JOURNAL OF APPLIED PHYSICS 116, 054908 (2014) Observed damage during Argon gas cluster depth profiles of compound semiconductors Anders J. Barlow,a) Jose F. Portoles, and Peter J. Cumpson National EPSRC XPS Users’ Service (NEXUS), School of Mechanical and Systems Engineering, Newcastle University, Newcastle upon Tyne NE1 7RU, United Kingdom (Received 9 May 2014; accepted 23 July 2014; published online 6 August 2014) Argon Gas Cluster Ion Beam (GCIB) sources have become very popular in XPS and SIMS in recent years, due to the minimal chemical damage they introduce in the depth-profiling of polymer and other organic materials. These GCIB sources are therefore particularly useful for depthprofiling polymer and organic materials, but also (though more slowly) the surfaces of inorganic materials such as semiconductors, due to the lower roughness expected in cluster ion sputtering compared to that introduced by monatomic ions. We have examined experimentally a set of five compound semiconductors, cadmium telluride (CdTe), gallium arsenide (GaAs), gallium phosphide (GaP), indium arsenide (InAs), and zinc selenide (ZnSe) and a high-j dielectric material, hafnium oxide (HfO), in their response to argon cluster profiling. An experimentally determined HfO etch rate of 0.025 nm/min (3.95 10 2 amu/atom in ion) for 6 keV Ar gas clusters is used in the depth scale conversion for the profiles of the semiconductor materials. The assumption has been that, since the damage introduced into polymer materials is low, even though sputter yields are high, then there is little likelihood of damaging inorganic materials at all with cluster ions. This seems true in most cases; however, in this work, we report for the first time that this damage can in fact be very significant in the case of InAs, causing the formation of metallic indium that is readily visible C 2014 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4892097] even to the naked eye. V I. INTRODUCTION Argon Gas Cluster Ion Beam (GCIB) sources were originally developed for wafer smoothing during semiconductor processing1,2 and advanced coatings,3 and subsequently the use of these sources for analytical applications was pioneered in SIMS.4,5 They have proved to be very useful in XPS.6 These GCIB sources produce ionised argon clusters having a size distribution ranging from around 500 to 5000 atoms, which can be narrowed by mass filtering provided by the manufacturer of the source. The average kinetic energy per atom is much lower than for conventional monatomic ion sources, leading to greatly reduced damage in sputter depth-profiling of polymers. The very low damage that these cluster ions introduce has been shown5 to allow practical depth-profiling of polymer and other organic samples,7 largely preserving sample chemistry. These GCIB sources are therefore particularly useful for depth-profiling polymer and organic materials. The assumption has been that, since the damage introduced into polymer materials is low, even though sputter yields are high, then there is little likelihood of damaging inorganic materials at all with cluster ions. In this work, we report for the first time that this damage can be very significant in the case of some compound semiconductors. Excellent progress has been made over the last five years in understanding cluster-surface interactions in an analytical context.8–10 In our laboratory, we have performed depth a) Author to whom correspondence should be addressed. Electronic mail: [email protected]. 0021-8979/2014/116(5)/054908/5/$30.00 profiles of many different organic layers and have characterised their behaviour during GCIB etching. Recently, we have published measurements of total sputter yields from a large number of organic and inorganic materials exposed to cluster ions11,12 indicating that the total sputter yield from inorganics is significantly less than that for organics over the range of cluster energies of practical significance in XPS and SIMS. This is of course beneficial when depth-profiling a polymer, for example, where it is desirable to have a high sputter yield, with a damage depth much less than the information depth of the analytical technique. With such low sputter yields, it is unlikely that a GCIB would be used for depth profiling through thick (i.e., more than a few tens of nanometres) inorganic layers, due to the prohibitively long times that would be required. However, these sources were originally developed for semiconductor processing by reducing surface roughness, so may be expected to be useful in depth-profiling the first 10 nm or so of semiconductor materials at extremely good depth resolution. It is therefore beneficial to understand how inorganic, and specifically semiconductor, surfaces respond to GCIB exposure, especially from the point of view of the XPS analyst, a technique where surface damage can be readily detected. This may also be of importance when attempting removal of organic contamination from a surface13 or when depth-profiling through an organic layer deposited on top of an inorganic substrate14 (for example, an etch-resist residue on a semiconductor surface). Prior to our experimental work, we expected the damage to inorganic materials to be unspectactular and perhaps undetectable. We have been surprised to find that in a few cases, this damage can be very significant, 116, 054908-1 C 2014 AIP Publishing LLC V [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 128.240.229.70 On: Wed, 29 Apr 2015 13:55:27 054908-2 Barlow, Portoles, and Cumpson J. Appl. Phys. 116, 054908 (2014) changing not just the surface chemistry of an inorganic semiconductor but even causing gross surface discoloration to the naked eye, as shown in Figure 1. In this work, GCIB sputter depth profiles were performed on a variety of compound semiconductors, CdTe, ZnSe, GaP, GaAs, and InAs, and a high-j dielectric material, HfO. For most of these materials, no detectable damage (taken as the reduction or preferential sputtering of species within the material) was observed. For a small subset, however, damage became evident in the XPS spectra after long etch times and eventually was visible to the naked eye on the surface of the semiconductor. All XPS analyses were performed in our Theta Probe XPS instrument (Thermo Scientific, East Grinstead, UK), utilising a monochromatic Al Ka X-ray source (1486.6 eV) operated at 100 W. While the Theta Probe is capable of operating in Parallel Angle-Resolved XPS (ARXPS) mode, the work discussed herein was collected in the “standard” lens mode, whereby all electrons were collected from 20 to 80 to the surface normal. Spectra were collected with pass energy of 40 eV and were the average of 10 sweeps. Charge neutralisation with a combination of low-energy electrons and argon ions was used throughout all experiments, and checks made that this low energy monatomic ion exposure alone did not lead to damage. The GCIB source is a Thermo Scientific MAGCIS gun, which throughout this work was operated at 6000 eV in “small” cluster mode, meaning a broad, semi-log distribution of Argon cluster size centred around 1000 atoms per cluster at an input pressure of 4 bars. On the low mass side, this distribution has a half-width of 200 atoms/cluster, and on the high mass side can extend to several thousand atoms per cluster. It is expected that the high mass clusters will have significantly less impact on the etch rate relative to the desired 1000 atom clusters. Neutral atoms are not expected to escape from the MAGCIS due to a 2 bend in the optics, and monoatomic ions are well below the cut-off of the MAGCIS mass filter. Further to this, multiply charged Ar clusters are not expected to be stable enough to emerge from the source.15 An ion beam raster of 1 mm 2 mm was used, with an X-ray spot size of 200 lm 400 lm, i.e., well within the area of the sputter crater. Throughout all experiments, the beam current was stable around 20 6 2 nA (as measured at the sample plate within the instrument). An etch cycle of 300 s was used for all depth profiles. Sputter time was converted to depth from experimentally determined etch rates on the high-j dielectric material, HfO, using a method detailed previously.12 During sputtering, a contact shadow mask aligned with the ion beam is used to pattern fine step-edges throughout the area of the sputter crater. The total etch depth and thus etch rate are then straightforward to measure using white-light interferometry (Newview 5000, Zygo, Middle-field, CT, USA). An example linescan across one step-edge is provided in Figure 2. The total crater depth is then taken as the step height in the profile, and multiple measurements were made across step-edges throughout the etch crater. From this analysis, an average etch rate of 0.025 nm/min was then determined for HfO, using a 6 keV Ar GCIB of 1000 atom clusters. Upon conversion, this relates to an etch rate of 6.77 10 24 cm3/ion, 6.77 10 27 cm3/atom in ion, and 3.95 10 2 amu/atom in ion. FIG. 1. Photograph of InAs surface after 6 keV and 2 keV GCIB depth profiles. The resultant damage after the 6 keV GCIB exposure is clearly visible. FIG. 2. Linescan across a step-edge from a white light interferometry topography map representative of the crater depth on HfO after etching. II. RESULTS Spectra acquired at two cumulative etch times during the depth profiles of compound semiconductors are given in Figure 3. Carbonaceous contamination for all samples was completely removed after a single etch cycle of 300 s at 6 keV (small clusters), and native oxides were typically removed after 3 to 6 etch cycles. For ease of comparison, the spectra from 2100 and 6000 s are shown here for all semiconductors studied. From the HfO experiments, this relates to an approximate depth of 0.9 nm and 2.5 nm, respectively. For each compound, a narrow-scan spectrum of most analytical interest for each element in the semiconductor was collected. Once the native oxides were removed through etching, the shape and position of the peaks observed in the spectra for each of the compounds remained the same throughout all 6000 s of exposure to the 6 keV small cluster beam. Peak positions and widths were measured for each of the spectra, taken from the lowest binding energy component of the doublets analysed (for example, the 5/2 component of a 3d doublet). The minimal change in either peak position or peak width (measured by FWHM) suggests that the compound semiconductors CdTe, GaAs, GaP, and ZnSe and the high-j dielectric HfO are etched by the 6 keV small cluster beam without damage. The measurements were repeated on InAs, and the results are presented in Figure 4. The native oxide on the InAs sample was rapidly removed by the 6 keV small cluster beam, with minimal oxide left after the first 300 s etch cycle. At this point, the In3d and As3d spectra were measured and observed to be as expected for InAs, with the In3d5/2 peak [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 128.240.229.70 On: Wed, 29 Apr 2015 13:55:27 054908-3 Barlow, Portoles, and Cumpson J. Appl. Phys. 116, 054908 (2014) FIG. 3. Analytical peaks of interest for two total etch times, 2100 s and 6000 s, during the depth profiles of (a) CdTe, (b) GaAs, (c) GaP, (d) ZnSe, and (e) HfO. Peak widths for each of the spectra are given in the inset tables and were taken from the lowest binding energy components of the spectra. position measured as 444.3 eV in good agreement with previous work.16,17 After a total etch time of 2100 s, however, broadening of the In3d peaks was observed, and reflected in the measured FWHM of the 3d5/2 component. After the full 6000 s depth profile, a noticeable “shoulder” was evident on the low binding energy side of the In3d5/2 peak. Minimal change was observed in the As3d peak, however. Analysis of the In4d doublet also showed a similar low binding energy shoulder after the full depth profile as shown in Figure 4(b). In both instances, the extra feature matches in binding energy what would be expected for In metal, suggesting that the cluster ion beam is reducing the In within the InAs to its metallic state. The In3d5/2 peak was fitted with two synthetic components, one at 444.3 eV for InAs and the other at 443.8 eV for In metal.16,17 Plotted in Figure 4(c) is the change in ratio of these features as the depth profile progressed. Note that we saw no evidence of X-rays accelerating the damage or sputter yield in any of these inorganic samples, whereas we have previously seen that X-ray enhancement of cluster ion sputter yields in certain polymers.18,19 To reduce ion-induced damage, the energy-per-atom can be reduced. For a 6 keV small cluster beam, we expect a distribution of energies around 6 eV per atom in the [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 128.240.229.70 On: Wed, 29 Apr 2015 13:55:27 054908-4 Barlow, Portoles, and Cumpson J. Appl. Phys. 116, 054908 (2014) FIG. 4. Spectra from the depth profile of InAs. The In3d and As3d spectra are given in (a) for three steps during the 6 keV small cluster depth profile. A feature becomes evident in the In3d spectra resulting from a reduction of In to the metallic state. This change in peak shape is also evident in (b) for the In4d spectra. (c) Peak fitting the In3d5/2 component with two synthetic peaks for InAs and In metal shows how this damage develops with etch time and depth. Reducing the ion energy to 2 keV in (d) allows the depth profile to be performed with no damage detected in the spectra. cluster (1000 atoms per cluster), which can be reduced to 2 eV per atom by reducing the ion energy to 2 keV. Figure 4(d) shows the In3d spectra for 3 stages during a depth profile of InAs using a 2 keV small Ar GCIB. Previously, damage was evident after 2100 s. With the reduced energy, even after 10 000 s of ion etching, no damage was evident in the In3d spectra. This is reflected by a visual inspection of the InAs surface after etching as was introduced in Figure 1. The 6 keV sputter locations are clearly visible as regions of damage where the InAs has been reduced to form metallic In at the surface. In the region of the 2 keV depth profile, however, there is no visible damage at the surface. Of course, one must remark that at reduced ion energy, it should be expected that the sputter yield will also decrease, thus leading to longer sputter times being required for a depth profile. However, the strength of a GCIB on inorganic surfaces lies in its ability to remove surface contamination and oxides with minimal and shallow damage to the surface of interest. The 2 keV small cluster beam was able to remove the surface oxide after 2000 s of etching. Continuing exposure to this beam up to 10 000 s saw no adverse impact to the chemistry at the surface, suggesting a useful margin in the removal of contamination or oxide. The higher energy 6 keV beam removed the oxide in 300 s; however, the peak shape showed signs of broadening immediately beyond this time, suggesting that the reduction of the InAs was occurring after a relatively brief exposure, leaving less margin for choice of sputtering time sufficient to remove the contamination and oxide. It might have been expected that exposure of an inorganic material to a GCIB would not result in damage at the surface which would in turn influence an analytical result, especially given the large difference in sputter yields between inorganics and organics for these ion beams. However, it is now apparent that some materials may suffer from damage due to these beams within a realistic experimental time frame. Thus, the proper use of GCIB sources becomes a balance between minimising damage to a surface of interest, while also minimising the length of time needed for a piece of analysis. Elsewhere, we have derived a “figure of merit” for removal of organic overlayers on inorganic substrates,15 indicating that organic overlayers are [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 128.240.229.70 On: Wed, 29 Apr 2015 13:55:27 054908-5 Barlow, Portoles, and Cumpson preferentially removed from the surfaces of inorganic substrates particularly for cluster beams with energy below around 4 eV/atom. This was obtained purely by consideration of the relative sputter rates of typical organic and inorganic materials. Our present work suggests that this low energy-per-atom, or lower, may be even more desirable for cases, like InAs, where the underlying inorganic material may be chemically modified by clusters of higher energyper-atom. III. CONCLUSIONS Five compound semiconductors and a high-j dielectric material were depth profiled using an Ar GCIB. CdTe, GaAs, GaP, ZnSe, and HfO all responded well to a 6 keV, 1000 atom Ar cluster beam with no appreciable damage observed over a 6000 s depth profile. Carbonaceous contamination was removed during the first 300 s etching step, and oxides were typically removed after 3 to 6 etch cycles. Damage was, however, observed in InAs, and XPS indicated this to be a reduction of indium to its metallic state. This was observed to begin occurring almost immediately after the first 300 s etch cycle, though became most prominent after 2100 s of exposure. After reducing the energy-per-atom to 2 eV, the depth profile could be performed for a cumulative etch time of 10 000 s without any appreciable damage. ACKNOWLEDGMENTS X-ray photoelectron spectra were obtained at the National EPSRC XPS Users’ Service (NEXUS) at Newcastle University, an EPSRC Mid-Range Facility. The J. Appl. Phys. 116, 054908 (2014) authors would like to thank Mr. Mike Foster for technical assistance and Dr. Naoko Sano for useful discussions. 1 I. Yamada, J. Matsuo, N. Toyoda, and A. Kirkpatrick, Mater. Sci. Eng., R 34, 231–295 (2001). L. P. Allen, T. G. Tetreault, C. Santeufemio, X. Li, W. D. Goodhue, D. Bliss, M. Tabat, K. S. Jones, G. Dallas, and D. Bakken, J. Electron. Mater. 32, 842–848 (2003). 3 A. J. Perry, S. J. Bull, A. Dommann, D. Rafaja, B. P. Wood, and M. Michler, Surf. Coat. Technol. 133–134, 253–258 (2000). 4 N. Winograd, Z. Postawa, J. Cheng, C. Szakai, J. Kozole, and B. J. Garrison, Appl. Surf. Sci. 252, 6836–6843 (2006). 5 C. M. Mahoney, Mass Spectrom. Rev. 29, 247–293 (2010). 6 T. Miyayama, N. Sanada, S. R. Bryan, J. S. Hammond, and M. Suzuki, Surf. Interface Anal. 42, 1453–1457 (2010). 7 N. Winograd, Surf. Interface Anal. 45, 3–8 (2013). 8 A. Delcorte, B. J. Garrison, and K. Hamraoui, Anal. Chem. 81, 6676–6686 (2009). 9 C. Anders, H. M. Urbassek, and R. E. Johnson, Phys. Rev. B 70, 155404 (2004). 10 T. Seki, T. Murase, and J. Matsuo, Nucl. Instrum. Methods Phys. Res., Sect. B 242, 179–181 (2006). 11 P. J. Cumpson, J. F. Portoles, N. Sano, and A. J. Barlow, J. Appl. Phys. 114, 124313 (2013). 12 P. J. Cumpson, J. F. Portoles, and N. Sano, J. Vac. Sci. Technol., A 31, 020605 (2013). 13 P. J. Cumpson and A. J. Barlow, “Optimum conditions for cleaning surfaces by argon gas cluster ions prior to XPS analysis,” Surf. Interface Anal. (unpublished). 14 P. J. Cumpson, J. F. Portoles, N. Sano, A. J. Barlow, and M. Birch, Surf. Interface Anal. 45, 1859–1868 (2013). 15 Discussion with Thermo-Scientific, East Grinstead, UK. 16 M. Procop, J. Electron Spectrosc. Relat. Phenom. 59, R1–R10 (1992). 17 G. Hollinger, R. Skheyta-Kabbani, and M. Gendry, Phys. Rev. B 49, 11159–11167 (1994). 18 P. J. Cumpson, J. F. Portoles, N. Sano, and A. J. Barlow, J. Vac. Sci. Technol., B 31, 021208 (2013). 19 P. J. Cumpson, J. F. Portoles, and N. Sano, Surf. Interface Anal. 45, 601–604 (2013). 2 [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 128.240.229.70 On: Wed, 29 Apr 2015 13:55:27
© Copyright 2026 Paperzz