Anal. Chem. 2008, 80, 3261-3269
Formation of High-Mass Cluster Ions from Compound Semiconductors Using Time-of-Flight Secondary Ion Mass Spectrometry with Cluster Primary Ions Robyn E. Goacher,† Hong Luo,‡ and Joseph A. Gardella, Jr.*,†
Departments of Chemistry and Physics, University at Buffalo, State University of New York, Buffalo, New York 14260
The detection of high-mass, nonstoichiometric, GaxAsy and InxPy secondary ion clusters using time-of-flight secondary ion mass spectrometry is reported for the first time. The GaxAsy and InxPy clusters are detected in both positive and negative ion spectra and extend to masses of at least 6000 dalton (Da). Consecutive clusters differ by the addition of one gallium (indium) atom. This leads to nonstoichiometric clusters at high mass (i.e., Ga15As3 at 1270 Da) which are metastable above a critical mass. The relative secondary ion yields of high-mass GaxAsy clusters detected using several primary ion sources (Cs+, Bi+, Bi3+, Bi32+, Bi52+, C60+, and C602+) are compared. The relative secondary ion yield of high-mass GaxAsy clusters is significantly enhanced by the use of cluster primary ions and the best relative secondary ion yield is obtained using Bi3+ primary ions. An application of the high-mass GaxAsy clusters is presented, in which these clusters are utilized to distinguish between contaminant levels of Ga and bulk GaAs structure in a depth profile of a MnAs/GaAs heterojunction. These results illustrate improved analysis of inorganic materials using cluster primary ions and break the paradigm of stoichiometric secondary cluster ion formation for SIMS of inorganic compounds. There has been much recent attention to novel results in secondary ion mass spectrometry (SIMS) using cluster primary ion sources.1 In particular, cluster primary ions are known to significantly increase the secondary ion yield of high-mass organic and biological molecular ions and molecular fragment ions.1 However, the use of cluster primary ions for the analysis of inorganic species has received less attention in the literature.2 In this paper we report the first detection of high-mass secondary ion clusters from gallium arsenide and indium phosphide using time-of-flight SIMS (ToF-SIMS). As is the case with the analysis * To whom correspondence should be addressed. E-mail: gardella@ acsu.buffalo.edu. Fax: 716-645-6963. † Department of Chemistry. ‡ Department of Physics. (1) Winograd, N. Anal. Chem. 2005, 77, 142A-149A. (2) Van Stipdonk, M. J.; Santiago, V.; Schweikert, E. A. Secondary Ion Mass Spec., SIMS XII, Proc. Int. Conf. 12th 2000, 291-294. 10.1021/ac7024656 CCC: $40.75 Published on Web 03/22/2008
© 2008 American Chemical Society
of organic species, the secondary ion yield of these inorganic clusters is enhanced using cluster primary ions. The paradigm for understanding the formation of high-mass secondary ion clusters from inorganic compounds was established by Campana and Colton in the 1980s with their SIMS analysis of alkali halide salts.3,4 In these papers, monatomic Ar and Xe primary ions were utilized and stoichiometric secondary ion clusters were detected from low mass to above 25 000 dalton (Da).3,4 In contrast to the work by Campana et al., the GaxAsy and InxPy secondary ion clusters reported in this paper are nonstoichiometric. GaxAsy and InxPy clusters have previously been studied using laser vaporization/mass spectrometry techniques by R. E. Smalley and D. M. Neumark’s research groups. Each of these groups has detected clusters containing up to ∼30 atoms with stoichiometric and nonstoichiometric ratios.5-8 Additionally, intensity oscillations between clusters having even and odd numbers of atoms were observed and explained in terms of closed and open electronic shells for even and odd neutral clusters, respectively.5,6,9,10 In this paper, we report analysis of the chemical composition of the GaxAsy and InxPy secondary ion clusters detected using ToFSIMS, including a comparison between these clusters and those reported by the Smalley and Neumark research groups. We also discuss the differences between positive and negative ion spectra, the complexation of primary ions with the samples, and the detection of metastable ion decay products in the spectra. Furthermore, a comparison of the efficiency of several primary ion species is presented in terms of the relative secondary ion yields of the high-mass GaxAsy clusters. Finally, an analytical application of the high-mass GaxAsy cluster ions is presented in a (3) Campana, J. E.; Colton, R. J.; Wyatt, J. R.; Bateman, R. H.; Green, B. N. Appl. Spectrosc. 1984, 38, 430-432. (4) Barlak, T. M.; Campana, J. E.; Wyatt, J. R.; Colton, R. J. J. Phys. Chem. 1983, 87, 3441-3445. (5) Asmis, K. R.; Taylor, T. R.; Neumark, D. M., Chem. Phys. Lett. 1999, 308, 347-354. (6) O’Brien, S. C.; Liu, Y.; Zhang, Q.; Heath, J. R.; Tittel, F. K.; Curl, R. F.; Smalley, R. E. J. Chem. Phys. 1986, 84, 4074-4079. (7) Liu, Y.; Zhang, Q. L.; Tittel, F. K.; Curl, R. F.; Smalley, R. E. J. Chem. Phys. 1986, 85, 7434-7441. (8) Zhang, Q. L.; Liu, Y.; Curl, R. F.; Tittel, F. K.; Smalley, R. E. J. Chem. Phys. 1988, 88, 1670-1677. (9) Lou, L.; Wang, L.; Chibante, L. P. F.; Laaksonen, R. T.; Nordlander, P.; Smalley, R. E. J. Chem. Phys. 1991, 94, 8015-8020. (10) Lou, L.; Nordlander, P.; Smalley, R. E. J. Chem. Phys. 1992, 97, 18581864.
Analytical Chemistry, Vol. 80, No. 9, May 1, 2008 3261
sputter depth profile of manganese arsenide and gallium arsenide, which is a material combination that may have importance for spintronics applications.11,12 EXPERIMENTAL SECTION Samples. Semi-insulating, undoped GaAs(100) single-crystal wafers were obtained from American Xtal Technology (Fremont, CA). InP(001) was the substrate of an InP/InAs quantum dot sample grown at the National Research Council Laboratories in Ottawa, Ontario. This sample was etched post-growth in hydrochloric acid to expose the substrate. The MnAs/GaAs sample was grown by molecular beam epitaxy (MBE) using a Riber 32 system. From the bottom up, this sample consists of a semi-insulating, undoped GaAs(100) substrate, an ∼120 nm layer of MBE-grown GaAs, and an ∼170 nm layer of MBE-grown MnAs, which was grown with a substrate temperature of 250 °C. All samples were analyzed as received. ToF-SIMS. A ToF-SIMS 5-100 instrument manufactured by IonTof Gmbh (Muenster, Germany) was used in this work. This instrument is equipped with a bismuth liquid metal ion gun and a dual-source column having cesium and C60 ion sources. With this instrument, each ion source may be used as the analytical primary ion source during acquisition under static conditions. For depth profiles, Cs or C60 must be used to erode the material while analysis is limited to the bismuth source. However, bismuth may be used in a DC mode to sputter-clean a region of the sample before static analysis. The analyzer of the ToF-SIMS 5 is designed so that the extractor voltage and drift path energy can be independently set by the user. In the normal analyzer mode, the extractor voltage and drift path energy are both equal to 2000 V. In the high-massresolution mode, the extractor voltage is 2000 V but the ion energy is reduced to 1000 V before the ions enter the field-free region. This allows for efficient collection of ions from the near-surface region of the sample but also for longer flight time and higher mass resolution. In both analyzer modes, a post-acceleration voltage of 10 kV is applied. The normal analyzer mode was used for all acquisitions in this work except for spectra acquired to investigate the change in flight time on the detection of metastable ion decay products. Semi-insulating GaAs was analyzed in the static mode using Cs+ (1.02 pA), Bi+ (0.11 pA), Bi3+ (0.41 pA), Bi32+ (0.15 pA), Bi52+ (0.06 pA), C60+ (0.03 pA), and C602+ (0.01 pA) primary ions. InP was also analyzed in the static mode using Bi3+ (0.17 pA). Bismuth ions were accelerated to 25 kV, and Cs and C60 ions were accelerated to 10 kV. Due to their double charge, Bi32+ and Bi52+ ions had a kinetic energy of 50 kV, and C602+ had a kinetic energy of 20 kV. Before each static analysis, a 2.5 × 105 µm2 region of the sample was sputter-cleaned for 30-120 s to remove organic contamination and the native oxide. The same species used to sputter-clean the samples was used for analysis (i.e., DC Bi sputter cleaning for Bi analysis). Additionally, a separate location on the sample was analyzed with each ion source. These precautions were taken to prevent the formation of artifacts in the spectra from (11) Dvakonov, M. I. Los Alamos Natl. Lab., Prepr. Arch., Condens. Matter 2004, 1-10. (12) Ramsteiner, M.; Hao, H. Y.; Kawaharazuka, A.; Zhu, H. J.; Kastner, M.; Hey, R.; Daweritz, L; Grahn, H. T.; Ploog, K. H.; Phys. Rev. B: Condens. Matter Mater. Phys. 2002, 66, 081304/1-081304/4.
3262
Analytical Chemistry, Vol. 80, No. 9, May 1, 2008
previous sputtering with different species. Analysis was restricted to the center of the sputter-cleaned area of the sample. Positive and negative secondary ion spectra were obtained using a cycle time of 400 µs to yield an upper mass limit of ∼15 000 Da for the normal analyzer mode. High-mass-resolution settings for the analyzer were used with Bi32+ primary ions to study GaAs. The pressure during analysis was better than 6 × 10-9 mbar for GaAs analysis and 9 × 10-8 mbar for InP analysis. A depth profile of semi-insulating, undoped GaAs(100) was performed using Bi32+ (0.24 pA) as the analytical species and C602+ (0.075 nA) as the sputtering species (9.0 × 104 µm2 sputter area, 4.2 × 103 µm2 analysis area, 0.06 nm/s sputter rate). Depth profiles of MnAs/GaAs were performed using Bi32+ (0.15 pA) as the analytical species and either 1 kV Cs (0.84 nA, 4.0 × 104 µm2 sputter area, 1.9 × 103 µm2 analysis area, 0.15 nm/s sputter rate) or C602+ (0.56 nA, 4.0 × 104 µm2 sputter area, 5.2 × 103 µm2 analysis area, 0.20 nm/s sputter rate) as the sputtering species. Carbon deposition during C60 sputtering was not a problem under the conditions used.13 All depth profiles were obtained with a 200 µs cycle time in the interlaced mode, where the flight time of the secondary ions is utilized for sputtering. Sputter crater depths were measured using a KLA-Tencor P-16+ surface profilometer at the Surface Interface Ontario facility in Toronto, Canada. Data Manipulation. IonSpec software (version 4.1.0.1) was used to manipulate the mass spectra. All spectra were calibrated to at least five peaks characteristic of the sample. To calculate the relative secondary ion yields of GaxAsy and InxPy clusters, a peak window was spread over the mass range for each cluster in the spectra and the Poisson-corrected14 integrated area was divided by the integrated intensity of a reference peak. For GaAs, the reference peaks were 69Ga+ in the positive ion spectra and 69GaAs- in the negative ion spectra. For InP, the sum of 113InP+,and 115InP+,- peaks was used as the reference in both positive and negative ion spectra. Mass resolution was greater than 5400 for normal mode spectra at 115 Da. Overlapping GaxAsy clusters were deconvoluted in a stepwise manner by using the isotopic ratios of the lowest mass (most Ga-rich) peaks which did not overlap with the next cluster to calculate what percentage of the neighboring overlapping peaks was due to the first cluster. RESULTS AND DISCUSSION The positive secondary ion mass spectra of InP and GaAs appear in Figure 1. In this figure, repeating high-mass GaxAsy clusters are clearly visible over the mass range of 500-5000 Da. Since the isotopes of gallium and arsenic are separated by only 4-6 mass units, clusters that have the same number of total atoms (i.e., n ) 6; Ga4As2 and Ga3As3) overlap to form what appears to be a single cluster in the spectrum. In the InP spectrum, InxPy clusters are well separated from each other due to the large mass difference between indium and phosphorus. The most intense peak series, which is distinctly visible in Figure 1A between 2500 and 10 000 Da, is the Inx series. GaxAsy and InxPy clusters are also present at lower masses, but the patterns are obscured in Figure 1 by other low mass peaks. (13) Gillen, G.; Batteas, J.; Michaels, C. A.; Chi, P.; Small, J.; Windsor, E.; Fahey, A.; Verkouteren, J.; Kim, K. J. Appl. Surf. Sci. 2006, 252, 6521-6525. (14) Stephan, T.; Zehnpfenning, J.; Benninghoven, A. J. Vac. Sci. Technol., A. 1994, 12, 405-410.
Table 1. Composition of the GaxAsy Clusters Centered at Approximately 360 and 640 Da, Determined by Fitting Experimental Positive and Negative Mode Data to Calculated Isotopic Distribution Patterns Ga5 positive ions (%) negative ions
3.1
Ga9
Figure 1. Positive secondary ion spectra of (A) InP and (B) GaAs acquired using Bi3+ primary ions.
It should be noted that the absolute intensity of the clusters above ∼3000 Da is low (3000 Da) of the spectra. Identification and Analysis of GaxAsy Secondary Ion Clusters. The repeating high-mass clusters shown in Figure 1B are separated by the mass of a single Ga atom. These clusters have been identified by comparison with isotopic distribution patterns calculated using IonSpec software and are composed of overlapping Gan-yAsy clusters where y ranges from 0 to 5 and n increases as the overall cluster mass increases. At sufficiently high mass, the clusters become highly nonstoichiometric. For example, the cluster centered at 1480 Da is composed of Ga19As2, Ga18As3, and Ga17As4. The formation of highly nonstoichiometric GaxAsy clusters is surprising at first because a stoichiometric cluster composition is expected from the paradigm established by Colton and Campana.3,4 However, in contrast to the clusters formed from salts, which must have a balance between positive and negative charges due to ionic bonding and electrostatic attraction, there is no strict requirement for the formation of stoichiometric clusters from the covalently bonded GaAs solid, and Ga metal is a stable bulk material. The Ga-rich clusters are detected in both positive and negative ion spectra, although the secondary ion yield of high-mass clusters in the negative ion spectra is less than that in the positive ion spectra for a given primary ion dose. This can be understood to be a result of the affinity of Ga to form positive ions. Clusters in the negative spectra are shifted toward higher mass compared with those in the positive spectra, indicating that a greater fraction of the overall cluster comes from a GaxAsy cluster with higher arsenic content, such as GaxAs4, as opposed to GaxAs. Examples of the change in composition between positive and negative ion spectra appear in Table 1 and in Figure 2, where the clusters at 360 and 640 Da have been decomposed into their composite GaxAsy clusters. The shift toward higher As content in the negative spectra is more severe in the 360 Da cluster than in the 640 Da cluster. The cluster at 360 Da also has some small contributions from Ga5 in the positive mode and As5 in the negative mode, but these single-element clusters do not have significant intensity at
positive ions (%) negative ions
360 Da Ga4As Ga3As2 91.1 5.4
5.8 55.8
640 Da Ga8As Ga7As2 15.7 5.0
39.3 42.9
Ga2As3
GaAs4
As5
35.7
2.7
0.4
Ga6As3
Ga5As4
44.2 48.3
0.8 3.6
Ga4As5
higher mass. Comparison of these two clusters also illustrates the trend that as the overall mass of the cluster increases, the contribution of GaxAs becomes smaller and the contributions of GaxAs3 and GaxAs4 become greater. Another interesting characteristic of the Gan-yAsy clusters shown in Figure 1B is the oscillating pattern between consecutive Gan-yAsy clusters (visible over the mass range of 600-2000 Da), which then transitions to a pseudo-exponential intensity decay. This pattern is present in both positive and negative ion spectra and is also illustrated in Figures 6 and 7 below. This oscillating pattern is consistent with the alternating even/ odd cluster intensities detected by Smalley and co-workers where odd n clusters are more intense than even n clusters.6-10 As explained by Smalley and co-workers, odd-numbered clusters have higher electron affinities and lower ionization potentials than evennumbered clusters because of the transition from an open-shell electronic structure for neutral odd clusters to a more stable closed-shell electronic structure for singly charged odd clusters.9,10 The transition to pseudo-exponential intensity decay occurs at ∼2000 Da (28-29 atoms), which coincides well with the highest mass GaxAsy clusters discussed by Smalley and co-workers.7,8 It is possible that if Smalley and co-workers had detected ions above this mass, they would also have seen the oscillating pattern transition to a simple exponential decay pattern. However, this transition may instead be a consequence of the Ga-rich nature of the clusters detected in this work. The clusters at the transition point in the mass spectrum have compositions centered on Ga25As3 and Ga26As3, and this may be enough Ga atoms for the electronic structure of the clusters to be dominated by metalliclike behavior with a sea of electrons, as opposed to localized bound electrons in either open or closed shells. Further calculations would be required to verify this hypothesis. Analysis of InxPy Secondary Ion Clusters. The secondary ion clusters formed from InP are in many ways similar to those discussed for GaAs. This is not completely surprising since both materials are III-V compound semiconductors and both have zincblende bulk structure. The InxPy clusters detected are also In-rich and are increasingly nonstoichiometric at higher mass. The number of phosphorus atoms, y, ranges from 0 to 6 but clusters having 5 or 6 P atoms are only sporadically detected in the spectra. Due to the larger mass separation between the InxPy clusters, the InP spectra can be easily broken down into series of Py, Inx, InxP, InxP2, InxP3, and InxP4 peaks. The secondary ion intensities for Analytical Chemistry, Vol. 80, No. 9, May 1, 2008
3263
Figure 2. Plots showing the measured positive (2) and negative (9) intensities for the GaAs clusters centered at 360 Da (A) and 640 Da (B), along with the fitted intensities obtained by multiplying the percentages in Table 1 by the appropriate isotopic distribution for each component cluster. The connected lines show the overall cluster shape, and each symbol is the intensity of an individual peak. The insets show the linear correlations between measured and fitted intensities: for positive ions, y ) 1.005x + 0.0039, R2 ) 0.9999; for negative ions, y ) 0.9965x + 0.0024, R2 ) 0.9984, where y is the fitted intensity and x is the measured intensity.
the aforementioned InxPy peak series relative to InP are presented in Figure 3. For the low-intensity cluster series in Figure 3, there appears to be a step in intensity at a certain point. This is because the data points at lower mass are the result of integrating and adding individual isotopic peaks for a given cluster, and the data points at higher mass are the result of integrating the entire cluster. This change in procedure is a consequence of the fact that the low mass clusters have to be resolved from contaminant and Bi-ionized species (see last section on primary ion complexation) while these species do not extend to higher mass, and the high-mass peaks are not well resolved within a cluster. Vertical lines in Figure 3 denote the transition from individual peak integration to integrating entire clusters. These lines are included to indicate where increased intensity is partially due to increased background signal from wider integration windows. The secondary ion yield of InxPy clusters relative to InP is generally higher in the positive ion spectra than in the negative ion spectra. Like with the GaAs clusters, this is due to the affinity of In to form positive ions. The only exceptions are the InP3, In2P3, and InP4 clusters and the Py series. The Inx peak series is detected throughout the spectra, which is in contrast to the negligible intensities of Gax clusters at high masses in the GaAs spectra. InxP, InxP3, and InxP4 clusters are also detected throughout the spectra while Py clusters are not detected past P7, and InxP2 clusters are not detected past In22P2. Like the GaAs spectra, there are some regions of oscillating high and low intensity in the InP spectra. This is not the case with the metallic Inx cluster series, but for the InxPy peak series there are oscillating intensity patterns at low mass that transition to a pseudo-exponential decay pattern when the clusters contain ∼17 In atoms. Thus, the low-mass InxPy clusters detected in this work using ToF-SIMS are consistent with the oscillating intensities of even (low) and odd (high) InxPx and Inx+1Px clusters described by Neumark and co-workers.5 Additionally, there are certain “magic numbers” evident in the spectra. Enhanced intensities are detected for the In13 and P5 clusters in the negative ion spectra and the In4P and In5P2 clusters in the positive ion spectra. Particularly low intensities for In13P3 in the positive ion spectra and In13P2 in the negative ion 3264 Analytical Chemistry, Vol. 80, No. 9, May 1, 2008
spectra are also observed. A detailed analysis of why these particular InxPy clusters have such different intensities will be the subject of future work. Formation of Ga-Rich and In-Rich Clusters. To test the hypothesis that Ga-rich GaxAsy clusters may form from a Ga-rich surface oxide, a depth profile was performed on semi-insulating GaAs. The peaks due to GaAs-oxide (GaO, AsO, GaO2, and AsO2) decreased to 50% maximum intensity at a depth of 1.4 nm while the GaxAsy clusters grew to 50% maximum intensity at a depth of 2.4 nm. Thus, the Ga-rich clusters are not formed because of the presence of a surface oxide and do not appear until after sputtering beyond this oxide. At the same time that the intensity of the GaxAsy clusters increased, the intensity of the elemental Ga and As signals decreased, indicating a change in the identity of the clusters formed. Since further sputtering is required after the disappearance of the oxide layer, it is possible that the Ga-rich clusters detected in this work result from the formation of a steady-state surface composition rich in Ga atoms and that prior to the erosion of the oxide, As was not selectively sputtered. It is known that arsenic generally has a higher sputtering yield than gallium,15 which supports this explanation. Additionally, indium-enriched surfaces have been measured after argon ion sputtering,16,17 which supports the detection of In-rich clusters in the spectra of InP. It is also possible that high-mass clusters form through recombination in the sputter plume. However, computer simulations18,19 describing the preferential sputtering of the surface and the probability of intact ejection of high-mass clusters would be required to support either of these conclusions. Detection of Metastable Ion Decay Products. As the mass of the GaxAsy and InxPy clusters increase, new sets of ions are detected in both positive and negative ion spectra. These peaks (15) Yamamura, Y.; Itikawa, Y.; Itoh, N., Nagoya University Institute of Plasma Physics Report IPPJ-AM-26, 1983. (16) Pan, J. S.; Wee, A. T. S.; Huan, C. H. A.; Tan, H. S.; Tan, K. L. J. Appl. Phys. 1996, 80, 6655-6660. (17) Yu, W.; Sullivan, J. L.; Saied, S. O.; Jones, G. A. C. Nucl. Instrum. Methods Phys. Res., Sect. B 1998, 135, 250-255. (18) Garrison, B. J.; Winograd, N. Chem. Phys. Lett. 1983, 97, 381-386. (19) Garrison, B. J.; Winograd, N.; Harrison, D. E. J. Chem. Phys. 1978, 69, 1440-1444.
Figure 3. Relative secondary ion intensity variations for the InxPy cluster series in positive ([) and negative (]) ion spectra acquired using Bi3+ primary ions. Series plotted are (A) Py, (B) Inx, (C) InxP, (D) InxP2, (E) InxP3, and (F) InxP4. Vertical lines denote the transition from individual peak integration to integrating over an entire cluster. Typical error is between 1 and 30% RSD.
are product ions resulting from the unimolecular dissociation of metastable ions in the flight tube of the spectrometer20 and are shown in Figure 4 for GaAs. (20) Cooks, R. G.; Beynon, J. H.; Caprioli, R. M.; Lester, G. R. Metastable Ions; Elsevier: New York, 1973.
In the GaAs spectra, the metastable ion product peaks appear first in the negative ion spectra, beginning as lower-mass shoulders at about 500 Da and becoming distinctly separated peaks by 850 Da. In the positive ion GaAs spectra, similar metastable ion product peaks appear as shoulders around 850 Da and become separated peaks around 1100 Da. These peaks are broad because of the wide range of energies of the product ions. This is in contrast to the sharp peaks present at the correct GaxAsy mass. Once a mass of ∼1550 Da is reached, the sharp peaks for what would be Ga22-yAsy clusters are no longer detected in either positive or negative ion spectra (Figure 4F). Since the high-mass clusters in the spectra are Ga-rich and gallium preferentially forms positive ions, it is sensible that more nonstoichiometric (higher mass) clusters would be less stable as negative ions. This interpretation explains why the sharp correctmass GaxAsy peak intensities drop more quickly in the negative ion spectra (compare positive and negative spectra in Figure 4 parts D and E) and why the metastable ion decay products are detected at lower mass in the negative spectra than in the positive spectra. It is worth noting that the broad peaks associated with the decay of metastable ions are not detected when a spectrum of GaAs is collected using high-mass-resolution conditions for the analyzer. As described in the Experimental Section, these conditions are different from the normal analyzer conditions because the ions travel through the field-free region of the flight tube at 1000 V instead of 2000 V. Thus, in the high-mass resolution mode, an ion of given mass has a longer flight time and it is expected that a metastable ion will decay at an earlier position within the TOF analyzer. If this decay occurs before a critical focusing point such as the reflectron, the products of the metastable ion decay may not be focused onto the detector. This could explain the absence of these peaks in the high-mass resolution spectra. It is also noticed in the high-mass-resolution mode that the sharp GaxAsy peaks disappear at 1550 Da (Figure 5). This is the same mass where the metastable ion decay products dominate the spectra in the normal mode (Figure 4F) and may be interpreted as the critical mass above which GaxAsy clusters are not stable under the conditions of this experiment. In the spectra of InP, broad metastable ion decay product peaks are also detected. The first of these peaks appears at ∼410 Da in the negative ion spectra and at ∼610 Da in the positive ion spectra. For InP, it is also noticed that above 1600 Da, the clusters in the negative ion spectrum are dominated by broad peaks and there is significant intensity of the broad peaks in the positive ion spectrum. Thus, the formation of metastable ions at a certain mass for nonstoichiometric clusters seems to have some generality. It should be noted that in unpublished work by other members of this research group, polymer secondary ion clusters above 1600 Da exhibit sharp peak shapes while peaks due to cluster ions from the silver substrate (e.g., Ag14) exhibit low-mass shoulders similar to those in the positive ion GaAs spectrum in Figure 4B. This contrast between metal and covalently bonded polymer clusters at the same mass means that the detection of broad peaks in the GaAs and InP spectra is not due to analyzer artifacts. The assignments of the metastable decay products for GaAs and InP is part of ongoing and future work. However, since the metastable decay products are detected close to the mass of Analytical Chemistry, Vol. 80, No. 9, May 1, 2008
3265
Figure 4. Series of GaxAsy cluster spectra acquired with Bi3+ primary ions showing the increased dominance of the products of metastable ion decay as the mass increases and the greater detection of such peaks at lower mass in the negative ion spectra.
Figure 5. Comparison of normal and high-mass resolution analyzer modes for positive ion GaAs spectra acquired with Bi32+ primary ions.
Figure 6. High-mass regions of the positive GaAs spectra acquired using Bi3+ and Cs+ primary ions.
predicted GaxAsy or InxPy clusters, it is likely that decay occurs via the loss of a single atom from the parent ion.20 This would be consistent with the laser fragmentation work done on GaAs clusters by Smalley and co-workers.7,8 The metastable ion product peaks are also detected at different masses in the positive and negative ion spectra. This discrepancy cannot be attributed to a difference in mass scale calibration due to the mass agreement between the sharp GaxAsy peaks, but it may indicate a different decay mechanism for positive and negative metastable ions. Dependence of Relative Secondary Ion Yield on Primary Ion Identity. The relative secondary ion yields of GaxAsy clusters detected using several primary ions have been calculated by dividing the cluster’s intensity by the intensity of the 69Ga+ or 69GaAs- peak. The primary ions used were Cs+, Bi+, Bi +,Bi 2+, 3 3 Bi52+, C60+, and C602+. GaxAsy clusters at high mass (>500 Da) were detected with all Bi and C60 primary ions but were not detected using Cs as the primary ion. The spectra collected using Cs primary ions were composed of organic species and low-mass GaAs species such as elemental Ga and As, Ga2, and GaAs. Additional Cs-cationized clusters such as GaCs, AsCs, GaCs2, and AsCs2 were also observed. Figure 6 shows a portion of the highmass positive ion spectra obtained using Bi3+ and Cs+ primary
ions. The contrast between the repeating GaxAsy clusters in the Bi3+ spectrum and the Cs+ spectrum illustrates the inability of Cs+ to generate detectable high-mass GaxAsy clusters. The evenodd intensity alternation discussed previously is also well illustrated in Figure 6. The relative secondary ion yields for GaxAsy clusters above 500 Da are portrayed in Figure 7 as a function of the primary ion species. Each data point in Figure 7 represents the integral intensity of the GaxAsy cluster centered at the specified mass, and the figure is intended to provide a compressed view of the spectra from each primary ion species so that overall trends in the secondary ion yields may be identified. Data acquired with Cs+ primary ions is not included in Figure 7 because GaxAsy clusters were not detected above 500 Da with this primary ion source. The oscillations that appear in the Bi1 and C60 data traces above 2500 Da for positive ions and above 2000 Da for negative ions arise from low signal-to-noise ratios and are not significant. Figure 7 reveals some important trends in the performance of the primary ions utilized in this study. Cluster primary ions such as Bi3, Bi5, and C60 yield more high-mass cluster secondary ions than monatomic Bi+ primary ions. Bi3+ was the best primary ion overall for efficient generation of high-mass GaxAsy clusters,
3266 Analytical Chemistry, Vol. 80, No. 9, May 1, 2008
Figure 7. Comparison of relative positive (A) and negative (B) secondary ion yields for high-mass GaxAsy clusters dependent on the identity of the primary ion. Key: C60+, ]; C602+, 9; Bi1+, 2; Bi32+, ×; Bi3+, [; Bi52+, b. Each data point represents the integral intensity of the GaxAsy cluster centered at the specified mass. Typical error is within 1-25% RSD.
followed by Bi52+, Bi32+, C60+, and C602+ as defined by the relative secondary ion yield. The noise level for C602+ in the negative mode was high, causing the trace for C602+ to rise above that for the bismuth clusters above 3500 Da. Interestingly, the relative secondary ion yields for high-mass GaxAsy clusters are inversely related to the energy per atom of the primary ion cluster. For bismuth ions, the energy values per atom are 8333 eV for Bi3+, 10 000 eV for Bi52+, 16 667 eV for Bi32+, and 25 000 eV for Bi+, and the relative secondary ion efficiency order is Bi3+ > Bi52+ > Bi32+ > Bi+. However, the energy of the primary ion is not an absolute criterion of relative secondary ion yield since 10 000 eV Cs does not yield high-mass secondary ion clusters like Bi52+ does. This illustrates a fundamental difference in the mechanism of sputtering for monatomic and cluster primary ions. The energy per atom for C60+ was 167 eV and was 333 eV
for C602+. This continues the trend of greater relative secondary ion yield for the primary ion cluster that has less energy per atom. However, the energy per atom of the C60 clusters may simply have been too low for efficient sputtering of the inorganic material, causing secondary ion yield to suffer in comparison with clusters of bismuth. Utility of Cluster Secondary Ions in Depth Profiles of Heterojunctions. In addition to the ability of cluster primary ion ToF-SIMS to generate high-mass cluster secondary ions from inorganic compounds, it is important to recognize the utility of such clusters in SIMS analysis. To illustrate the extra information that can be gained from such cluster secondary ions, a depth profile of MnAs over GaAs is presented in Figure 8. In this depth profile, additional peaks were detected that contained carbon atoms (i.e., Ga3C) and these carbon-complexed Analytical Chemistry, Vol. 80, No. 9, May 1, 2008
3267
Figure 8. Positive ion depth profile of MnAs/GaAs using C602+ as the sputter species and Bi32+ as the analytical species. Traces denoted (C) include signal from carbon-complexed peaks as well as non-carbon-complexed peaks.
Figure 9. (A and B) BiGa7-yAsy clusters (right-hand side starred peaks) and Ga10-yAsy clusters in positive (A) and negative (B) ion spectra. (C-E) Positive ion GaAs spectra showing differences in high-mass clusters when the sample is (C) precleaned with bismuth sputtering alone, (D) when concurrent C602+ sputtering occurs in a depth profile, and (E) when concurrent Cs+ sputtering occurs in a depth profile. A gray vertical line representing the midpoint of the Ga13As3 cluster has been added to aid the eye.
peaks were added into the traces for the non-carbon-complexed peaks. The peaks from GaAs are separated into two categories: peaks due to elemental Ga and diatomic clusters such as Ga2 and GaAs and peaks due to high-mass GaxAsy(C) clusters. The MnAs peaks are also split into two groups: those due to Mn and Mn2 and those fromMnxAsy clusters. Thus, the profile compares the information provided by elemental and small cluster traces with the information provided by large clusters from both MnAs and GaAs. In the uppermost 100 nm of the profile, the GaxAsy clusters have low, nonzero intensity due to mass interference between certain MnxAsy(C) and GaxAsy peaks. Meanwhile, the elemental Ga and small Ga clusters have high near-surface intensity that drops before the interface and rises again. Both 69Ga and 71Ga isotopes follow this intensity profile, so the near-surface intensity is not due to mass interference between 71Ga and MnO. The distinct difference between the low, level signal for the GaxAsy cluster ions and the high, varying level of elemental Ga and small Ga cluster signals indicates that Ga is present in the MnAs layer 3268 Analytical Chemistry, Vol. 80, No. 9, May 1, 2008
as a contaminant. Significant intensity for GaxAsy species is only obtained upon sputtering to the bulk structure of the GaAs layer, so monitoring these species in a depth profile allows the analyst to distinguish between trace and bulk gallium. Furthermore, in this depth profile, the MnAs clusters decay to 50% intensity at a depth of 147.7 nm, which is the same depth at which Ga elemental signals rise to 50% intensity. However, the Mn elemental trace persists to 149.5 nm, indicating that there is either some diffusion of Mn into the GaAs layer or that there is sputter-induced intermixing of elemental Mn that does not affect the signal for MnAs clusters. Thus, tracking the clusters in this depth profile provides more information about the sample than if Ga and Mn alone were monitored. In this case, Ga contamination in the MnAs layer and potential Mn diffusion are revealed. Complexation of Primary Ions with the Sample. Clusters containing bismuth atoms are detected in the positive and negative GaAs spectra up to BiGa10-yAsy. These clusters are separated from the GaxAsy peaks by 0.2 Da with sufficient resolution to separate the peaks (Figure 9 A and B). Thus, the presence of these peaks
does not influence the calculations in Table 1 and Figure 2. Interestingly, because Bi is in the same group as As and has five valence electrons, the trend of high-intensity odd number clusters and low-intensity even number clusters is also evident in the BiGan-yAsy series. This means that the Gan-yAsy and BiGan-yAsy clusters for a given n value have inverted intensity oscillations because the addition of a Bi atom to a closed shell even n GaAs cluster makes an open-shell odd cluster. The percent contribution to the spectrum of the BiGan-yAsy clusters vs Gan-yAsy clusters ranges from 0.3 to 57.7%, with most contributions in the range of a few to 10 percent. Outstandingly intense BiGan-yAsy clusters are BiGa3-yAsy in the positive ion spectra and BiGa5-yAsy in the negative ion spectra. BizInxPy clusters are also detected in positive and negative ion spectra of InP. In particular, BizInx peaks are detected with z ) 1, 2, and 3 and up to 12, 16, and 8 In atoms, respectively; BizInxP peaks are detected with z ) 1 and 2 and up to 11 and 12 In atoms, respectively; and BiInxP2 peaks are detected with up to 16 In atoms. The Bi-ionized InxPy clusters have a very large range of intensities compared with that of the non-Bi-ionized clusters, ranging from 0.0002 to 160%. For all peak series, the Bi-ionized clusters make up a larger portion of the negative ion spectra than the positive ion spectra. Additionally, a higher percentage of clusters contain one Bi atom compared with two or three atoms, with three atoms being least probable. The oscillating even-odd intensity patterns are also observed in the BizInxPy series where peaks having an odd number of total atoms are more intense. There are also some even-odd oscillations in the BizInx series which are not present in the Inx series. Aside from Bi incorporation in the samples, additional complexation of the sputtering species with the sample is observed in depth profiles. In the depth profile discussed above, GaAs was sputter eroded with C602+ and analyzed with Bi32+ and additional peaks were detected in the spectrum that can be attributed to the addition of carbon atoms and hydrocarbon species to GaxAsy clusters (Figure 9D). In addition to the intense Ga3C cluster mentioned previously, clusters appear in the spectrum of the GaAs layer above 1000 Da as shoulders on the GaxAsy clusters. These shoulder clusters are ∼22-29 Da below the GaxAsy clusters and are not detected in spectra where GaAs is both sputter-cleaned and analyzed with Bi (Figure 9C). The shoulder clusters have low intensity and poor mass resolution, but an apparent 2 Da spacing between the peaks indicates that they are composed in part of Gax. Because of the low mass resolution, the peaks have yet to be firmly identified. However, because these peaks are detected only in the presence of C60 sputtering and the mass separation falls in the range of hydrocarbon fragments, the authors presently attribute these peaks to the complexation of C60 fragments with the GaAs. The shoulder peaks identified in the C602+-sputtered spectrum are not detected when the MnAs/GaAs sample is depth profiled using Cs+ as the sputtering species. However, with Cs sputtering, the intensity of the regular GaxAsy clusters, which were detected in the C60 and bismuth static spectra and the C60 depth profile,
are greatly reduced and an entirely new set of clusters appear in the spectrum. Some of these clusters have been identified as complexes of GaxAsyCsn, where n is 1, 2, or 3. These Cs-containing cluster species are so intense below 500 Da that they overshadow the unreacted GaxAsy clusters. Above 500 Da, low-intensity clusters appear ∼17 Da below the regular GaxAsy clusters. These peaks are shown in Figure 9E where the remnants of the regular GaxAsy clusters are just visible by comparison with the Bi-sputtered spectrum (Figure 9C). The exact composition of individual high-mass clusters in the Cs-sputtered GaAs spectrum has yet to be determined. However, since Cs-cationized clusters have been positively identified at lower mass, it is likely that these peaks are GaxAsyCsn clusters. It is interesting to note that the Cs-containing cluster peaks are not detected in the static spectrum where Cs is the primary analytical ion. In order for these peaks to be detected, the reaction of GaAs with Cs must be followed by desorption with an efficient cluster primary ion such as Bi32+. CONCLUSION The detection of Ga-rich and In-rich, nonstoichiometric, highmass GaxAsy and InxPy clusters represents a new formation mechanism for inorganic secondary ion clusters in SIMS, which is distinguished from the stoichiometric secondary ion cluster formation from ionic solids. Additionally, the significant enhancement of cluster secondary ion yield when cluster primary ions are used illustrates the value of using cluster primary ions in the analysis of inorganic species. The utility of the cluster secondary ions has been shown by their ability to distinguish between contaminant levels of a species and bulk structure in depth profiles of samples containing GaAs layers. This type of analysis may prove to be valuable in the analysis of semiconductor heterojunctions for the electronics industry and in depth profiles of other inorganic species. The differences in the spectra with different sputtering species (Bi, Cs, and C60) illustrate the complicating effects of reactions between the sputtering species and the sample. These reactions may severely change the spectrum, as is the case with Cs sputtering, or may create new peaks from one layer that produce unexpected mass interference with another layer, as is the case with some MnAs-hydrocarbon peaks and GaxAsy peaks in depth profiles where C60 was used. ACKNOWLEDGMENT We thank Dr. Ian Sellers at the University at Buffalo for the provision of the InP sample, Professor Troy Wood at the University at Buffalo for helpful discussions, and Dr. Rana Sodhi and SurfaceInterface Ontario at the University of Toronto for the use of their profilometer. We also gratefully acknowledge the financial support of a NSF IGERT Fellowship (DGE 0114330) to R.E.G. and a NSF Research Grant (CHE 0616916) and NSF Major Research Instrumentation Grant (CHE 0619728) to J.A.G. Received for review December 3, 2007. Accepted February 20, 2008. AC7024656
Analytical Chemistry, Vol. 80, No. 9, May 1, 2008
3269
