Article pubs.acs.org/ac
Large O2 Cluster Ions as Sputter Beam for ToF-SIMS Depth Profiling of Alkali Metals in Thin SiO2 Films Sabine Holzer,†,‡ Stefan Krivec,*,† Sven Kayser,§ Julia Zakel,§ and Herbert Hutter‡ †
Infineon Technologies Austria AG, Siemensstraße 2, 9500 Villach, Austria Vienna University of Technology, Institute of Chemical Technologies and Analytics, Getreidemarkt 9, 1060 Vienna, Austria § ION-TOF GmbH, Heisenbergstraße 15, 48149 Münster, Germany ‡
ABSTRACT: A sputter beam, consisting of large O2 clusters, was used to record depth profiles of alkali metal ions (Me+) within thin SiO2 layers. The O2 gas cluster ion beam (O2-GCIB) exhibits an erosion rate comparable to the frequently used O2+ projectiles. However, because of its high sputter yield the necessary beam current is considerably lower (factor 50), resulting in a decreased amount of excess charges at the SiO2 surface. Hence, a reduced electric field is obtained within the remaining dielectric layer. This drastically mitigates the Me+ migration artifact, commonly observed in depth profiles of various dielectric materials, if analyzed by time-of-flight secondary ion mass spectrometry (ToFSIMS) in dual beam mode. It is shown, that the application of O2GCIB results in a negligible residual ion migration for Na+ and K+. This enables artifact-free depth profiling with high sensitivity and low operational effort. Furthermore, insight into the migration behavior of Me+ during O2+ sputtering is given by switching the sputter beam from O2+ to O2 clusters and vice versa. K+ is found to be transported through the SiO2 layer only within the proceeding sputter front. For Na+ a steadily increasing fraction is observed, which migrates through the unaffected SiO2 layer toward the adjacent Si/SiO2 interface. he test of inorganic bulk materials or thin dielectric films on the presence of impurities or trace elements is a frequently encountered analytical task in the semiconductor industry. Many reliability issues in electronic devices arise due to alkali metal ions (Me+), which are located within dielectric layers. The most prominent representatives are Na+ and K+. Regarding such issues, time-of-flight secondary ion mass spectrometry (ToF-SIMS) has become one of the major techniques for chemical analysis. Its outstanding detection limits and its ability to record depth profiles are the main advantages concerning a variety of analytical questions. Depending on the investigated element, one can drastically enhance its ionization probability for the generation of secondary ions (SI). The SI signal intensity for depth profiles can be increased, for example, by choosing a proper projectile for sample erosion. After a series of recoils, caused by the impacting primary ion, alkali metals tend to generate positive ions upon the escape from the sample surface. Because of its reactive nature and its high electronegativity, O2+ with energies up to a few kiloelectronvolts are typically used as sputter projectile for Me+ detection.1 However, because of the total amount of incident ions per sputter cycle, one has to consider a massive charge-up of the sample surface for dielectric materials. Modern ToF-SIMS instruments are able to counteract most of the upcoming charging problems during depth profiling by choosing appropriate measurement conditions in dual beam mode and by using a low energy electron shower.2 For depth
T
© XXXX American Chemical Society
profiling applications, where the relevant information is not located in near surface regions, the sample can additionally be covered by a thin conductive layer. During ion bombardment auxiliary secondary electrons are produced, which support the compensation of excess charges.3 Unfortunately, this supplementary deposition step can result in an uncontrolled load of environmental impurities on top of the sample surface. An inherent problem during positive sputter-ion bombardment is the generation of an electric field within the dielectric layer, due to insufficient charge compensation at the sample surface. Once hit by the sputter projectile, charge transfer mechanisms may take place.4,5 Positive charges are generated in the topmost sample region and alkali metals are released from potential binding partners. As a result, subsequent migration processes of positively charged Me+ toward the cathode occur. Especially for nanometer-thick dielectric films, typical measurement conditions for O2+ sputtering can cause electric field strengths of up to MV cm−1. As a consequence, severely distorted Me+ distributions are measured, easily leading to misinterpretations of the obtained results. In literature many approaches can be found as countermeasures for numerous SIMS techniques.6−10 However, most of the proposed methods are either very time-consuming due to the necessity of Received: October 28, 2016 Accepted: January 25, 2017 Published: January 25, 2017 A
DOI: 10.1021/acs.analchem.6b04222 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
out by two independently operating ion beams: 1 keV O2+, using a dual source column with a sputter current of about 248 nA, or 20 keV O2 clusters, using a GCIB column with a sputter current of 5.1 nA. The single charged O2 clusters exhibited an average cluster size of about 1300 molecules per cluster. The sputter crater size was set to 300 × 300 μm2. Depth profiles were acquired in noninterlaced mode using a sequence of 1 analysis frame, 1 sputter frame (meander shaped) and a pause of 1 s. All measurements were carried out at room temperature, including O2 gas flooding (vacuum level in idle status = 8 × 10−7 mbar) and low energy electron flooding for charge compensation. Intensities of recorded ToF-SIMS depth profiles were converted into a concentration scale by considering the total ion dose given by actual implantation parameters. Environmental impurities, as detected by reference SiO2 layers, were taken into account. Depth calibration was undertaken by using the 30Si+ signal as depth position marker, indicating the material interfaces for the beginning, as well as the end of the SiO2 layer. The sputter rate was individually determined for the two different sputter projectiles and subsequently used for depth calibration. The erosion rate of Si was obtained from a bare Si reference sample by crater depth measurements using a digital holographic microscope (DHM R1000, Lyncée tecDHM, Lausanne, Switzerland).
optimized settings, are lacking in sensitivity because of the absence of signal enhancing effects or cannot be implemented in ToF-SIMS depth profiling. A promising approach is the use of a C60 sputter beam, which is proposed to reduce surface charging during depth profiling. A comparable low sputter current is sufficient to obtain a reasonable erosion rate.11 This systematic is also addressed by gas cluster ion beams (GCIB), circumventing the limitations of C60 bombardment.12 In the past decade, large Ar clusters have found a broad field of applications for the analysis of organic materials. These single charged Ar clusters typically consist of more than 1000 Ar atoms per cluster. Therefore, each Ar atom exhibits an impact energy of only a few eV/atom, resulting in almost no damage at the sample surface. Hence, nearly no fragmentation of the investigated compounds occurs.13−15 The benefit of surface charge reduction is thereby even larger than for C60 projectiles. GCIBs can be operated with a variety of gas sources. A beam of large singly charged O2 cluster ions combines the advantage of a GCIB with the remaining signal enhancing effect of O2 for positively charged SI.14 This makes it very attractive as a highly sensitive analysis technique for the determination of Me+ distributions within dielectric materials.
■
EXPERIMENTAL SECTION All wafer manufacturing processes were performed under ISO 4 clean room conditions at Infineon Technologies Austria AG in Villach. A 200-nm-thick SiO2 layer was thermally grown on a Sb-doped Si substrate (ρ < 0.3 Ω·cm) using a vertical furnace at 1000 °C. O2 and HCl served as reactive gases and N2 as inert carrier gas. To ensure electrical contact of the Si substrate the backside of the wafer was metallized via physical vapor deposition technique. Finally, the wafer was cut into 1 × 1 cm2 samples. Ion implantation was performed by Helmholtz Zentrum Dresden in Rossendorf, using three different types of alkali metal ions (Na+, K+, and Li+). Appropriate acceleration energies were chosen to accommodate the entire implantation profile within the 200-nm-thick SiO2 layer, as well as to obtain similar values for projected ion range (Rp) and standard deviation (σ): 7 keV for Li+, 25 keV for Na+, and 45 keV for K+. Note that Rp represents the position of maximum concentration within SiO2 or in other words the average ion implantation depth. σ indicates the broadening of the ion implantation profile. The total implanted Me+ dose was 1 × 1014 cm−2 with an ion beam tilt of 7°. The predicted ion distributions were determined via Monte Carlo simulations using the TRIM code,16 specified by implantation parameters and the SiO2 density (ρ(SiO2) = 2.32 g/cm3). A SiO2 layer without any implanted ions served as reference sample. Optionally, a 40 nm thin Au−Pd layer was sputtered on top of the SiO2 layer via physical vapor deposition. On the one side this metallic conductive layer ensures charge compensation during analysis and on the other side increases the Na+ and K+ contamination level at the sample surface. ToF-SIMS depth profiles were measured using a TOF.SIMS 5 time-of-flight secondary ion mass spectrometer, manufactured by ION-TOF GmbH in Münster, Germany. Measurements were operated in dual beam depth profiling mode and 2 kV extraction voltage for SI. Bi1+ was used as pulsed primary ion projectile with an acceleration energy of 15 keV. Analysis was done in spectrometry mode. (2.4 pA pulsed target current, 6.6 kHz repetition rate) with a field of view of 100 × 100 μm2 and a raster resolution of 128 × 128 pixels. Sputtering was carried
■
RESULTS AND DISCUSSION Depth Profiling: O2+ versus O2 Clusters. To evaluate whether O2-GCIB is suitable for an artifact-free analysis of Me+ distributions within thin dielectric SiO2 layers, a comparison with depth profiles, which are recorded by a common molecular O2+ sputter beam, is undertaken. In Figure 1a and b one can immediately identify the appearing measuring artifact by O2+ depth profiling. Severely distorted Li+ and Na+ distributions are obtained, which are triggered by ion migration. Additionally, an accumulation at the SiO2/Si interface is measured. The K+ signal in Figure 1c is still located almost entirely within the SiO2 layer. Nevertheless, in contrast to simulated data the maximum of the implanted ion distribution has a pronounced offset of about 30 nm. Taking the increased decay length toward the sample bulk into account, it clearly indicates K+ migration during depth profiling. As stated in literature, the mobility of K+ within SiO2 under the influence of an electric field in the MV cm−1 regime is significantly reduced in contrast to Na+ and Li+.17 The electric field arising from surface charges, which are induced during O2+ sputtering, is comparable and thus we suggest the lower extent of migration for K+ is due to its comparable large ionic radius. All graphs in Figure 1 include the expected Me+ distribution upon implantation, as deduced from data simulation. For Na+ and K+ the analyzed ion distributions, measured by O2-GCIB, match perfectly with the simulated ion implantation profiles. Na+ slightly deviates from the simulated ion distribution in the vicinity of the SiO2/Si interface. The Li+ implantation profile exhibits a slightly larger decay. Despite using O2 clusters as sputter projectiles, there obviously still exists a residual ion migration during depth profiling. To assess the occurring residual ion migration for O2-GCIB measurements, Table 1 lists the relative amounts of Me+, which are detected outside of the simulated ion distribution after reaching the simulated Rp. This amount of ions is taken as quantity for the residual migration artifact. Note that any surface contamination, determined by the reference SiO2 layer, is considered within these values. Accordingly, O2+ depth B
DOI: 10.1021/acs.analchem.6b04222 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Table 1. Comparison of the Simulated Data for Alkali Metal Ion Distribution within SiO2 with the Measured ToF-SIMS Depth Profiles, Using O2-GCIB for Sputtering mobile ion
analysis technique
Li+
simulated O2-GCIB simulated O2-GCIB simulated O2-GCIB
Na+ K+
relocated fraction of Me+ (%) 6.62 0.35 0.03
Rp (nm)
σ (nm)
53.0 53.7 49.2 44.5 49.1 44.3
25.7 30.2 21.8 21.9 19.0 19.6
during O2-GCIB depth profiling, resulting in an inaccurate measured ion distribution. In addition to the generation of surface charges the neutralization of an incoming O2+ sputter projectile triggers the mobilization of Me+, which is initially associated with a counterpart with high electronegativity, for example, an oxygen within SiO2.4 Obviously, this ion mobilization still takes place during O2 cluster bombardment, even if the incoming projectile exhibits an average energy of only some eV per O2 molecule. However, to get an erosion rate in SiO2 comparable to O2+ sputtering O2-GCIB requires a factor 50 lower sputter current with the chosen settings for operation. This circumstance is also valid for other cluster projectiles as e.g. C6011 and can be explained by a much larger sputter yield per incoming cluster projectile in comparison to O2+. The total amount of surface charges is considered to be drastically reduced. Hence, the resulting electric field within the SiO2 film is also significantly lowered. As a consequence, Me+ migration is remarkably decelerated by mitigating its driving force. In addition a decrease of charges potentially leads to more efficient charge compensation at the sample surface. Since there evolves no significant atomic mixing zone upon impact of large O2 clusters,14 the ion bombardment leads to a different state of surface near regions in comparison to O2+.18 Thus, the enhanced diffusivity within a disordered topmost sample layer, as observed for O2+ bombardment, does not take place. To show the accuracy of the obtained O2-GCIB depth profiles, Rp and σ are determined for each Me+ distribution in Figure 1, by fitting the recorded data with a Gaussian shaped function. The extracted values are compared to simulated values in Table 1. For Na+ as well as K+ the measured Rp is shifted from the simulated Rp about 5 nm further toward the SiO2 sample surface. This can be explained by the running-in characteristic of the erosion rate, being higher at the beginning of the analysis. Together with a reduced depth resolution of O2GCIB this deviation is reasonable.19 The presented values confirm the ability of an artifact-free measurement of Na+ and K+ distribution within SiO2 by using O2 clusters as sputter projectiles. For Li+, which exhibits a smaller ionic radius compared to Na+ and K+, no offset is observed and the measured and the evaluated Rp fit together perfectly. This can be explained by a residual migration artifact of Li+ toward the SiO 2/Si interface, which is indicated by a broadened implantation profile in Figure 1a and a larger σ in Table 1. Me+ Migration Behavior during O2+ Sputtering. Some years ago the question of the actual location of Na+ during ToF-SIMS depth profiling using O2+ as sputter projectiles was already addressed by Krivec et al.6 Experiments, comprising a decelerated ion migration, were undertaken by reducing the sample temperature. Considering the generated atomic mixing
Figure 1. ToF-SIMS depth profiles of (a) Li+, (b) Na+, and (c) K+ implanted SiO2 layers using different sputter beams: O2+ (blue solid lines) and O2 clusters (red solid lines). Reference SiO2 layer with no ions implanted, measured by O2 clusters (dashed lines), and simulated ion distribution (dotted lines).
profiling for Li+ as well as Na+ leads to 100% ion migration. The calculated fractions show an increase of ion migration according to Li+ < Na+ < K+. This trend suggests that the migration artifact depends on the ionic radius of the migrating species. Via sputtering with O2 clusters a nearly migration free depth profile of Na+ and K+ with acceptable uncertainties (