Factors Affecting the Quantification of Boron in SiO2 and Si by Sputtered Neutral Mass Spectrometryкод для вставкиСкачать
SURFACE AND INTERFACE ANALYSIS, VOL. 24, 371-374 (1996) Factors Affecting the Quantification of Boron in SiOz and Si by Sputtered Neutral Mass Spectrometry Michael L. Wise, Netzer Moriya and Stephen W. Downey* Lucent Technologies,Bell Laboratories, 600 Mountain Ave., Murray Hill, NJ 07974, USA Boron implanted in SiOz and Si is characterized using sputtered neutral mass spectrometry (SNMS). Nonresonant, ultrahigh-intensity postionization finds that a fraction of boron not sputtered as neutral atoms from SiO, and Si is partially present in the form of BO and BSi molecules. Total boron detection is matrix sensitive. A deficit in the boron measured in SiO, suggests that boron may go partially undetected due to the substantial production of secondary ions during sputtering. INTRODUCTION EXPERIMENTAL The measurement of the distribution of boron in thin layers of Si and SiO, continues to be of great importance to microelectronic technologies. As device sizes shrink below 0.5 pn, the determination of boron concentrations and gradients by sputter-based techniques like secondar ion mass spectrometry (SIMS) in and through 100 f l sub-layer of SiO, is exceedingly difficult due to stringent depth resolution requirements and strong matrix effects. The accurate determination of boron in the thin SiO, gate oxides is effectively precluded. Postionization detection of neutrals often reduces matrix effects because the atomization via sputtering and laser ionization are separate processes. As a result, many times only changes in relative sputter rates have any substantial effect on measured signals during depth profiling through thin layers. The application of ultrahigh-intensity (> loi4 W ern-,) laser pulses allows non-resonant ionization of all sputtered species.'*' Consequently, in many cases, the sputtered species are detected with nearly uniform sensitivity. Constant detection efficiency of all sputtered species, including many molecules, can result in more accurate and standardless determination of dopant profiles in thin layers of differing materials. However, if the assumption that the majority of sputtered species are neutral is not valid, then the manner in which postionization is implemented is critical for obtaining satisfactory results in thinlayer depth profiling. In this work, ultrahigh-intensity postionization mass spectrometry is used to profile ion-implanted "B in SiO, and Si. Three different sputtering ions are utilized. Perhaps surprisingly, total boron detection is found to be matrix sensitive. Those factors creating this postionization 'matrix effect' and possible solutions to this circumstance are discussed. The postionization experiments are performed with a modified magnetic-sector SIMS instrument in which the output of the laser system is directed into the ionization region of the mass spectrometer.' The laser is a regeneratively amplified Ti :sapphire system based on an argonion pumped, self-mode-locked oscillator and a frequency-doubled YAG pumped amplifier. Output pulses at 800 nm contain up to 1 mJ of energy in a pulse width of 100-150 fs with a focused diameter of -50 pm. The repetition rate is -1 kHz. The pulse width is measured by a scanning Michelson interferometer-type:autocorrelator with 10 fs resolution. The average intensity present in each pulse can be in excess of loi4 W cm-,. The ion source is a mass-filtered duoplasmatron capable of producing O,', Xe' and Ar'. The ion energy on target is adjustable from 2 to 10 keV, but for these experiments it is 3 keV. At this energy, the incident angle of the primary ion beam on the sample is estimated to be 60" off-normal. The operating current is 1-2.5 pA and the beam is -100 pm in diameter. The primary ion beam is pulsed for postionization. Laser pulses entering the chamber are synchronized with a pulsed, primary ion beam for sample sputtering. An ion pulse precedes a laser pulse by -2 ps and is -0.5 ps in duration. All the timing delays are set to maximize the laser beam/sample atom cloud overlap in space and time for the greatest detected signal while minimizing sample erosion during data acquisition. Depth profiles consist of alternating cycles of continuous, rastered primary ion patterns for material removal and pulsed sputtering for static analysis of the center of the crater bottom. The SIMS data are obtained in the same fashion while the laser is blocked, with the sample bias lowered 300 V to select those ions produced at the sample surface to pass through the energy analyzer. The sample is loo0 A of thermally grown SiO, on Si(l00) which is "B-implanted (26 keV, 1 x 10'' cm-') such that the total area density in each material is the * Author to whom correspondence should be addressed. CCC 0142-2421/96/06037 1-04 0 1996 by John Wiley & Sons, Ltd Received 9 August 1995 Accepted 22 January 1996 M. L. WISE, N. MORIYA AND S. W. DOWNEY 372 same, 5 x 1014 anT2. This was confirmed by the "B[p, a] nuclear reaction analytical technique3 on the layered sample and on a portion of the sample with the oxide film removed by HF e t ~ h i n gThe . ~ peak of the implant, located at the interface, has a boron concentration of 9 x 1019~ m - ~ . can be observed while sputtering with O,+ across the interface, while a three orders of magnitude drop in signal can be seen on the Xe+- and Ar+-sputtered samples. These changes are the classic SIMS matrix effect. Predictably, the presence of oxygen in the matrix considerably enhances positive secondary ion yield^.^ The matrix effect for O2 sputtering while transitioning from SiO, to Si is far less severe than that for Xe' and Ar+ sputtering due to the greater oxygen content of the Si substrate. A postionization depth profile of the same sample using Xe+ sputtering is shown in Fig. 2. A depth profile using Ar + sputtering (not shown) is almost identical. Because of the non-selective, universal detection of sputtered neutrals, a variety of species are tracked with one laser wavelength (800 nm). A survey of the possible sputtered species from the two substrates indicates that, in addition to IIB+ and "Si', 2eSi160+(mIe = 441, + RESULTS A N D DISCUSSION Three conventional SIMS profiles of the "B-implanted 1O00 A SiO, on Si sample using Xe+, Ar' and 02+ sputtering are shown in Fig. 1. We note that no significant sputtering rate changes between these two materials are observed when using these primary ions. An order of magnitude drop in the secondary ion signal Depth (A) Figure 1. Normalized "B SlMS depth profiles of O,+-,Ar+- and Xe+-sputtered 1000 A SiO, on Si implanted with B (26 keV, 1 x I O l 6 cm-*). The peak of the implant is at the interface where a sharp decrease in SlMS signal is found. 1 o6 t o5 - 6 ._ 10' (I) c 'a s ..-E 1 o3 L B 1 o2 10 ' 0 1000 2000 Depth 3000 (A) Figure 2. Postionization depth profile of the l1 B-implanted I O O O A SiO, on Si sample obtained using Xe+ sputtering. The average intensity of the focused laser beam is 4 x 1O'* W cm-*. QUANTIFICATION OF BORON IN SiO, AND Si BY SNMS 11B160' (m/e = 27) and 11B28Si' (m/e = 39) are detectable. Interestingly, the B+ profle displays a sudden increase at the SiO,/Si interface, opposite to the matrix effect encountered in the SIMS results. Integration of the prolile indicates that 30% of the atomic boron is detected in the SiO, with 70% detected in the Si. Molecular BO' and BSi' also represent a large fraction of the total detected neutral boron under these sputtering conditions. We cannot prove that BSi and BO are detected with the same efficiency as atomic boron. It is suspected that the molecules are detected somewhat more efficiently because of their expected lower velocities. The molecular percentages stated are probably an upper limit to their sputtered fractions. Ignoring any surface peaks, which contain hydrocarbon isobars and environmental boron, BO ' accounts for 40% of the total detected boron in the oxide while BSi' accounts for 30% of the total detected boron in the Si. Nonetheless, adding the signals from BO' and BSi' to the B' signal does not significantly change the disproportionality in the measured boron in the two materials. Less total neutral boron is sputtered from the SiO, than Si. ,' sputtering Postionization of the sample with 0 detects the same sputtered species as those obtained with noble gas ion sputtering (Fig. 3). Likewise, a disproportionality in measured boron in the two materials is observed but is less severe. Integrating the atomic signal indicates that 40% of the atomic boron is detected in the SiO,, as compared to 60% detected in the Si substrate. Molecular contributions are different with O2 sputtering than with the noble gas sputtering; BO' accounts for 30% of the total boron detected in the oxide, while BSi+ and BO+ account for only 20% of the total boron detected in the Si. Again, accounting for the boron detected as BO' and BSi+ does not significantly alter the deficit in detected boron in the SO,. The magnitude of the sputtered neutral signals from Si are noted to be lower when using 0 2 + as compared to Ar+ or Xe'. Many factors can affect the signals obtained with a postionization technique. Ionization efficiency, sputter + 373 yield, angular and velocity distributions of sputtered neutrals, molecular contributions, instrumental biases, as well as the production of secondary ions during sputtering can all affect the measured signals. Unfortunately, these different effects are extremely difficult to distinguish from one another when looking only at relative signals. In these experiments, a difference in the postionization detection of boron in SiO, and Si is apparent. Assuming that the molecular species in the two materials are being detected uniformly and with similar efficiency to the atomic boron, molecular contributions are not able to compensate for the disproportionality. Likewise, differences in ionization efficiency, instrumental biases and angular/velocity distributions are unlikely to significantly affect the measurement of the same species from two different matrices. One factor which may be especially important in this specific chemical system is the production of secondary ions. During the postionization experiments the sample is biased such that secondary ions produced by the sputtering are not allowed through the energy spectrum analyzer, nor are they subjected to the laser light, because the high extraction field sweeps them out of the laser focal spot during the delay optimized for the neutrals. For this reason, no secondary ions are counted during the postionization experiments. Correspondingly, significant production of secondary ions detracts from the sputtered neutral species which are probed by the laser. Figure 1 indicates that the "B' SIMS signal has a large matrix effect at the interface of the two substrates, especially when using the noble gas ions for sputtering. This large decrease in secondary ion yield when going from the SiO, to the Si may explain the inverse trend in the postionization signals due to the suppressionn of sputtered neutrals by secondary ion production. The effect of secondary ion matrix effects on postionization signals may be surprising. Nonetheless, under certain sputtering conditions, a large fraction of the total sputtered species may be secondary ions. In these experiments, SIMS signals from the SiO, are always greater than the postionization signals, even with the 1 o2 1 0' 1 1 28 + 8 8 0 500 1000 2000 1500 Depth 2500 3000 (A) Figure 3. Postionization depth profile of the "B-implanted 1OOOA SiO, on Si sample obtained usingo,' sputtering. M. L. WISE, N. MORIYA AND S. W. DOWNEY 314 short (500 ns) sputtering pulses. The extraction parameters of the SIMS and postionization are different enough to prevent an effective comparison of absolute ion us. neutral yields, but the observation supports the contention that ions are a large fraction of the sputtered material in the SiO, . Absolute secondary ion yields (secondary ion/incident ion) have been measured for a few metallic species using secondary ion mass spectrometers with known secondary ion transmissions6-' and using other techniques.' The absolute secondary ion yield for Si+ from oxygen-covered silicon using 3 keV Ar sputtering was determined by Benninghoven to be O X 6 , ' Without the adsorbed oxygen, this yield drops to 8.4 x 10-3. The current Si+ postionization results mirror these absolute secondary ion yields. Figure 2 indicates that the Si+ postionization signal changes by a factor of 4.1 when going from SiO, to Si. Because SiO, and Si have the same sputter rates with Xe+ sputtering, a factor of 2.1 increase in Si' when transitioning to the Si substrate is predicted based on the greater density of elemental Si in the Si matrix. The additional factor of two observed in the postionization signal can be explained by a secondary ion yield in the SiO, approximately equal to the yield of neutrals. Upon transitioning to the Si substrate, this secondary ion yield is reduced by a factor of 1000 and the corresponding yield of neutrals is doubled. While the absolute secondary ion yield of B+ has not been measured, boron and silicon in general display similar relative sensitivity factors (RSFs) when sputtered from SiO, and Si with O,'.' For this reason, it may be reasonable to assume that the secondary ion fraction of B + from SiO, is high. A secondary ion yield of 30-50% in SiO, (B' plus BO') would explain the undetected neutral boron in the SiO,. As indicated by the SIMS results in Fig. 1, this secondary ion fraction would drop to 5% or less in the Si substrate with the corresponding increase of the boron postionization signal. The matrix effect observed here would not be too much different for + any other type of SNMS technique because of the inherent secondary ion production in this chemical system. It is noteworthy that this matrix pair may be the most severe with respect to this effect. Sputtered neutral mass spectroscopy can be successfully applied to the analysis of many other material systems where secondary ion emission is low (<10%).The importance of SiO, and Si to the microelectronics industry warrants this special investigation. To alleviate this matrix effect caused by secondary ions, an alternative ion extraction scheme, like that commonly used in time-of-flight mass spectrometry, would be beneficial. If sputtering is performed in a fieldfree region, then the neutrals and ions move roughly with the same velocities. A pulsed extraction field coincident with the laser pulse would accept all ions from the ionizing volume, regardless of their origin, thus summing the secondary ions with the postionized neutrals. In this scenario, secondary ions as well as neutrals are subjected to the intense laser field, increasing the possible contribution from multiply charged ions to the mass spectrum. Time-of-fight detector with adequate mass resolution to separate isobars would be ideal for this type of measurement. An intense peak anticipated from the sputtered matrix constituents would have to be handled by rapid mass blanking or detector gain adjustment. CONCLUSIONS Postionization techniques have determined how boron sputtered from SiO, and Si fractionates into atoms, ions and molecules. Using Xe', Ar' and 0,' sputtering, a significant amount of boron was sputtered from SiO, and Si as BO and BSi molecules. The shape of the boron profiles is affected by secondary ion emission, which in the present experimental geometry subtracts from the sputtered neutrals available for postionization. REFERENCES 1. C. H. Becker and J. S. Hovis, J . Vac. Sci. Technol.A 12, 2352 (1994). 2. M. L. Wise, A. B. Emerson and S. W. Downey, Anal. Chem. 67, 4033 (1995). 3. L. C. Feldman and J. W. Mayer, Fundamentals of Surface and Thin Film Analysis. North-Holland, New York (1 986). 4. A. 6. Emerson, Y. Ma, M. L. Wise, M . L. Green and S. W. Downey, J. Vac. Sci. Technol. J. Vac. Sci. Technol., B14, 301 (1996). 5. R. G. Wilson, F. A. Stevie and C. W. Magee, Secondary Ion Mass Spectrometry: a practical Handbook for Depth Profiling and Bulk IrnpurityAnalysis.Wiley, New York (1989). 6. A. Benninghoven, Surf. Sci. 53,596 (1975). 7. A. Benninghoven, F. G. Riidenauer and H. W. Werner, Secondary Ion Mass Spectrometry: Basic Concepts, Instrumental Aspects, Applications and Trends. Wiley, New York (1 987). 8. P. Williams, I. S. T. Tsong and S. Tsuji, Nucl. Instrum. Methods 170,591 (1980). 9. W. Husinsky, J. Vac. Sci. Technol. B 3,1546 (1985).