low operating costs, and modest vacuum requirements.Although quadrupole ICP-MS has matured into a powerful technique, there are still several remaining weaknesses, such as isobaric overlaps, matrix interferences, limited precision (especially for isotope ratios) and limitations in elemental coverage when measuring rapid transient signals.3 The two latter shortcomings are a direct consequence of the fact that quadrupole ICP-MS is a sequential (scanned) mass spectrometer.As an alternative, Myers et al. Ærst described in 1993 an ICP-MS based upon a time-of-Øight (TOF) mass spectrometer placed orthogonal to the ion source.4 In later work, using the same system, results concerning optimization, characteristics and performance for isotope ratios were published.5±7 TOFMS is inherently a simple type of mass spectrometer with high transmission efÆciency. The three most important assets are simultaneous ion extraction, quasi-simultaneous nature of spectral generation and extremely high data acquisition speed over the whole mass spectrum.Ions are generated and extracted from the ion source through a sampler and a skimmer cone. The ions are then guided into an acceleration region where a high voltage pulse is imposed, accelerating the ions to a similar kinetic energy and subsequently directing them into the Æeld-free Øight tube. Lighter ions will consequently travel faster than heavier ions, therefore, ions of different mass (m/z) will separate into their individual mass components as they travel through the Øight tube and generate a complete mass spectrum. Generally, TOFMS was previously only used with pulsed ion sources, such as glow discharge (GD) matrix-assisted laser desorption/ionization, since the ions have to enter the Øight tube in the form of discrete packets (i.e.modulation). Since the repetition rate of a TOF instrument is determined by the Øight time of the heaviest ion through the Æeld-free region, and with a Øight time for atomic species of less than 40 ms, more than 20 000 spectra per s can be generated in ICP-TOF-MS.The high repetition rate and simultaneous detection capabilities of the TOF mass spectrometer not only reduce analysis time and improve precision on isotope ratios, but also make the TOF instrument an ideal choice for the analysis of transient signals. Especially for a better precision on isotope ratios, simultaneous extraction of ions from the plasma improves the performance compared to quadrupole ICP-MS through a better elimination of Øicker noise.ICP-TOFMS thereby overcomes some of the remaining weaknesses of quadrupole ICP-MS. Recently, the application of ICP-TOFMS for transient signal detection generated by ETV and laser ablation have been reported.8,9 Axial ICP-TOFMS was made commercially available by LECO (St. Joseph, MI, USA) in 1998.Compared with orthogonal ICP-TOFMS, an axial arrangement has less divergence of ions, which minimizes mass bias and increases resolution. When comparing time-of-Øight systems with quadrupole systems the duty factor is an important parameter. This is the fraction of ions that are extracted for mass analysis. In the axial system used here approximately 10% of the ions are extracted for analysis. It results in a lower overall sensitivitywhen compared with quadrupole systems used for detection of a single m/z.This is a direct consequence of the ion source modulation. An important feature of the axial ICP-TOFMS is its scheme to avoid detector overload. In quadrupole ICP-MS, the high background obtained at certain m/z from the plasma gas, entrained atmospheric gases and water can be eliminated by skipping over certain dc/rf settings that allow their transmission. In contrast to this an ICP-TOFMS instrument accepts ions of all m/z into the acceleration region regardless of whether they are of analytical interest or not.This may cause detector overload or even destruction by the intense Øux of certain matrix ions. Given this situation, ICP-TOFMS has to be equipped with an additional device to avoid detector overload, called the ``Transverse Rejection Ion Pulse (TRIP)''. This device involves the application of a high voltage pulse imposed at the time when a matrix ion passes by an ion gate in order to remove it from the ion beam.In some cases, analyte ions are also concomitant with matrix ions from the sample. Therefore, J. Anal. At. Spectrom., 1999, 14, 1807±1814 1807 This journal is # The Royal Society of Chemistry 1999additional deØection windows need to be implemented besides certain standard deØections for background from the plasma such as argides and ions derived from water ions. More details may be found in a previous publication.10 The time spent in determining 66 isotopes (the maximum number that can be measured with 30 bins per isotope acquisition window) is only determined by the integration time and number of replicates selected resulting in considerable time-saving. Each bin corresponds to a 2 ns mass window and over the whole mass spectrum 2000 bins are distributed.Note that time and mass can be used interchangeably when describing TOF mass spectrometry. Through the software the analyst can control the width of the isotope acquisition window.By collection of, e.g., a 15 bin mass window for each isotope, 2000/15~133 isotopes can be measured. Naturally a slight sensitivity loss is encountered in such a case. It should also be mentioned that a maximum of 19 bins can be used by the software for calculating the peak area so that peaks at higher mass are not fully integrated. This may inØict problems for the accuracy of isotope ratio measurements. Using quadrupole ICP-MS, the time spent for analysis is directly proportional to the number of isotopes multiplied by the integration time including settling time for the rods.The analysis time using ICP-TOFMS is thus independent of the number of isotopes being measured within the limits described above. In this paper, a set of results obtained by ICP-TOFMS will be presented. The results include basic analytical performance and Ægures of merit like detection limits, signal stability, precision of isotope ratios and accuracy of analytical results using a standard reference material.The instrumentation is competitive with quadrupole ICP-MS in terms of detection limits and precision, but offers the additional advantage of simultaneous ion extraction, which is important for achieving high precision of isotope ratios. The analysis time is reduced and hence sample throughput is increased thanks to the high speed. Rapid spectral generation and quasi-simultaneous detection capabilities also results in complete elemental coverage in micro volumes and rapid transients.Experimental Instrumentation The ICP instrument used in this study is shown in Fig. 1 (axial ICP-TOFMS, Renaissance, LECO). The basic elements include a plasma torch for generation of a continuous ion beam, an ion transmission system including a sampler, a skimmer, an extraction cone (ion lens 1), an ion optical array to position the ion beam for accelerating and repelling the ion beam segments in a modulated fashion, an ion mirror and a detector.Segments of ions enter the Øight tube and are directed to the ion mirror where they are refocused and reØected back towards the detector. Use of an ion mirror results in a higher resolution, lower level of background noise and provides a doubling of the Øight path, thereby allowing a more compact size of the vacuum system. The instrument has a dual frequency rf generator, which generates frequencies of 27.12 MHz and 40.68 MHz; in this work 40.68 MHz was used.For an outline of the performance for the two different frequencies, readers can Ænd useful information in an article by Vickers et al.11 The torch cartridge assembly with a standard torch (Fassel type) is mounted horizontally in the torch box. A Wu±Hieftje FAST spray chamber and a Meinhard nebulizer are used for sample introduction. Sample solutions are delivered at a Øow rate of 1 ml min21 using a peristaltic pump (Gilson, Villiers Le Bel, France, Miniplus 3), where the pump speed can be controlled from the software or directly on the pump.Mass Øow controllers control the gas Øow rates; the plasma (coolant), nebulizer and auxiliary (intermediate) argon Øows are preset at 16, 0.8 and 1.0 l min21, respectively, with a primary argon input pressure of 50 psi. The torch position is adjustable via precise software control in the X, Y and Z directions in relation to the sampling oriÆce for optimization purposes. The torch position together with forward power, coolant, auxiliary and nebulizer Øows are important initial parameters for achieving maximum signal-to-noise ratio.The sampler and skimmer oriÆces are maintained at ground potential. The sampler is made from nickel with an aperture diameter of 0.9 mm. Because of the high temperature from the plasma, the sampler is mounted on a water-cooled plate, which forms the front wall of the vacuum system. An external recirculating chiller (Model CFT-75, Neslab Instruments, Inc, Newington, NH, USA) is used to supply cooling water also for the load coil.The skimmer cone is more sharply angled than the sampler cone, with an aperture diameter of about 0.5 mm. The distance between the sampler and skimmer is 7 mm (standard for this instrument). The vacuum chamber of the mass spectrometer is divided into three separate chambers as shown in Fig. 1. The Ærst stage vacuum chamber is immediately behind the sampler cone. The Ærst chamber's pressure is pumped down with a rotary vane vacuum pump to approximately 2±3 Torr.The expanding gas from the ICP is sub-sampled through skimmer oriÆce into the second-stage vacuum region. The second-stage vacuum is pumped down with a turbo molecular pump, to a pressure of 1023 Torr. Behind the skimmer, there are ion stacks, which control the ion Øow (as shown in Fig. 1). The ions selected by the modulation optic sends all ions into the dc acceleration region of the TOF-MS where all ions reach a similar kinetic energy.The third stagevacuumispumpedby a second turbo molecularpump to 1026 Torr, the actual pressure when the ICP is being operated. At the opposite end of the Øight tube, an ion mirror is placed to double the Øight path of the ions and focus their energy spreads. The detector is located above the ion-optics, opposite the ion mirror. In front of the detector are three grids used to reduce the number of low-energy background ions striking the detector surface.The detector is a discrete dynode multiplier (ETP Model AF831H, Ermington, NSW, Australia). The detector is constructed so that ion counting and analog signals are obtained simultaneously. Ion counting is used for the most sensitive work near the detection limit while analog mode is used to extend the dynamic range at higher signal levels since the signal from the ion counting mode is not linear at concentrations above 10 ng ml21 for the most sensitive isotopes.Optimization of the instrument Three standard plasma conditions, corresponding to ``normal'', ``hot'' and ``cold'' plasma conditions, were adapted for this work. The plasma and instrumental settings used are indicated in Table 1. Slight modiÆcations to these settings are normally necessary on a daily basis. It is obvious from the table that a number of important parameters need be studied and optimized to obtain adequate sensitivity and resolution. Following plasma ignition and prior to optimization, it is important to ensure that deØection settings are valid, so thatOzand Arzare deØected to Fig. 1 Schematic diagram of LECO Renaissance axial ICP-TOFMS system used in this work. The function and operation of the various components are described in the text. The Øight path is 260.5 m. 1808 J. Anal. At. Spectrom., 1999, 14, 1807±1814avoid detector overload. Normally, a 100 ng ml21 mass calibration solution is used to optimize the instrument by maximizing the signal intensity for 115In.The optimization starts with adjustments of the torch position by changingX±Y(vertical position of the plasma relative to the sampler aperture) and Z (horizontal position which also deÆnes the sampling depth) positions for a maximum 115In signal. From that optimum the nebulizer Øow, together with forward power, is changed to obtain a new maximum. A number of parameters of the mass spectrometer are also important for obtaining maximum signal intensity such as the potential on ion lenses 1 and 2, Einzel lenses 1 and 2 and X- and Y-steering.For optimal resolution 208Pb and 209Bi should be observed by adjusting reØectron (ion mirror) settings so as to obtain symmetrical peaks with no tailing or fronting. It is also possible to measure the resolution directly in the optimization menu in the software to reach a value ofw400. Initial optimization of the cool plasma conditions is obtained by Ærst setting the plasma power and nebulizer Øow for minimum background at 40Arz.Under cool plasma conditions, the deØection at m/z 40 should be removed. Mass calibration is mandatory following all optimization events and prior to data collection, especially if parameters in the mass spectrometer have been changed. The Øight time (mass calibration) for each isotope is normally highly stable and does not need to be recalibrated once this has been done properly. Reagents For the performance studies described here, sample solutions were prepared in the laboratory from analytical-reagent grade standard solutions and diluted on a balance (Sartorius LA620S, Go» ttingen, Germany) to 2% HNO3 (by addition of 65% HNO3 Merck, Darmstadt, Germany, Suprapur grade).A 10 mg ml21 mass calibration solution containing Li, Mg, Sc, Co, Y, In, Ce, Ba, Pb and Bi was used. Rare earth elements were separated into two groups to avoid isobaric overlaps, 10 mg ml21 stock standard solutions of Pr, Sm, Tb, Dy, Ho, Tm, and Yb, and a second standard containing La, Nd, Eu, Cd, Er and Lu.For high ionization elements stock solutions containing 10 mg ml21 of Be, Ti, Ge, Y, Mo, Rh, Pd, Sb and Hf and another solution containing Zr, Ru, Cd, Ce, W, Ir, Pt, Tl, Th and U were used. Similarly, 10 mg ml21 stock solutions containing memory prone elements such as B, Nb, Se, Ta, Hg, Au and Te were also measured separately to obtain information on the detection limits. All these standards were obtained from LECO.A 10 mg ml21 mixed standard solution of Al, Sb, As, Ba, Be, Bi, Cd, Ca, Cr, Cs, Co, Cu, Ga, Fe, Rb, Li, Mg, Mn, Na, Ni, K, Se, Ag, Na, Sr, Tl, U, V and Zn (Perkin Elmer, U» berlingen, Germany, PE-mix) was also used throughout. Deionized water was obtained from a Milli-Q system (Millipore, Milford, MA, USA). The isotopic reference materials were obtained from the European Commission's Joint Research Center, Institute for Reference Materials and Measurements, Geel, Belgium. The isotopic reference material for magnesium (IRMM-009) was taken from a Øame-sealed quartz ampoule with aMg content of approximately 1023 mol kg21. The ampoule contained 4 ml of acidiÆed solution (0.2 mol dm21 HNO3).For Pt (IRMM-010), the Pt-wire supplied was dissolved in aqua regia, and then diluted to 1 mg ml21 for stock solution. For Rb another certiÆed reference material was used (NIST-SRM-984). The stock solutions were diluted to a concentration of 100 ng ml21 for Mg and 200 ng ml21 for Pt prior to isotope ratio measurement and all solutions were acidiÆed to 2%HNO3 (Merck, Suprapur).Table 1 Typical operating conditions for ICP-TOFMS.Mass calibration data included illustrating typical values only since they change slightly for different optimizations. Ion deØection settings reported are only valid for normal and hot plasma ICP Source Normal plasma Hot plasma Cool plasma Forward Power/kW 1.27 1.49 0.83 Plasma Flow/l min21 15.4 14.0 16.1 Auxiliary Flow/l min21 1.29 0.701 1.1 Nebulizer Flow/l min21 0.801 0.841 1.0 Mass Spectrometer/V Flight Tube ±1470 21470 21485 ReØectron Low 199 199 193 ReØectron High 1530 1530 1524 X Steering ±1470 21470 21470 Y Steering ±1610 21620 21620 Einzel Lens 1 ±1250 21200 21353 Einzel Lens 2 ±797 2810 2748 Ion Lens 1 ±220 2393 2125 Ion Lens 2 ±19 247 2130 Detector ±2790 22790 22700 Mass calibration/ns Li-7 6746 Mg-24 12138 Co-59 18800 Y-89 22996 In-115 26084 Ba-138 28536 Ce-140 28740 Pb-208 34942 Bi-209 35024 Ion DeØectionsa DeØected Ions Start/ms Width/ms DeØ. 1 (12±19 amu) Cz, Nz, Oz, H2Oz, H3Oz, 0.988 0.12 DeØ. 2 (28±36 amu) N2 z, NOz, O2 z, O2Hz, 36Arz 1.302 0.09 DeØ. 3 (40 amu) 40Arz 1.480 0.1 DeØ. 4 (56 amu) ArOz 1.528 0.03 DeØ. 5 (80 amu) Ar2 z 2.068 0.03 aRemoval of plasma matrix ions. J. Anal. At. Spectrom., 1999, 14, 1807±1814 1809Application–assessment of accuracy using NIST 1643d One milliliter portions of NIST Standard Reference Material 1643d were transferred to eight different 50 ml polypropylene centrifuge tubes (Falcon1, Becton Dickinson & Co, Franklin Lakes, NJ, USA) weighed precisely and then diluted 10 times using Milli-Q water.Small volumes (50±200 ml) of PE solution were added to the centrifuge tubes to increase the initial concentrations 0.5 to 2 times.12 Following adjustments of the Ænal volume with Milli-Q water, so as to dilute the NIST 1643d sample equally for all added concentrations, the following concentrations of the different elements in the PEmix were obtained: 0, 0.158, 0.319, 1.035, 4.913, 8.457, 19.86 and 39.33 ng ml21.The resulting solutions were then analysed by ICP-TOFMS monitoring the following 33 isotopes: 7Li, 9Be, 11B, 23Na, 24Mg, 27Al, 39K, 40Ca, 51V, 50Cr, 52Cr, 53Cr, 55Mn, 54Fe, 56Fe, 58Ni, 59Co, 63Cu, 64Zn, 75As, 76Se, 77Se, 80Se, 82Se, 88Sr, 89Y, 98Mo, 107Ag, 114Cd, 121Sb, 138Ba, 205Tl and 208Pb. A 10 s integration time was used throughout and each sample was analysed with 6 replicates (repeated twice).Certain elements such as Mg, Na, Ca and K were present at mg l21 levels and deØection windows were implemented at those masses. Iron and selenium were severely affected by background from 56ArO and 80ArAr, respectively. The isotopes selected were nevertheless supposed to cover all elements certiÆed in NIST 1643d, regardless of whether instrumental parameters and deØection settings would render some of them non-quantiÆable.The concentrations were calculated in an Excel spreadsheet by dividing the intercept of the y-axis with the slope for the standard addition curve element by element. Measurements of isotope ratios The isotope ratios were measured in the analog mode at high concentrations since these measurement conditions resulted in the best precision. No dead time correction is necessary using the analog mode and obviously the contribution from fundamental noise is smaller at higher concentrations (high signal intensities).Since it was established that close proximity of deØection pulses affect the accuracy of isotope ratios, the TRIP deØections were minimized prior to measurements of isotope ratios. Two deØections for plasma matrix ions with an m/z value close to magnesium were adjusted to have as low impact on magnesium as possible at the same time making sure that the detector was not overloaded. The deØection at m/z 80 was completely removed when measuring 85Rb/87Rb.For platinum, hot plasma conditions were used. In order to obtain backgroundcorrected ratios subtraction of blanks was performed on the basis of the average signals observed at signal-free adjacent masses. For magnesium, the signal at m/z 22 was used for subtraction of the continuum background, and for rubidium and platinum, masses 86 and 189 were used, respectively. In addition, blank solutions were measured to check for presence of possible isobaric overlaps.Results and discussion Full spectral scan spectra A mass spectrum obtained in the analog mode of a solution of 100 ng ml21 mass calibration solution is shown in Fig. 2. The spectrum was obtained under normal plasma operating conditions used for multi-element analysis (Table 1). The background spectrum is similar to those obtained by quadrupole ICP-MS. The Ærst group of background peak results from oxygen and hydrogen (or water) between m/z 16 and 19. The second group is found at m/z 28, 30, 32 and 33 which is due to N2 z, NOz, O2 z, O2Hz.The last major background peaks are found at m/z 40 and 41 and result from 40Arz and 41ArHzinterfering in the determination of 39K and 40Ca. The presence of ions, such as 56ArOz and 80Ar2 z, complicates the determination of trace levels of 56Fe and 80Se. The analog channel spectrum obtained under hot plasma conditions for 100 ng ml21 PE-mix standard is shown in Fig. 3. The observed mass spectra are simple enough for unit mass resolution, which is sufÆcient for most analytical purposes. From Fig. 2, the presence of CeOz at m/z 156, YOz at 105, and UOz at 256 is more evident than signals at the doubly charged mass positions, for example, Ba2z (m/z 69), or Ce2z (m/z 70). The percentage of oxide formation CeOz to Cez is 24.7% at 100 ng ml21 in this example. It is clear that with this instrument, special attention should be given to the likelihood of oxide interferences for certain elements.Oxide formation can be largely eliminated by using an ultrasonic nebulizer with a membrane dryer if desired. Resolving power In the TOF-MS system, the resolving power depends, in part, upon the initial velocity spread of the ions along the Øight-tube axis,13 in part, on the initial spatial distribution of the ion packet along the Øight-tube axis. Here, a two-stage acceleration technique is implemented to partially correct for the initial spatial distribution of the ions and to improve resolving power.14 Moreover, an ion mirror or reØectron is employed to accomplish energy focusing of ions of the same mass with slightly different kinetic energies.An additional advantage of using the ion mirror is that it works as a photon stop that lowers the background count to 1±3 counts per s21 across each m/z. For 100 ng ml21 of Pb and Bi (as shown in Fig. 2 and 3), the resolution is w415 at 10% peak height. This resolving power is sufÆcient for most purposes and a similar performance is possible in quadrupole ICP-MS.For 1 ng ml21 of Bi the resolution increases to 615 at 10% peak height. At low m/z the resolution is signiÆcantly better than for quadrupole ICP-MS. It is noteworthy that peaks become broader, and resolution becomes worse under cool plasma conditions, for reasons that cannot be explained at the present time. Detection limits Detection limits were obtained using the ion counting mode of the detector. No dead time correction was used to improve linearity.Each peak was integrated over 19 successive bins (38 ns of mass window), and the background was determined by integration of the same number of bins from the blank solution at the same m/z position. The integration time was set to 10 s and standards and blanks were determined with 10 replicates. Net signals were taken as the difference between Fig. 2 ICP-TOFMS mass spectrum for 100 ng ml21 mass calibration solution under normal plasma operating conditions.Integration time was 1 s using full spectral scan mode of instrument (2000 bins equally distributed over the whole mass range). 1810 J. Anal. At. Spectrom., 1999, 14, 1807±1814signal and blank levels. Three times the standard deviation of the blank response at the given m/z position divided by the slope of the calibration graph for each isotope was used to calculate detection limits. Typical detection limits of ICPTOFMS are reported in Table 2 together with typical values for quadrupole ICP-MS, ionization energies, number of stable isotopes for each element and abundance of the reported isotope. In general the detection limits are somewhat worse than those reported for quadrupole ICP-MS.The lower sensitivity of approximately 10 times for ICP-TOFMS system originates from the modulation of the ion source. Due to difÆculties in the sensitive measurement of low mass analytes such as 39K, 40Ca and 56Fe under normal plasma conditions, cool plasma conditions were used for these elements. In fact, accurate determinations of 39K, 41Ca and 56Fe are virtually impossible under normal operating conditions, because of serious interferences from argon and argonrelated ions such as 56ArOz.Jiang and Houk et al. reported that operating conditions of the ICP-MS plasma could be optimized to reduce or completely eliminate poly-atomic spectral background associated with certain argide ionic species.15 By employing a lower rf power and higher nebulizer Øow rate, they observed that NOz, O2 z and H3Oz largely dominated the background spectrum, while most Ar-related ions are considerably reduced which make it possible to measure isotopes such as 39K, 40Ca and 56Fe.16 Here, the forward power and nebulizer Øow rate were optimized with a blank solution at m/z 39, 40 and 56, until the background levels were 5±10 cps.The working conditions are listed in detail in Table 1. The three isotopes 39K, 40Ca and 56Fe reach detection limits which are slightly higher than those of other isotopes.It should be pointed out that the instrumentation was not operated in a clean room, hence contamination from the outside atmosphere, acids and Milli-Q water dictates the conditions under which the detection limits were obtained. Moreover, the Milli-Q water apparatus is not placed in the same building as the ICP-TOFMS equipment. Under optimal clean room conditions slightly lower values are attainable. Signal stability Variations of the signals for 7Li, 24Mg, 39K, 40Ca, 56Fe, 133Cs and 85Rb over 4 h under cool plasma conditions are shown in Fig. 4. Long term variations over 4 h for 45Sc, 59Co, 89Y, 115In, 138Ba, 140Ce, 207Pb and 209Bi under normal plasma conditions are shown in Fig. 5. The data collection started 1 h after the plasma was ignited. The working conditions used are as listed in Table 1. The relative standard deviations and correlation coefÆcients following linear regression of the data displayed in Fig. 4 and 5 are given in Tables 3 and 4, respectively. It is clear for the cool plasma that no drift is present over 4 h. On the other hand for 45Sc, 59Co, 89Y, 115In, 138Ba, 140Ce, 207Pb and 209Bi under normal plasma conditions a clear downward drift is present, as shown in Table 4 and Fig. 5. The values for the correlation coefÆcient, r, from 20.895 to 20.948, indicate a time-correlated dependence conÆrming downward drift.It is noteworthy that at the beginning of the measurement when the signals were clearly decreasing, the short term drift over 30 min is only 1.1%RSDfor 208Pb to 4.1% RSD for 59Co, which is better than literature values reported using quadrupole ICP-MS.11,20 Long term drift for ratios of different elements–internal standardization The ions in ICP-TOFMS are sampled from the plasma at the same instant in time, and extracted synchronously from the primary ion beam, so ratioing of the signals from two different isotopes should theoretically overcome transport efÆciency changes from the peristaltic pump, nebulizer Øuctuations and Øuctuations in the plasma tail.Raw signal Øuctuations thus track each other, as can be seen in Fig. 5, and ratioing cancels out Øicker noise sources. Obviously, fundamental noise (shot noise) cannot be overcome.12 The uncertainty, or variation, of the collected signal is nevertheless largely determined by Poisson statistics. By accumulating more signal (higher concentration) the standard deviation can be reduced further in a predictable way. For w30 observations the data should be Fig. 3 ICP-TOFMS mass spectrum for 100 ng ml21 PE-mix standard solution under hot plasma conditions, integration time was 1 s using full spectral scan mode of instrument. Fig. 4 Signal stability over 4 h under cool plasma operating conditions using ion counting mode. Concentration of elements, 4 ng ml21 and a 10 s integration time was used.Fig. 5 Signal stability over 4 h under normal plasma operating conditions using ion counting mode. Concentration of elements, 1 ng ml21 and a 10 s integration time was used. J. Anal. At. Spectrom., 1999, 14, 1807±1814 1811Table 2 Detection limits (3s criterion, pg ml21) using axial ICP-TOFMS and literature values for quadrupole ICP-MS.17 For ICP-TOFMS working conditions see Table 1 Isotopes Detection limits of ICP-TOFMS Detection limits of Quad-ICP-MS17 1st Ionization energy18/eV Number of stable isotopes19 Abundance of major isotope19(%) ICP-TOFMS working conditionsc 7Li 0.5 v1 5.392 2 92.41 C 9Be 23 v1 9.322 1 100 H 11B 165 10±100 8.298 2 80.1 H 23Na 5 10±100 5.139 1 100 C 24Mg 5 10±100 7.646 3 78.99 N 27Al 9 10±100 5.986 1 100 N 28Si 2005 100±1000 8.151 3 92.23 H 39K 6 100±1000a 4.341 3 93.26 C 40Ca 29 100±1000a 6.113 6 96.94 C 45Sc 26 10±100 6.54 1 100 H 48Ti 28 10±100 6.82 5 73.72 H 51V 11 1±10 6.74 2 99.75 H 52Cr 55 1±10 6.766 4 83.79 H 55Mn 2 1±10 7.435 1 100 C 56Fe 36 10±100a 7.870 4 91.75 C 58Ni 15 1±10 7.635 5 68.08 N 59Co 17 v1 7.86 1 100 H 63Cu 4 1±10 7.726 2 69.17 C 64Zn 57 1±10 9.394 5 48.63 H 69Ga 6 v1 5.999 2 60.11 N 74Ge 36 10±100 7.889 5 36.28 H 75As 53 1±10 9.81 1 100 H 82Seb 641 10±100 9.752 6 8.73b H 85Rb 2 v1 4.177 2 72.17 C 88Sr 4 v1 5.695 4 82.58 H 89Y 3 v1 6.38 1 100 H 90Zr 3 1±10 6.84 5 51.45 H 98Mo 19 1±10 7.099 7 24.13 H 103Rh 4 1±10 7.46 1 100 H 102Ru 16 1±10 7.37 7 31.55 H 106Pd 12 1±10 8.34 6 27.33 H 107Ag 7 1±10 7.576 2 51.84 H 114Cd 7 1±10 8.993 8 28.73 H 115In 1 v1 5.786 2 95.71 C 120Sn 21 1±10 7.334 10 32.58 H 121Sb 14 1±10 8.641 2 57.21 H 130Te 31 10±100 9.009 8 34.08 H 133Cs 1 v1 3.894 1 100 C 138Ba 4 v1 5.212 7 71.70 H 139La 3 v1 5.577 2 99.91 H 140Ce 2 v1 5.47 4 88.45 H 141Pr 2 v1 5.42 1 100 H 142Nd 7 v1 5.49 7 27.25 H 152Sm 7 v1 5.63 7 26.75 H 153Eu 4 v1 5.67 2 52.19 H 158Gd 11 v1 6.14 7 24.84 H 159Tb 2 v1 5.85 1 100 H 164Dy 8 v1 5.93 7 28.18 H 165Ho 2 v1 6.02 1 100 H 166Er 5 v1 6.10 6 33.61 H 169Tm 2 v1 6.18 1 100 H 174Yb 4 v1 6.254 7 31.83 H 175Lu 2 v1 5.426 2 97.41 H 180Hf 7 1±10 7.0 6 35.08 H 184W 15 1±10 7.98 5 30.64 H 191Ir 2 1±10 9.1 2 62.7 H 196Pt 16 1±10 9.0 6 33.83 H 197Au 9 1±10 9.225 1 100 H 202Hg 6 v1 10.437 7 29.86 H 205Tl 2 v1 6.108 2 70.48 H 208Pb 4 v1 7.416 4 52.4 N 209Bi 3 v1 7.289 1 100 H 232Th 5 v1 – 1 100 H 238U 5 v1 – 3 99.27 H aIt is not clear from Ref. 17 under which plasma operation conditions the detection limits were obtained for K, Ca and Fe; the values reported suggest that cool plasma was not used.bIt should be noted that 82Se is not the most abundant isotope in contrast to all the other isotopes mentioned above. cN, H, and C are abbreviations of normal-, hot- and cool plasma. 1812 J. Anal. At. Spectrom., 1999, 14, 1807±1814normally distributed around the mean as outlined by Hooton et al.21 The following holds true for data that are normally distributed: 68.3% of the observations are within x°1s, 95.5% within x°2s and 99.7% within 3s. As can be seen from Table 5, the data generally shows a normal distribution demonstrating the high stability of isotope ratio measurements.Here, elements with similar chemical properties, ionization energies or atomic mass were measured as ratios under hot plasma conditions, as to outline the stability of internal standardization. The results can be directly compared with the raw signals in Table 3 and 4.Repeatability and reproducibility Under normal plasma conditions using the working conditions listed in Table 1, a 10 ng ml21 mass calibration solution was used to investigate the repeatability and reproducibility of the system. Employing a 10 s integration time and 5 replicates, the signals for 138Ba, 209Bi, 140Ce, 59Co, 115In, 208Pb, 45Sc and 89Y were monitored, and the repeatability was calculated from 30 discontinuous determinations, nebulizing a blank solution between each measurement for 2 min.The measurements were repeated after 24 h using the working conditions used the previous day and re-optimized monitoring m/z 115. Another 30 discontinuous measurements were performed, and the resulting signals together with the 30 previous measurements were used to calculate the reproducibility, which is shown in Table 6. The measurements were initiated following 2 h of instrumental warm-up. The time spent for 30 discontinuous measurements was about 1.5 h.It can be seen from the Table 6 that for the measured isotopes, the instrument gives very good repeatability and acceptable reproducibility. Analysis of NIST 1643d The results reported for the NIST standard reference material 1643d are generally in agreement with the certiÆed concentrations (Table 7). Out of 26 certiÆed elements only 18 were useful for determinations in the instrumental and analytical conditions used. Since uncertainties both in the slope of the regression line and the intercept of the Y-axis are present, a check of the robustness for the regression data is appropriate.In this example the data obtained for Tl is used. It should be noted that Tl had the worst correlation coefÆcient for all concentrations reported in Table 7, (r2w0.998), the others were better than this. Using the lower 95% conÆdence interval of the slope and intercept obtained using the data PAK-Excel data analysis tool, a concentration of 7.46 ng ml21 is obtained and on the high side 8.03 ng ml21.With a certiÆed concentration of 7.28°0.25 ng ml21 for Tl, the regression data are robust enough to ensure that a result of reasonable quality is maintained. The basic idea of standard additions is thus to make a standard out of the sample and thereby compensate for matrix effects.12 It should be noted that the method of standard addition is less precise than interpolation techniques.22 Measurement of isotope ratios Precisions for seven isotope ratios are reported in Table 8.Even though the deØections were minimized, the accuracy for 25Mg/24Mg, and 26Mg/24Mg is still affected to a rather high degree. Mass bias, given as deviation in % from the true ratio, has been found to be high in the low mass range, 29.2% for 26Mg/24Mg and approximately 1% in the high mass range, 194Pt/195Pt. We will report in detail on this effect in another paper. The precision for isotope ratios of magnesium and rubidium is better than 0.1% under normal working conditions.The slightly deteriorated precision for platinum-ratios possibly results from the relatively poorer response even using a concentration of 200 ng ml21. A higher concentration could not be used since 194Pt, 195Pt and 196Pt were not properly resolved under such conditions. Contrary to observations made with quadrupole ICP-MS,23,24 memory effects were not signiÆcant. Using 192Pt which is the least abundant isotope, a RSD of 0.74% for 192Pt/195Pt could be obtained.In this circumstance the relative contribution from fundamental noise to the 192Pt signal is high due to the low signal intensity, thus resulting in a higher RSD. A more detailed study of accuracy, stability of mass bias and precision for isotope ratios is currently performed in our laboratory. It can be concluded that the precision for isotope ratios using axial ICP-TOFMS is approximately ten times better than ratios measured by quadrupole ICP-MS.25±27 It should also be underlined that this high precision is obtainable for a large number of isotope Table 3 Signal stability for cool plasma, concentration is 4 ng ml21 RSD (%) Isotope 30 min 4 h Correlation coefÆcient r, 4 h 7Li 0.6 1.0 20.295 23Na 0.7 1.0 20.628 24Mg 0.8 1.7 20.186 39K 0.7 2.0 0.071 40Ca 1.8 2.1 0.204 56Fe 2.1 4.3 0.406 85Rb 1.2 1.7 20.427 115In 1.6 2.3 0.640 133Cs 1.2 1.1 0.689 Table 4 Signal stability for normal plasma, concentration is 1 ng ml21 Isotope RSD (%) Correlation coefÆcient r, 4 h 30 min 4 h 45Sc 3.8 11.4 20.947 59Co 4.1 12.8 20.942 89Y 2.5 11.2 20.943 115In 1.9 9.4 20.948 138Ba 1.6 8.7 20.895 140Ce 2.5 10.5 20.933 208Pb 1.1 6.9 20.924 209Bi 1.3 6.7 20.911 Table 5 Signal stability for measurements of intensity ratios over 4 h under normal plasma conditions (concentration is 1 ng ml21) Intensity ratio 138Bi/115In 140Ce/115In 59Co/89Y 24Mg/45Sc 208Pb/209Bi Mean intensity, x 1.0119 1.0383 0.4770 0.3017 0.7233 Standard deviation, s 0.0086 0.0140 0.0091 0.0051 0.0047 % RSD 0.85 1.35 1.91 1.70 0.65 Distribution of measurements Mean°1s 74.1 75.4 75.9 74.1 73.2 Mean°2s 96.4 96.0 94.6 95.1 96.9 Mean°3s 100.0 100.0 100.0 100.0 100.0 J.Anal. At. Spectrom., 1999, 14, 1807±1814 1813ratios in the same measurement. When using quadrupole ICPMS one is limited to one isotope pair at a time if the best possible precision is desired. Conclusions Axial ICP-TOFMS is highly competitive to quadrupole ICPMS despite somewhat lower sensitivity and higher detection limits.Extremely rapid data acquisition makes it possible to obtain qualitative and quantitative information for 66 isotopes in a very short time. Simultaneous ion extraction likewise results in high precision for isotope ratios. Oxide formation is however somewhat more prominent than for quadrupole ICPMS which deserves special attention when measuring certain elements. Accuracy has been demonstrated through the analysis of a standard reference material.Acknowledgements The authors wish to thank Dr. M. Berglund (IRMM, Geel, Belgium) for help with isotope ratio measurements. Dr Janos Fucsko (LECO) is acknowledged for supplying mixed standards of a large number of elements. The Renaissance LECO ICPTOFMS was purchased through funds provided by the Belgian National Lottery, and the work is Ænally supported by FWO, Brussels, Belgium. H.E. also gratefully acknowledges a visiting postdoctoral fellowship from FWO.The LECO Corporation, St. Joseph, MI, USA is acknowledged for supplying Fig. 1. References 1 R. S. Houk, V. A. Fassel, G. D. Flesch, H. J. Svec, A. L. Gray and C. E. Taylor, Anal. Chem., 1980, 52, 2283. 2 A. R. Date and A. L. Gray, Analyst, 1981, 106, 1255. 3 P. P. Mahoney, S. J. Ray and G. M. Hieftje, Appl. Spectrosc., 1997, 51, 16A. 4 D. P. Myers and G. M. Hieftje, Microchem. J., 1993, 48, 259. 5 D. P. Myers, G. Li, P. Yang and G. M. Hieftje, J. Am. Soc.Mass Spectrom., 1995, 6, 400. 6 D. P. Myers, G. Li, P. Yang and G. M. Hieftje, J. Am. Soc. Mass Spectrom., 1995, 6, 411. 7 D. P. Myers, G. Li, P. Yang and G. M. Hieftje, J. Am. Soc. Mass Spectrom., 1995, 6, 920. 8 P. P. Mahoney, S. J. Ray, G. Li and G. M. Hieftje, Anal. Chem., 1999, 71, 1378. 9 P. P. Mahoney, G. Li and G. M. Hieftje, J. Anal. At. Spectrom., 1996, 11, 401. 10 H. Emteborg, X. Tian and F. C. Adams, J. Anal. At. Spectrom., 1999, 14, 1567. 11 G. H. Vickers, D.A. Wilson and G. M. Hieftje, J. Anal. At. Spectrom., 1989, 4, 749. 12 J. D. Ingle and S. R. Crouch, Spectrochemical Analysis, Prentice Hall, Englewood Cliffs, NJ, USA, ISBN 0-13-826876-2. 13 R. J. Cotter, Anal. Chem., 1992, 64, 1027. 14 W. C. Wiley and I. H. McLaren, Rev. Sci. Instrum., 1955, 26, 1150. 15 S. J. Jiang, R. S. Houk and M. A. Stevens, Anal. Chem., 1988, 60, 1217. 16 S. D. Tanner, J. Anal. At. Spectrom., 1995, 10, 905. 17 E. H. Evans, J. J. Giglio, T. M. Castillano and J.A. Caruso, Inductively Coupled and Microwave Induced Plasma Sources for Mass Spectrometry, The Royal Society of Chemistry, Cambridge, UK, 1995, ch. 3, pp. 35. 18 Chemical Rubber Company Handbook of Chemistry and Physics, 64th edition, CRC Inc., Boca Raton, Fl, USA, 1984. 19 K. J. R. Rosman and P. D. P. Taylor, Pure & Appl. Chem., 1998, 70(1), 217. 20 J. J. Thompson and R. S. Houk, Appl. Spectrosc., 1987, 41, 801. 21 K. A. H. Hooton and M. L. Parsons, Anal. Chem., 1973, 45, 2218. 22 J. C. Miller and J. N. Miller, Statistics for Analytical Chemistry, Ellis Horwood Limited, NY, USA, ch. 5, p. 117. 23 F. Vanhaecke, M. van Holderbeke, L. Moens and R. Dams, J. Anal. At. Spectrom., 1996, 11, 543. 24 L. Moens, F. Vanhaecke, J. Riondato and R. Dams, J. Anal. At. Spectrom., 1995, 10, 569. 25 G. P. Russ and J. M. Bazan, Spectrochim. Acta, PartB, 1987, 42, 49. 26 H. P. Longerich, B. J. Fryer and D. F. Strong, Spectrochim,. Acta, Part B, 1987, 42, 39. 27 B. T. G. Ting and M. Janghorbani, J. Anal. At. Spectrom., 1988, 3, 325. Paper 9/06762K Table 6 Repeatability (within run precision) and reproducibility (between run precision), n~30 for repeatability sets and n~60 for reproducibility data 138Ba 209Bi 140Ce 59Co 115In 208Pb 45Sc 89Y Day 1 Average intensity 7415 9775 8522 1492 7090 6596 1231 4481 Repeatability (%) 2.1 1.4 2.2 3.6 2.2 1.6 2.7 2.6 Day 2 Average intensity 8088 9644 9436 1920 7859 6471 1460 5261 Repeatability (%) 1.3 2.4 1.2 1.7 1.1 2.4 1.0 1.1 Day 1 and Day 2 Average intensity 7752 9710 8979 1706 7475 6533 1346 4873 Reproducibility (%) 4.7 2.1 5.4 12.9 5.5 2.2 8.8 8.2 Table 7 Comparison of certiÆed concentrations and concentrations measured using axial ICP-TOFMS of trace elements in NIST1643d. Spread is given as 95% conÆdence intervals for n~6 Concentration/ng ml21 Element Measured CertiÆed Li 16.37°0.34 16.5°0.55 Be 13.08°0.75 12.53°0.28 As 54.1°0.91 56.02°0.73 Zn 70.25°0.88 72.48°0.65 Cu 20.7°0.22 20.5°3.8 Tl 7.77°0.05 7.28°0.25 Cd 6.63°0.18 6.47°0.37 Pb 18.51°0.16 18.15°0.64 Mn 36.61°0.35 37.66°0.83 Co 24.27°0.20 25.00°0.59 Cr 18.46°0.44 18.53°0.20 Sb 53.75°0.62 54.1°1.1 Al 122.1°1.54 127.6°3.5 B 141.9°3.6 144.8°5.2 Mo 106.8°0.78 112.9°1.7 Ni 53.1°0.67 58.1°2.7 Table 8 Mass bias and precision of isotope ratios for isotopic reference materials measured using axial ICP-TOFMS (30 s integration time, 10 replicates). Mass bias is given as deviation in % from certiÆed ratio Isotope ratio CertiÆed ratio Measured ratio Mass bias (%) RSD (%) 25Mg/24Mg 0.1266 0.1213 24.19 0.07 26Mg/24Mg 0.1393 0.1265 29.19 0.09 85Rb/87Rb 2.5932 2.5307 22.41 0.07 192Pt/195Pt 0.02315 0.02389 3.20 0.7 194Pt/195Pt 0.9731 0.9584 21.51 0.1 196Pt/195Pt 0.7465 0.7549 1.13 0.2 198Pt/195Pt 0.2178 0.2230 2.39 0.2 1814 J. Anal. At. Spectrom., 1999, 14, 1807±1814