Studies on the quantitative analysis of trace elements in single SiC crystals using laser ablation-ICP-MS{ Erwin Hoffmann,*a Christian Lu»dke,a Jochen Skole,a Heike Stephanowitzb and Gu»nther Wagnerc aInstitut fu»r Spektrochemie und angewandte Spektroskopie (ISAS), Institutsteil Berlin, Rudower Chaussee 5 D-12489 Berlin, Germany bGesellschaft zur Fo»rderung angewandter Optik, Optoelektronik, Quantenoptik und Spektroskopie e.V., Rudower Chaussee 5 D-12489 Berlin, Germany cInstitut fu»r Kristallzu»chtung, Rudower Chaussee 6 D-12489 Berlin, Germany Received 14th June 1999, Accepted 25th August 1999 A technique has been developed for the simultaneous determination of the trace elements Al, Cu, Mn, V, Ti and Fe in single SiC crystals by laser ablation-ICP-MS.The wavelength of 1064 nm of an Nd:YAG laser operating in the `free-running' mode is used for ablation, because in the Q-switched mode the signal-to-noise ratio was too low to obtain the analytical reproducibility required.Sample preparation is simple, as only puriÆcation of the crystal surface with HF is necessary. To prepare calibration standards, multi-element solutions were added to pure SiC powder. The powder was dried, homogenized and pressed to pellets using carbon powder as binding material. The veriÆcation of the calibration was carried out with powdered SiC reference materials. Acceptable recovery rates of between 95 and 110% were obtained. The limits of detection were between 361029 g g21 (V) and 461028 g g21 (Cu).Relative standard deviations measured at trace element concentrations of about 161026 g g21 were less than 10% when the intensities of 10 craters each with 500 laser shots were averaged. Silicon carbide (SiC) forms colourless crystals of hexagonal or rhombic structure. SiC turns glistening-green to blue-black when other atoms are incorporated into the crystal lattice. A number of modiÆcations exist which differ from each other by the structure formed while the crystal is growing.1 SiC crystals promise special applications in the electronics industry because of their semiconductor properties and, for these purposes, they are required to be of very high purity.In order to guarantee the purity of this material, it is necessary to determine quantitatively the concentrations of a number of trace elements, such as Al, V, Cu, Fe, Mn and Ti. Detection limits should not exceed 161027 g g21. A variety of instrumental techniques have been applied to determine trace impurities in SiC powder.2,3 These include instrumental neutron activation analysis (INAA),4 X-ray Øuorescence spectrometry (TXRF),5,6 analysis by ICP-OES4,7 or ICP-MS8 subsequent to sample dissolution without and with separation of the analyte elements from the matrix and direct analysis by slurry atomization ICP-OES,4 slurry atomization ICP-MS9 and laser ablation-ICP-MS.2 Procedures with sample dissolution suffer from high blank signals because, in most cases, fuming sulfuric acid is used.Therefore the detection power for environmental elements such as Al, Fe and Cu is poor.10 Slurry atomization of SiC powders combined with ICP-OES or ICP-MS is more advantageous. However, the technique is limited because powder particles with a diameter larger than 10 mm can be neither transported by a carrier gas nor completely evaporated in an ICP.11 Slurry ICP-OES has detection limits for the elements mentioned above in the range from 361027 g g21 for Ti to 2.161027 g g21 for Al.12 For most elements, the TXRF method has detection limits higher than those of optical atomic spectrometry.13 However, the repeatability of the determinations is quite satisfactory.TXRF is suited for trace element determination in bulk material when matrix separation and element enrichment are possible or when the concentrations of the analytical elements are in the upper 1026 g g21 range. In this case, the technique is advantageous because it is fast and sampling is easy.SiC crystals are extremely difÆcult to dissolve quantitatively for solution analysis. Equally, it is difÆcult to powder SiC crystals to the small particle size and narrow size distribution necessary for slurry sampling. These procedures are not only time consuming, but bear an increased contamination risk. Because of this, a technique is required that is capable of analysing the SiC crystals without using a complicated sample pretreatment. Laser ablation linked to ICP-MS (LA-ICP-MS) is in principle suited to solve the analytical problem of trace element determination in solids such as SiC crystals.14 The aim of this work was to develop a technique for the quantitative analysis of trace elements of analytical interest in single SiC crystals using LA-ICP-MS.It is planned to use the technique for the process control of SiC crystal production. Additionally, the results of this study are intended to demonstrate the potential of LA-ICP-MS as a quantitative analytical technique for the determination of impurity elements in materials such as SiC. Unfortunately, there are no certiÆed SiC reference materials which are suitable for use as calibration standards for trace element determinations.However, a powder of satisfactory purity is available which can be spiked with analyte elements. {Dedicated to Prof. Dr. D. Klockow on the occasion of his 65th birthday. J.Anal. At. Spectrom., 1999, 14, 1679±1684 1679 This Journal is # The Royal Society of Chemistry 1999Experimental Instrumentation An Elan 6000 ICP mass spectrometer (Perkin-Elmer SCIEX Instruments, Toronto, Canada) was used for this work. The system incorporates a 40 MHz free-running radiofrequency generator as the ionization source coupled with an adiabatic plasma sampling interface.15 A Model 320 laser sampler (Perkin-Elmer SCIEX Instruments) was used for sampling into the ICP.The system is based on an Nd:YAG laser, running at 1064 nm. This can be operated in either `freerunning' or `Q-switched' mode. Additionally, equipment for higher harmonic generation (532 nm, 355 nm, 266 nm) and wavelength separation is installed.16 Inside the sample chamber used throughout this study is a small moveable sample stage with a mounting for single SiC crystals. The ablated material was transferred from the laser sampler to the Elan 6000 using 1 m of 5 mm id PVC tubing as used by Gray.17 Preparation of calibration standards To compensate for matrix effects, it is necessary to match the standards to the sample material.Consequently, a puriÆed SiC powder (concentration of the analyte elements lower than the required detection limits mentioned above) was used for the preparation of the calibration standards; multi-element solution standards were added. The wet powder was dried and homogenized in an SiC mortar. Materials in a powdered state cannot be ablated by a laser beam, as the shock wave of the laser pulse scatters the powder particles. It is therefore necessary to immobilize the material, which is possible by use of a binder,14 by pressing to a pellet or by both.The latter procedure was selected in this study. Pellets were prepared by mixing the doped SiC powder with homogeneous carbon powder (weight ratio 3 : 1 respectively) as binding agent, followed by mechanical shaking. Approximately 3 g of the mixture was pressed in a 20 mm diameter pellet disc under a pressure of approximately 8 bar for 5 min.A special sample preparation was not necessary. Handling was carefully carried out under clean bench conditions to protect the sample surface from contamination. Samples and the SiC powder used for the preparation of the standards were received from the Institut fu» r Kristallzu» chtung, 12489 Berlin, Germany. SiC reference materials and reagents were purchased from the following providers: (a) Standard MRC 780-1 from IRSID, B.P. 320, 57214 Maizieres-Le�s-Metz Cedex, France; (b) NIST standard reference material 112b from National Bureau of Standards, Gaithersburg, USA; and (c) SiC F400 from ESK, 52428 Ju» lich, Germany and HF and HNO3 from Merck KgaA, 64293 Darmstadt, Germany. Time-resolved signal studies Initial studies showed that t sensitivity of the LA-ICP-MS instrument was adequate for the determination of the analyte elements over the required concentration range.Therefore, a special optimization of the sensitivity was not necessary. However, the precision of the analytical signals was insufÆcient when the laser was operated in the Q-switched mode as it is usually used. Time-resolved laser ablation signals of the analyte elements were investigated with the aim to improve the reproducibility. Measurements were made at various power settings and laser wavelengths. Figs. 1±3 show signal versus time plots for single craters at m/z (mass to charge ratio) 63 (63Cu) and m/z 65 (65Cu).The sample was an SiC crystal with about 161026 g g21 Cu concentration. Figs. 1 and 2 show the results obtained with 1064 nm pulses for 0.23 J and 161022 J, respectively. For the results in Fig. 3, the fourth harmonic at 266 nm with 1.461023 J pulse energy was used. It was found that the ion number ratio of m/z 63 to m/z 65 differs from the isotope ratio of 63Cu to 65Cu by more than a factor of two, and that all the time-resolved laser ablation signals have a number of ion intensity spikes which result in much higher signal deviations than we are used to determining with solution nebulization.It was found that the intensity spikes at m/z 63 are not correlated to those at m/z 65 and, therefore, they cannot be caused by Cu ions, as also becomes apparent from the comparisons shown in Figs. 1±3. This conØicts with the explanation given in the literature that intensity spikes are a result of unstable plasma conditions which are caused by ablated particles of different diameter passing into the plasma.14 When the laser is operated in the free-running mode, spikes are not observed.In Fig. 4, the signal versus time plot of a signal in the free-running mode is shown (1064 nm, 0.038 J pulse energy, 10 Hz pulse frequency). The ion number ratio of m/z 63 to m/z 65 (2.22°0.06) was found to be in good agreement with the isotope ratio of 63Cu to 65Cu (2.247).The reproducibility was adequate for the trace element determination required. Fig. 1 Ion intensity of 63Cu and 65Cu as a function of the ablation time (laser wavelength, 1064 nm; Q-switched mode; pulse frequency, 10 Hz; pulse energy, 0.23 J). Fig. 2 Ion intensity of 63Cu and 65Cu as a function of the ablation time (laser wavelength, 1064 nm; Q-switched mode; pulse frequency, 10 Hz; pulse energy, 161022 J). Fig. 3 Ion intensity of 63Cu and 65Cu as a function of the ablation time (laser wavelength, 266 nm; Q-switched mode; pulse frequency, 10 Hz; pulse energy, 0.0014 J). 1680 J. Anal. At. Spectrom., 1999, 14, 1679±1684Although it has been reported that ion intensity spikes can interfere with analyte signals, the processes leading to this phenomenon are not well known.18,19 Therefore, further investigations were performed to attempt to understand the signal Øuctuations of LA-ICP-MS using SiC as target material. Fig. 5 shows the distribution of the time-resolved ablation signals at m/z 65 (Q-switched: 1064 nm, 0.18 J, 10 Hz, about 161026 g g21 Cu, 30 s measuring time, 12 measurements; free running: 0.26 J, the other parameters are the same as for the Qswitched mode). The x-axis represents the ranges of the signal height.The following ranges were selected: 10±20; º90±100; 100±200; º900±1000; 1000±2000; º9000±10 000 ions per 40 ms. The number of measured signals in each range was normalized (divided by the number of the Ærst range) so that the graphs of the Q-switched mode and the free-running mode could be shown in the same Ægure.The Q-switched graph has two maxima, indicating two groups of signals, Qs I and Qs II, which must be distinguished. We found that the signals of the Qs II group cause the low reproducibility measured in the initial investigations. In contrast, the reproducibility of the signals of Qs I is high and comparable to that of solution nebulization. The distribution of the signal heights in the freerunning mode is similar to that of the Qs I group as shown in Fig. 5. The two signal ensembles of the Q-switched laser mode were also found when the ion intensities of the other analyte isotopes were measured. The signals of the Qs II group can be easily separated from those of the Qs I group by a statistical test of outlying observation.20 Fig. 6 shows the distribution (in %) of the number of the Qs II signals as a function of the mass to charge ratio.Spikes at mass to charge ratios larger than m/z 85 were not observed. To clarify the origin of the Qs I and Qs II ensembles, ion number ratios were calculated. The ratio of Qs I at m/z 63 to Qs I at m/z 65 was found to be 2.20°0.06, which is near the theoretical value of the isotope ratio of 63Cu to 65Cu. The corresponding ratio of Qs II was 0.75°0.13. We can conclude from this that the Qs II group in Fig. 5 originates from electrically charged particles (clusters) with the same charge to mass ratio as the 65Cu isotope.The time-resolved studies provide a possible explanation of the noisy nature of the laser ablation signals in the Q-switched mode. It can be supposed that particles broken off by the shock of the short laser pulse and transported by the carrier gas into the inductively coupled plasma have such a size that they are not vaporized but only ionized. Then the ionized clusters are transferred into the mass analyser and detected as ions, thus interfering with the analytical signals according to their charge to mass ratios.The ablation process of a free-running pulse is different from the Q-switched mode. It takes more time [Q-switched: (5± 10)61029 s; free-running: more than 161026 s], and consequently thermal processes are dominant during the ablation, as the melted rim of the crater in Fig. 7(a) conÆrms. Obviously, the ablated material is more homogeneously dispersed and is completely vaporized in the inductively coupled plasma.Fig. 7(b) shows a picture of a crater generated by one Qswitched pulse. The crater wall shows the layer structure of the material which was ablated without melting. Instrumental operating conditions As has been described above, it was found that the signals obtained from laser pulses in the free-running mode gave a better reproducibility and reliability of the results for the required analytical elements than the signals from Q-switched laser pulses.Consequently, the free-running mode of laser operation was used for the analytical procedure. The maximum signal-to-noise ratio was obtained if the laser was set at a pulse energy of 0.23 J and sharply focused at the surface of the sample. Although the pulse reproducibility of the laser was good, the amount of material ablated varied considerably from crater to crater. The precision of the analytical results is, therefore, Fig. 4 Ion intensity of 63Cu and 65Cu as a function of the ablation time (laser wavelength, 1064 nm; free-running mode; pulse frequency, 10 Hz; pulse energy, 0.038 J; SiC crystal; about 161026 g g21 Cu).Fig. 5 Number of signals (normalized) as a function of the signal groups at m/z 65: –, Q-switched mode; - - -, free-running mode. Fig. 6 Number of type Qs II signals as a function of the charge to mass ratio (laser wavelength, 1064 nm; pulse frequency, 10 Hz; Q-switched mode; pulse energy, 0.18 J; number of pulses, 300; SiC crystal).J. Anal. At. Spectrom., 1999, 14, 1679±1684 1681greatly improved by use of internal standardization. 29Si as well as 30Si can be used as internal standards for the determination of trace elements in SiC. They are of sufÆciently low abundance (4.7% for 29Si, 3.1% for 30Si) so that the detector is not overloaded. 12C and 13C have to be excluded as internal standards because carbon is used as binder in the calibration standard pellets. Suitable analyte isotopes could be found for the required elements.Only Al has a small background ion interference caused by the 28Si isotope. The interference is a result of abundance sensitivity overlap, because the concentration ratio between Si and Al is extremely high. A summary of the operating conditions and the isotopes selected for the analyte elements is shown in Table 1. Results Calibration, limit of detection, precision Calibration curves with least-squares regression correlation coefÆcients of better than 0.99 were obtained for all the elements analysed.The slope value determines the sensitivity of the element determination. The relative standard deviations of the slope values are listed in Table 2. Additionally, the relative signal standard deviations (analyte concentration of about 161026 g g21) for one crater ablation are given in Table 2. Table 2 also contains the limits of detection (LOD) of the elements of interest. Non-spiked SiC powder was used for the determination of the detection limits.After mixing the SiC with carbon powder (see `Preparation of calibration standards') and pressing to a pellet, we carried out 11 measurements and calculated the detection limits using the 3s criterion for the background deviations of the signals. To verify the calibration, the SiC powder reference materials mentioned above were prepared in the same way as the calibration standards and analysed. The results are shown and compared with the certiÆed values in Table 3.The averaged recovery rates deÆned as the ratio `concentration found' to `concentration given' in per cent are also shown in Table 3. The concentrations found agree acceptably with the certiÆed concentrations. Repeatability Measurement precision and sample homogeneity are two factors affecting the repeatability of the laser ablation technique. Laser ablation creates only small craters with vaporized sample masses in the mg to mg range depending on the pulse energy and pulse number used.The analytical results, therefore, are only representative of the small amounts of material ablated by the laser beam. Consequently, one of the advantages of laser sampling–the capability of spatially resolved analysis–simultaneously turns out to be a disadvantage when bulk representative concentration values are Fig. 7 Laser crater in SiC: (a) free-running mode (1064 nm; single pulse; 0.25 J; crater diameter, about 0.4 mm); (b) Q-switched mode (1064 nm; single pulse; 0.25 J; crater diameter, about 0.4 mm).Table 1 Instrumental and analytical parameters Laser Wavelength 1064 nm Pulse energy 230 mJ Mode Free running Frequency 10 Hz Ablation time per crater 50 s Crater diameter About 200 mm Crater depth About 200 mm Ablated mass About 20 mg Number of craters 10 ICP Electric power 1 kW Transport gas 1 l min21 Dwell time 100 ms Replicates 40 Resolution High Sample Preparation Surface puriÆcation with HF Calibration SiC powder, doped with multi-element solution standards, bound with carbon powder, pressed to pellets (8 bar) Validation SiC reference materials Isotopes 48Ti, 51V, 55Mn, 56Fe, 60Ni, 63Cu, 27Al Internal standard 30Si, 29Si Table 2 Detection limits (based on 3s criterion), relative standard deviation of the slope of the calibration curve (eight measurements at each concentration, seven concentration values for each element), variation of the concentration from crater to crater in a single SiC crystal, relative standard deviation of the single crater ablation Relative standard deviations (%) Concentration variation from crater to crater (%) Isotope Detection limit/mg g21 Slope Single crater 27Al 0.020 1.2 2.8 17 49Ti 0.025 2.9 3.8 25 51V 0.003 1.4 3.2 20 55Mn 0.030 4.1 4.8 30 57Fe 0.020 1.8 3.1 22 63Cu 0.040 1.3 3.0 24 1682 J.Anal. At. Spectrom., 1999, 14, 1679±1684required. To assess the accuracy of the method when applied to real samples, the homogeneity of a typical industrial plant SiC crystal was studied.In the Ærst step, the surface impurity of the sample was studied. SiC crystals grow at such a high temperature that the elements which have to be determined are in the vapour phase during the whole process. After crystallization, the crystal begins to cool down. Condensation on the surface and diffusion into the crystal takes place. Therefore, the surface and outer crystal layers are expected to show increased trace element content.To determine the concentration±depth proÆle, the laser was defocused so that a larger area of the sample was ablated. Then the signals of 10 laser pulses were measured and the element concentrations were determined with the help of the calibration curves. Concentration±depth proÆles of Cu before and after puriÆcation with HF are shown in Fig. 8, indicating that an acid-cleaned surface is necessary to obtain reliable bulk analytical results. Further studies were focused on the question of whether signiÆcant differences can be observed between the signal precision of the one-crater ablation and the repeatability of replicate measurements carried out at different sites (distance between two laser spots, 1 mm) of the sample.The calculation of the correlation coefÆcients showed the statistical independence of the analytical signals for one-crater ablation.21 The relative standard deviations for the signals of 40 measured craters are given in Table 2.It was found that the concentrations of 10 craters have to be averaged to obtain repeatabilities within the 10% level for the bulk analysis of SiC crystals. Conclusion and outlook A rapid technique, involving only puriÆcation of the surface by HF as sample pretreatment, has been developed for the determination of trace elements in single SiC crystals by LAICP- MS. If the laser is operated in the Q-switched mode, the signal deviations of the elements investigated are high, owing to the spiky nature of the ablation signals.However, in the freerunning mode, the relative standard deviations are comparable to those usually obtained by solution methods and adequate for the purpose of this work. Limits of detection are less than 561028 g g21 for all the analytical elements investigated. Our studies have indicated that the low signal-to-noise ratio in the Q-switched mode is due to large ablation particles, which are transported into the plasma by the carrier gas but do not vaporize.In the free-running laser mode, particles introduced into the plasma are small enough to vaporize. It is assumed that the large particles created at the beginning of ablation are vaporized due to absorption of subsequent laser spikes in the free-running mode. However, the free-running mode has two analytical disadvantages: the radiation intensity of the Nd:YAG wavelength is usually too low to generate higher harmonics, and fractional volatilization of analyte elements with low boiling points may take place.To use the advantages of the Q-switched mode, the ablation can be combined with a second laser pulse which arrives at the sample surface with a short time delay. The second pulse vaporizes the large particles which were broken off the sample by the shock wave.22 The second laser beam can be generated by splitting the original beam with a mirror or a prism. A time difference between the arrival of both laser beams at the sample surface can be obtained if the distances which the beams have to run are different.The difference between the beam paths must be about 3 m, which can be realized with the use of several beam reØections at mirrors. The ability to obtain spatial information on element distribution is one of the advantages of the laser system. This aspect can also be important for bulk analysis, which has been demonstrated by investigating the sample homogeneity.Acknowledgements The Ænancial support by the Senatsverwaltung fu» r Wissenschaft, Forschung und Kultur des Landes Berlin and the Bundesministerium fu» r Bildung und Forschung is gratefully acknowledged. H. Stephanowitz also thanks the `Gesellschaft zur Fo»rderung angewandter Optik, Optoelektronik, Quantenelektronik und Spektroskopie e.V.' for a grant. References 1 G. P. Fritz and E. Matern, in Carbosilanes, Springer, Berlin, 1986. 2 H. Nickel and J. A. C. Broekaert, Fresenius' J.Anal. Chem., 1999, 363, 145. 3 J. A. C. Broekaert, R. Brandt, F. Leis, C. Pilger, D. Pollmann, P. Tscho» pel and G. To» lg, J. Anal. At. Spectrom., 1994, 9, 1063. 4 T. Graule, A. von Bohlen, J. A. C. Broekaert, E. Grallath, R. Klockenka»mper, P. Tscho» pel and G. To» lg, Fresenius' J. Anal. Chem., 1989, 335, 637. 5 A. von Bohlen, R. Eller, R. Klockenka»mper and G. To» lg, Anal. Chem., 1987, 59, 2551. 6 M. Franek and V. Krivan, Fresenius' J. Anal. Chem., 1992, 342, 118. 7 G. Zaray, F. Leis, T. Kantor, J. Hassler and G. To» lg, Fresenius' J. Anal. Chem., 1993, 346, 1042. 8 C. Pilger, F. Leis, P. Tscho» pel, J. A. C. Broekaert and G. To» lg, Fresenius' J. Anal. Chem., 1995, 351, 110. 9 F. Kohl, N. Jakubowski, R. Brandt, C. Pilger and J. A. C. Broekaert, Fresenius' J. Anal. Chem., 1997, 359, 317. Table 3 CertiÆed concentrations and measured concentrations with reference to the appropriate reference material MRC 780-1 concentration/mg g21 NIST 112b concentration/mg g21 F400 concentration/mg g21 Isotope Certif.Measured Certif. Measured Certif. Measured Recovery rate averaged (%) 27Al 18 600 18 520°500 4200a 4600°500 1200 1262°100 105°7 49Ti – – 230a 220°15 280 265°19 95°7 51V – – – – 1100 1095°60 100°5 55Mn 290 293°6 – – 34 35°2 102°4 57Fe 13 040 14 500°1000 1300 1450°150 3100 3360°250 110°8 63Cu – – – – 90 94°3 105°6 aGiven but non-certiÆed concentration. Fig. 8 Cu concentration as a function of the number of pulses (depth proÆle) (free-running mode; pulse energy, 0.23 J; wavelength, 1064 nm; SiC crystal; crater diameter, about 0.5 mm; maximum crater depth estimated at 0.2 mm). J. Anal. At. Spectrom., 1999, 14, 1679±1684 168310 T. Graule, PhD Thesis, University of Dortmund, Germany, 1988. 11 B. Raeymackers, T. Graule, J. A. C. Broekaert, F. Adams and P. Tscho» pel, Spectrochim. Acta, Part B, 1988, 43, 923. 12 L. E. Ebdon and M. R. Cave, Analyst, 1982, 107, 172. 13 L. E. Ebdon and J. R. Wilkinson, J. Anal. At. Spectrom., 1987, 2, 325. 14 S. A. Darke, S. E. Long, C. J. Pickford and J. F. Tyson, Fresenius' J. Anal. Chem., 1990, 337, 284. 15 D. J. Douglas and J. B. French, Spectrochim. Acta, Part B, 1986, 41, 197. 16 E. Hoffmann, C. Lu»dke and H. Stephanowitz, Fresenius' J. Anal. Chem., 1996, 355, 900. 17 A. L. Gray, Analyst, 1985, 110, 551. 18 S. T. G. Anderson, R. V. D. Robert and H. N. Farrer, J. Anal. At. Spectrom., 1992, 7, 1195. 19 P. M. Outrige, W. Doherty and D. C. Gregoire, Spectrochim. Acta, Part B, 1996, 51, 1451. 20 U. Graf and H. J. Henning, Mitteilungsblatt Mathematische Statistik, 1952, 4, 1. 21 H. Scholze, E. Hoffmann, C. Lu»dke and A. Platalla, Fresenius' J. Anal. Chem., 1996, 355, 892. 22 J. Uebbing, J. Brust, W. Sdorra, F. Leis and K. Niemax, Appl. Spectrosc., 1991, 45, 1419. Paper 9/04734D 1684 J. Anal. At. Spectrom., 1999, 14, 1679±1684