Introduction

Depth profiling using time-of-flight secondary ion mass spectrometry (ToF-SIMS) has been widely utilized in semiconductor industry for more than 20 years [1]. Recently, it has become increasingly popular in the fields of biology, geology, and novel material research [2]. However, unlike in semiconductor industry, the majority of samples in these new application fields are insulators.

In ToF-SIMS depth profiling analysis, a dual-beam operation composed of a sputtering and an analysis beam is typically used. The analysis beam is optimized for collecting high-quality mass spectra, and the sputtering beam is optimized for high sputter rate, optimum depth resolution, and optimum ionization yield [1]. Two operation modes are generally performed: interlaced mode and non-interlaced mode. In interlaced mode, the beam is active between two analysis shots that are operated quasi-simultaneously. It is commonly applied for depth profiling of conductive or semiconductive samples, having the advantage of time-saving and the ability to provide full depth information [1]. However, for depth profiling of insulating samples, the employment of a low energy (≤10 eV) electron beam is not sufficiently effective to compensate the intensive charge accumulation at the sputtering interface, especially for traditional O2 + and Cs+ sputtering beams (low energy (0.2–2.0 kV) and high current (tens to hundreds nA)) [3]. To this end, non-interlaced mode is proposed, in which the sputtering phases (typically, several seconds each) and the analysis phases (several seconds each) are separated, and charge compensation time slots (several seconds each) are added in between.

Although the non-interlaced mode can effectively alleviate the charging effect to a limited extent, its measurement is normally 3–5 times longer than the interlaced mode if a similar depth resolution is required, making the deep depth profiling (e.g., 1–10 μm) of insulating samples a time-consuming task [3].

Since argon cluster (Arn +) ion sources were implemented in SIMS [4, 5] and subsequently used in X-ray photoelectron spectroscopy (XPS) [6] and ultraviolet photoelectron spectroscopy (UPS) [7], it has been widely applied for the molecular depth profiling of polymers, biological molecules, and other organic materials because they can retain molecular information during the erosion process. In addition, they can provide high sputtering rate, allowing the fast depth profiling of above soft materials [815]. However, the application of Arn + ion source to depth profiling of inorganic insulating materials has rarely been attempted, possibly because of the lesser efficiency-enhancing ionization yield compared with O2 + and Cs+ ion sources for both positive and negative ions.

In this work, we evaluated the performance of an Arn + sputtering source in ToF-SIMS compared with traditional O2 + and Cs+ ion sources in terms of efficiency, accurate interfacial chemical information, and mass resolution in depth profiling of several representative insulating samples. A leached simulated borosilicate nuclear waste glass(SON68), a La0.93Sr0.07CrO3 thin film on a SrTiO3 substrate, and a SrTiO3/SrCrO3 bi-layer film on a SrTiO3 substrate were selected as model systems for our study because of their fundamental and practical importance[1626].

Experimental

Materials and Sample Preparation

The components, preparation, and leaching procedure of the SON68 glass have been described elsewhere [20, 27]. In brief, it was made by batching carbonates and oxides of the various metals, melting in a furnace, and quenching on a stainless steel plate. This glass was then pulverized and re-melted to ensure a homogeneous solid. The second melt was poured into molds to produce rectangular bars. Coupons (~10 × 10 × 1 mm3) were cut from the bars and both sides of the coupons were polished. To study the diffusion behavior of Li, the SON68 coupons (natural isotopic abundance for all elements) were submerged in a solution of a dimethyl sulfoxide (DMSO) with dissolved LiCl (6Li enriched, 6Li:7Li = 95:5). The coupons were in contact with the solution at 150°C for 10 d. When the coupons were removed from the vessel, they were rinsed with clean DMSO, water, then ethanol, and dried in an oven.

Epitaxial La0.93Sr0.07CrO3 (in brief, LaSrCrO3) thin films with a thickness of ~66 nm were grown on TiO2-terminated SrTiO3 (001) substrates by molecular beam epitaxy (MBE). The detailed preparation process has been described elsewhere [26]. In brief, the substrates were loaded into an ultrahigh vacuum chamber and heated at 700°C prior to growth. La, Sr, and Cr were evaporated for high-temperature effusion cells, and flux rates were calibrated using a quartz crystal microbalance (QCM). The O2 partial pressure was kept at ~2.0 × 10–6 Torr during growth. The procedures for epitaxial growth of SrTiO3/SrCrO3 on SrTiO3 are essentially the same as those of La0.93Sr0.07CrO3, except that the oxygen line was changed to isotope 18O2. A 50 nm thick SrCrO3 film was grown first on TiO2-terminated SrTiO3, then followed by 50 nm thick SrTiO3.

Depth Profiling

Dual beam depth profiling experiments were performed on a ToF-SIMS instrument (TOF.SIMS5; IONTOF GmbH, Münster, Germany) in the Environmental Molecular Sciences Laboratory (EMSL), located at Pacific Northwest National Laboratory (PNNL). Three sputtering ion sources were available: Arn + (2.5–20 keV), O2 + (0.2–2.0 keV), and Cs+ (0.2–2.0 keV) sources. A 25 keV pulsed Bi+ ion beam was used as the analysis beam, and the analysis area was 100 × 100 μm2 or 50 × 50 μm2 at the center of the sputter crater. Interlaced mode was applied to most measurements because it is more time-efficient than non-interlaced mode. Non-interlaced mode was applied to compare signal intensity, mass resolution, and also to prove the accuracy of interfacial chemical information. Charge compensation was used for all depth profiling measurements. The details for the adjustment of charge compensation and additional information for ToF-SIMS measurement can be found in our previous publication [28].

The depths and crater shapes of sputter craters were measured using a Dektak 6 M stylus profilometer (UT, USA). For convenience, a constant sputter rate was adopted in each measurement. Tapping mode atomic force microscope (AFM) (Digital Instruments (Veeco) Nanoscope III Multimode, Tonawanda, New York, USA) was used to characterize the roughness of the crater bottoms. The craters in LaSrCrO3/SrTiO3 thin film were imaged using TESP silicon probes (Veeco probes) (42 N/m nominal stiffness), operating at a set point of 80% of its free amplitude. Both height and phase data were collected simultaneously during the characterization.

Results and Discussion

Sputter Rate

In absolute terms, the sputter rate of the Arn + source on inorganic materials may not be as good as that of the O2 + or Cs+ sources. For example, on the updated IONTOF instrument (TOF.SIMS5) in EMSL, with a 300 × 300 μm2 sputtering area, the highest sputter rate of the Arn + ion source on a SON68 glass sample is about 0.73 nm/s (20 keV Ar1500 +, 12 nA). This is lower than 1.5 nm/s for the O2 + source (2.0 keV O2 +, 600 nA) and 0.82 nm/s for the Cs+ source (2.0 keV Cs+, 150 nA) (Table 1). However, because the focus of 20 keV Arn + ion beam is better than either the Cs+ beam or O2 + beam, a smaller sputter area is required for the same size analysis area when the Arn + source is used, which allows a higher sputter rate achieved in comparison with either the Cs+ beam or O2 + beam (Figure S1 in the Supplemental Information).

Table 1 The Highest Sputter Rates of Different Ion Sources Over Different Analysis Areas on a SON68 Glass Sample

For dual beam depth profiling, the sputter area needs to be considerably larger than the analysis area to avoid “crater edge” effect, and differs from one sputtering source to another for the same analysis area. We perform depth profiling of the LaSrCrO3/SrTiO3 sample (Supplementary Figure S2) using different sputtering sources at the same size of analysis area (100 × 100 μm2). The depth resolution, which can be defined as the depth interval over which the signal intensity of a selected ion varies from 16% to 84% at the sharp interface, is an important parameter to characterize the quality of depth profiling, and used as reference to decide the sputter size of a sputtering source requested for a certain analysis area. Figure 1 shows the relationship between the depth resolution and the sputter size for the three ion sources. The results show that when the sputter size is equal to or larger than 150 × 150 μm2, the depth resolution of 20 keV Arn + sputtering becomes stable. As a comparison, 200 × 200 μm2 is required for the Cs+ source, and 300 × 300 μm2 is required for the O2 + source. It means that the minimum sputter sizes for a 100 × 100 μm2 analysis area are 150 × 150, 200 × 200, and 300 × 300 μm2 for the Arn +, the Cs+, and the O2 + sources, respectively. Accordingly, the sputter rate of the 20 keV Arn + beam can reach 2.9 nm/s, which is faster than those of the 2.0 keV O2 + beam (1.5 nm/s) and the 2.0 keV Cs+ beam (~1.9 nm/s). When a 50 × 50 μm2 analysis area is required, the performance improvement of the Arn + beam over the O2 + or Cs+ beams is even more pronounced (Table 1).

Figure 1
figure 1

Depth resolution obtained at the interface of the LaSrCrO3/SrTiO3 film. A 100 μm × 100 μm analysis area (with 25 keV Bi+ beam) was used, and the Cr+ profile (Supplementary Figure S2 in the Supporting Materials) was chosen to calculate depth resolution

Surface/Interface Charging

Surface charging is a major challenge to ToF-SIMS depth profiling of insulating materials. Accumulated charges at the sputtering interface can change the energy distribution of the emitted secondary ions, degrading their transmission and detection by the mass spectrometer, which results in both low signal intensity and poor mass resolution. Despite the fact that a flood gun can be used for charge compensation, this does not work efficiently if a high sputter rate is needed because high sputter rates require both high sputter currents and an interlaced sputtering mode. Thus, insufficient charge compensation could be detrimental to the analysis of certain types of materials. For example, Li+ diffusion in the SON68 glass during leaching can easily reach several μm deep. When a 2.0 keV O2 + sputtering beam was used to characterize Li+ diffusion in the glass with interlaced mode, low 7Li+ and 6Li+ signals were observed. More seriously, a sudden anomalous drop of Li+ signals was observed at a depth of ~2500 nm as shown in Figure 2. This is not an uncommon occurrence in glass depth profiling. Similar situations have been observed when using an O2 + beam in deep depth profiling of other glass materials, such as a glass microscope slide (data not shown here). Apparently, the accumulation of a larger amount of charges at the sputtering interface is responsive for the unreasonable experimental data. Though non-interlaced mode can solve this problem, the measurement time needs to be ~3 to 5 times longer.

Figure 2
figure 2

A comparison of interlaced mode depth profiles of 7Li (dash line) and 6Li (solid line) in a SON68 glass sample using 20 keV Arn +(red) and 2 keV O2 + (blue) sputtering beams. (a) Intensity profiles; (b) ratio profiles

Figure 3 and Table 2 show that signal intensity and mass resolution may improve if the energy and/or current of the sputtering beam is reduced, for example, changing the energy from 2.0 keV to 1.0 keV. But the sputter rates dramatically decrease as a consequence (Table 1), which in turn increase the measurement times as does the use of non-interlaced mode.

Figure 3
figure 3

A comparison of (a) 28Si+ and (b) 18O peaks between different sputtering beams and modes on a SON68 glass sample. A 25 keV Bi+ beam was used as analysis beam. For each spectrum, same setting of the analysis (Bi+) beam (e.g., beam current, pulse width, pixel number, and frame number) was used

Table 2 A Comparison of Mass Resolution and Signal Intensity of 28Si+ and 18O Peaks with Different Sputtering Beams and Sputtering Modes

In a previous publication, we observed that a cluster ion sputtering source might reduce charging at the sputtering interface and enhance signal intensity of secondary ions [29]. In this study, when a 20 keV Arn + ion source was used for the SON68 glass sample, much stronger Li+ signals were observed (Figure 2). The intensity of 7Li+ and 6Li+ was nearly 100 times higher than that observed by using the 2.0 keV O2 + source. In addition, no sudden drop of Li+ signals (as well as the total ion signal, data not shown) occurred at ~2500 nm deep, even until 20 μm depth was reached.

It is worthy to note that using Arn + sputtering under interlaced mode provides not only high sputter rates and strong signal intensities but also good mass resolution. Figure 3 shows the 28Si+ and 18O- peaks from the SON68 sample under different conditions of O2 +, Cs+, and Arn + sputtering. The mass resolution values are summarized in Table 2. The mass resolutions of the 28Si+ peak from interlaced mode 2.0 and 1.0 keV O2 + sputtering are about 1600 and 5000, respectively, whereas the mass resolution from interlaced mode 20 keV Arn + sputtering is about 8200. Non-interlaced mode sputtering provides the best mass resolution, ~9900 for 2.0 keV O2 + and ~8600 for 20 keV Arn +, which can be attributed to an effective charge compensation. A more interesting observation is that the mass resolution of 18O- from Arn + sputtering is better than that from non-interlaced Cs+ sputtering. A plausible explanation is that Cs implantation changes the electric properties of the sputtering interface, whereas Arn + sputtering does not result in this effect. This virtue is very valuable in glass corrosion research. As mentioned earlier, isotopically enriched glass samples have been introduced to elucidate the corrosion mechanisms of glass, and SIMS has been a major technique to characterize isotopic depth profiles in these samples [27]. However, mass interferences have been a problem for interlaced mode O2 + depth profiling. For example, for the ToF-SIMS analysis of 44Ca and 57Fe enriched SON68 glass samples, an interference peak, 28Si16O+ (m/z = 43.972), is very close to 44Ca+ (m/z = 43.956), and a minimum mass resolution of 2700 is required to separate the pair of ions. Similarly, 40Ca16O+ (m/z = 55.958) is very close to 56Fe+ (m/z = 55.935), and a minimum mass resolution of 2500 is required to distinguish them. In practice, at least twice the minimum mass resolution is required to obtain reasonable measurement precision. The traditional solution to obtain decent mass resolution in depth profiling of insulator samples is to use the non-interlaced mode, but the cost in time may not be affordable for many applications.

We already show that with the use of an Arn + sputtering beam, high sputter rates, high mass resolution (>8000), and reasonable signal intensities can be obtained simultaneously. Compared with O2 + or Cs+ sputtering, depth profiling of glass samples using an Arn + sputtering source can dramatically reduce experimental time (9- to 15-fold time-saving) without compromise of data quality. The 9- to 15-fold time-saving is estimated because both interlaced mode and smaller sputter area can be used for “high-speed” Arn + sputtering, but none of them is practical for traditional “high-speed” O2 + or Cs+ sputtering. This advantage is especially important for deep depth profiling (e.g., ≥3 μm) because the measurement time is very long if Cs+ or O2 + sputtering is used.

In Table 2, the signal intensity and mass resolution of the 28Si+ or 18O- peak with non-interlaced mode 20 keV Arn + sputtering is only slightly (5%–15%) better than that from interlaced mode. This means that with interlaced mode 20 keV Arn + sputtering, the charge compensation is very effective, though it may not be as good as that with non-interlaced mode. As a comparison, interlaced and non-interlaced mode using 2.0 keV O2 + sputtering show ~6 times difference in mass resolution and ~560 times difference in signal intensity for the 28Si+ signal. Interlaced and non-interlaced mode using 2.0 keV Cs+ sputtering show ~7 times difference in mass resolution and ~30 times difference in signal intensity for the 18O- signal. Apparently, Arn + sputtering leads to much lower charge residuals at the sputtering interface, which may be attributed to two reasons. First, as shown in Table 2, the sputtering yield of a 20 keV Arn + ion is about 0.88 nm3/ion, much larger than that of a 2.0 keV Cs+ ion (0.079 nm3/ion) or an O2 + ion (0.036 nm3/ion). Therefore, the implanted charge number with Arn + sputtering is much lower. At the same time, a big Arn + cluster (e.g., n = 1500 for the 20 keV Arn + used in this work) would break down to many single Ar atoms and small Ar clusters at the sputtering interface. These Ar atoms and small Ar clusters can take some charges away from the sputtering interface. The advantages of Arn + sputtering over traditional O2 + and Cs+ sputtering in depth profiling of insulator samples have been schematically illustrated in Figure 4.

Figure 4
figure 4

A schematic illustration of the advantage of Arn + sputtering over O2 + and Cs+ sputtering in depth profiling of insulating samples. Although implantation of O and Cs atoms may improve ionization yield, O2 + and Cs+ sputtering introduces too many residual charges at the sputtering interface on insulating samples. As a comparison, a low charging sputtering interface can be obtained with Arn + sputtering because (1) the high sputter yield per Arn + ion leads to much less charges implanted in to the sputtering interface; and (2) a big Arn + ion would break into many single Ar atoms and small Ar clusters, which can take some charge away from the sputtering interface

The advantage that the Arn + sputtering can provide a low charging sputtering interface during interlaced mode depth profiling of insulating samples was verified further by analyzing the multilayer functional thin oxide films on insulating substrate. This sample was prepared by growing a 50 nm SrCrO3 thin film and then a 50 nm SrTiO3 thin film on SrTiO3 substrate (SrTiO3/SrCrO3/SrTiO3) in 18O2 atmosphere, and 18O is found to be diffused into the SrTiO3 substrate with a significant depth (~10 μm as shown in Figure 5a). An Arn + sputtering beam (20 keV) with interlaced mode was used to measure 18O diffusion behavior. It took about 4 h (sputter rate was about 1.1 nm/s, 10.0 nA Arn + beam, 200 × 200 μm2 sputter area) to perform this measurement. The results show that the 18O diffuses into the SrTiO3 substrate for about 12 μm. As a comparison, if a 2.0 keV Cs+ sputtering source is used with interlaced mode, the sputter rate is about 0.31 nm/s (100 nA Cs+ beam, 300 × 300 μm2 sputter area). The 18O-/(18O + 16O-) depth profile with Cs+ sputtering is shown in Figure 5b. It was surprising to observe that in a range from the SrTiO3/SrCrO3 interface (~50 nm deep) to a depth of ~400 nm, the Cs+ data is significantly higher and noisier than the Arn + data. To validate the data reliability, we reduced the Arn + beam current, to make its sputter rate similar to the Cs+ source sputter rate, and the data from this measurement match well with the data from the high-current Arn + sputtering data. In addition, non-interlaced mode depth profiling using the Cs+ sputtering source shows a result similar to the Arn + data (Figure 5b). These results suggest that there was a problem with interlaced mode depth profiling using the 2.0 keV Cs+ beam. After carefully checking the O- signal intensity and peak shape, we found that the total O signal (18O + 16O) with 2.0 keV Cs+ interlaced mode had a big drop from ~50 nm to ~100 nm (in the SrCrO3 layer). The total O signal showed a jump at the SrCrO3/substrate interface (~100 nm deep) and gradually increases until ~400 nm, where the intensity of the total O signal is similar to that from the surface SrTiO3 layer (0–50 nm) (Figure 5c). At the same time, the O peak centers shift in this depth range (data are not shown here). The signal drop and peak shift can be attributed to the different electrical conductivity in the different layers, which leads to a dramatic and continuous change in the charging state during the interlaced Cs+ sputtering. The non-interlaced mode can be used to avoid this pitfall (Figure 5c), but the cost in time may be an issue. As a comparison, interlaced mode Arn + sputtering data show limited signal intensity variations and peak shifting. These imply that erroneous results may manifest themselves because of a variable charging state with interlaced mode Cs+ depth profiling, whereas Arn + sputtering beam can be used with interlaced mode without serious charging effect, providing accurate chemical information in the interface of multilayer oxide films on insulating substrate.

Figure 5
figure 5

16O and 18O depth profiles in a SrTiO3/SrCrO3 thin film sample on SrTiO3 substrate. (a) 20 keV 12 nA Arn + sputtering beam under interlaced mode. (b)18O/ (16O + 18O) depth profiles using different sputtering sources and sputtering modes. (c) Depth profiles of normalized oxygen signal (16O + 18O) with different sputtering sources and sputtering modes

Ionization Yield

Compared with traditional O2 + and Cs+ sputtering sources, a major concern with the use of Arn + sputtering in inorganic depth profiling is the lack of the capability to enhance ionization yield. For example, the Si+ and Si- signals are weak if an Arn + sputtering beam is used to perform depth profiling of a silicon wafer. Our data show that compared with 20 keV Arn + sputtering, 2.0 keV Cs+ sputtering can provide ~2000 times signal enhancement for Si, and 2.0 keV O2 + sputtering can provide ~50 times signal enhancement for Si+ on a silicon wafer. However, ionization yield enhancement may not be a problem for glasses or other oxide samples because these materials are generally ionic structures, guaranteeing reasonable ionization yields with any sputtering source. To compare the ionization yield difference, non-interlaced mode depth profiling with O2 +, Cs+, and Arn + sputtering was performed on the SON68 glass sample. The 28Si+ and 18O- peaks were used as representative signals (Figure 3 and Table 2). If defining the 28Si+ peak area from 2.0 keV O2 + non-interlaced mode sputtering was 100%, the relative peak area of the 28Si+ arising from 20 keV Arn + non-interlaced mode sputtering was ~50%. This indicates that for this glass sample, the ionization yield enhancement using O2 + sputtering is only twice that using 20 keV Arn + sputtering, which is much smaller than the enhancement value (~50 times) of Si+ on a silicon wafer. Similarly, Cs+ sputtering provides only a mild ionization yield enhancement (~3 times) compared with Arn + sputtering, and is also much smaller than that (~2000 times) for Si- signal on a silicon wafer. Similar results have been observed on other glass and oxide samples (data not shown here).

Roughness of Sputtering Interface

As we discussed above, the high sputtering yield of a 20 keV Arn + ion leads to a low charging sputtering interface, allowing us to perform fast depth profiling of glass and functional oxide samples with very limited sacrifice of signal intensity and mass resolution. However, there are still some issues related to the use of this sputtering source. For example, it may increase the roughness of the sputtering interface if one compares the interface to that obtained with O2 + and Cs+ sources. Figure 6 shows the AFM images in the crater bottom on the LaSrCrO3/SrTiO3 thin film sample after sputtering with 20 keV Arn +, 2.0 keV O2 +, and 2.0 keV Cs+ beam, respectively. The RMS (root mean square) roughness of the Arn + crater bottom is about 9–10 nm, the comparable RMS values of the O2 + and the Cs+ craters are about 4–5 and 3–4 nm, respectively. These results correspond to the data shown in Figure 1, where the best depth resolution of the 20 keV Arn + sputtering at the LaSrCrO3/SrTiO3 interface is about 5 nm, approximately two times that of the 2.0 keV Cs+ or the 2.0 keV O2 + sputtering (2–3 nm). These results are not in agreement with the general concept that cluster sputtering ions provide a much smoother sputtering interface than single atom sputtering ions used with similar energy [30]. However, we noticed that previous reports in the literature normally compared the roughness induced by various sputtering sources used with similar energy, where the roughness induced by single atom ions is much worse than cluster ions. If the energy used to different sputtering sources is largely different, the situation may become complicated because the roughness induced by a sputtering ion may increase with increase in the used energy. For example, focused ion beam and atom probe tomography (FIB-APT) experiments showed that 30 keV Ga+ ion beam can cause damage in a Si sample as deep as 40 nm, but the damage layer induced by 2 keV Ga+ is less than 1 nm thick [31]. Taking this into account, the poor roughness induced by Ar cluster sputtering sources observed in the present work may be attributed to the high energy (20 keV) used during the sputtering; the comparable energy used during the Cs+ or O2 + sputtering was 2.0 keV.

Figure 6
figure 6

AFM images at the crater center areas on the 66 nm thick LaSrCrO3 film on SrTiO3 substrate after sputtering with (a) 20 keVArn +, (b) 2.0 keV O2 +, (c) 2.0 keV Cs+ beams. All crater depth is about 100 nm

Conclusions

We have demonstrated the superior performance of an Arn + sputtering source relative to traditional O2 + and Cs+ sputtering sources for depth profiling of various inorganic insulating samples in terms of their sputter rates and accuracy of interfacial chemical information. Because of the high sputter rate and low charge accumulation of the Arn + sputtering source, interlaced mode of dual-beam depth profiling can be easily performed on insulating samples without compromising the depth and interfacial chemical information. Although Arn + sputtering sources may cause some roughness problem, the measurement time of depth profiling of an insulating sample using an Arn + sputtering source can be dramatically reduced compared with traditional O2 + and Cs+ sputtering sources, making it a feasible sputtering source for magnetic SIMS, XPS, and other depth profiling besides ToF-SIMS. The implementation of an Arn + sputtering source can enhance the capabilities for analysis of insulating materials and, thus, expand applications in the field of glass corrosion, study of functional thin films on insulating substrates, and many other potential fields related to insulating materials with complex layered structures.