Proximity gettering of silicon wafers using CH3O multielement molecular ion implantation technique

We investigated the fundamental characteristics of a CH3O-ion-implanted silicon epitaxial wafer with our previously developed multielement molecular ion implantation technique and compared this technique with a conventional implantation technique, i.e., “carbon cluster ion implantation”. We found that the CH3O ion projection range has a 10-fold higher oxygen concentration than the carbon cluster ion projection range after epitaxial growth. We also found 50 nm silicon {111} stacking faults in the CH3O ion projection range. Such defects were not observed in the carbon cluster ion projection range. From nickel gettering test results, proximity gettering of nickel contaminations by CH3O ion implantation was found to be more effective than that by C2H3 ion implantation. Therefore, we speculate that the CH3O ion projection range improves the gettering capability of metallic impurity contaminants through the formation of complex point defects formed by vacancies and that oxygen implanted at a high concentration and silicon {111} stacking faults are new gettering sinks.


Introduction
Advanced CMOS image sensors currently require high imaging sensitivity and device performance. However, one of the most important technological issues is the removal of metallic impurity contaminants from the device active range during CMOS image sensor fabrication. [1][2][3][4][5] To address this issue, it is strongly required that a silicon wafer should have a high gettering capability.
To import the gettering capability of silicon epitaxial wafers, intrinsic gettering (IG) and extrinsic gettering (EG) techniques are mainly used in CMOS device fabrication. Figure 1 shows a schematic image of these gettering techniques. The IG technique involves the formation of an oxygen precipitate as a gettering sink during device heat treatment. [6][7][8][9] The EG technique involves the deposition of a polysilicon layer used as a gettering sink on the back of a silicon wafer. 10,11) However, the temperature of the current device fabrication process continuously decreases year by year. In the case of the IG technique, the oxygen precipitate cannot form sufficient amounts of gettering sinks. In the case of the EG technique, the polysilicon layer is far from the device active region. Thus, it is difficult to diffuse the metallic impurity to the gettering sinks. For these reasons, a conventional gettering technique cannot perform sufficient gettering in the CMOS image sensor manufacturing process.
To solve this technical issue, we developed proximity gettering wafers called "carbon-cluster-ion-implanted epitaxial silicon wafers". [12][13][14][15][16] In our previous study, we demonstrated that carbon-cluster-ion-implanted epitaxial silicon wafers have three useful characteristics for advanced CMOS image sensors, namely, strong gettering capability for metallic impurity contaminants, out-diffusion barrier effect on oxygen impurities from a silicon substrate, and hydrogen passivation effect. 14) Although these wafers are useful for improving the performance of CMOS image sensors, the demand for highperformance wafers is expected to continue in the future. Therefore, silicon epitaxial wafers require a much higher metallic impurity gettering capability. To obtain epitaxial silicon wafers with higher gettering capability, we focused on ion implantation techniques.
In general, it is known that the defects formed by ion implantation function as gettering sinks. Kuroi and coworkers reported on a gettering technique involving high-energy ion implantation using various elements such as oxygen, silicon, and boron. [17][18][19] The origin of this gettering sink is the secondary defect induced by high-energy ion implantation. On the other hand, Okuyama et al. reported that the carbon cluster ion implantation does not form secondary defects in the projection range after heat treatment. 16) The gettering sink induced by carbon cluster ion implantation is a carbon and interstitial silicon complex formed in the carbon cluster ion projection range. From these studies, it is considered that ion implantation techniques can form various types of gettering sinks depending on implantation elements or conditions.  Content from this work may be used under the terms of the Creative Commons Attribution 4.0 license. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
The conventional carbon cluster ion implantation technique implants cluster ions consisting of carbon and hydrogen. To improve this technique, we have developed a new molecular ion implantation technique called "multi element molecular ion implantation". 20) This technique characterized that implant molecular ions consist of three elements with one additional element in addition to carbon and hydrogen. In this article, we use CH 3 O ion implantation. We expect that a CH 3 O-ion-implanted silicon epitaxial wafer will have a higher metallic impurity contamination gettering capability with the formation of new defects or complexes, while retaining the characteristics of conventional carbon-clusterion-implanted epitaxial silicon wafers.
In this study, we demonstrate the diffusion behavior of implanted elements and the defect formation behavior in the CH 3 O ion projection range after epitaxial growth, and also the gettering capability test on nickel. 20) 2. Experimental methods Figure 2 shows the sample preparation process used in this study. 12 in. n-type (100) phosphorus-doped Czochralski (CZ) silicon single-crystal wafers were implanted with CH 3 O multielement molecular ions and carbon cluster C 2 H 3 ions at room temperature. The CH 3 O and C 2 H 3 ion implantations used Nissin's ion implanter "CLARIS ® ". [21][22][23] The implantation energy was 80 keV=ion. The implantation dose of the CH 3 O ions was 5.0 × 10 14 -2.0 × 10 15 ions=cm 2 converted to a carbon dose of 5.0 × 10 14 -2.0 × 10 15 carbon atoms=cm 2 . The implantation dose of C 2 H 3 ions was 3.75 × 10 14 -5.0 × 10 14 ions=cm 2 converted to a carbon dose of 7.5 × 10 14 -1.0 × 10 15 carbon atoms=cm 2 . This carbon dose is the same as 7.5 × 10 14 -1.0 × 10 15 ions=cm 2 in CH 3 O ions. The ion beam currents of CH 3 O and C 2 H 3 were the same, i.e., 550 µA. The epitaxial growth of the samples was conducted after CH 3 O ion and carbon cluster ion implantation. The epitaxial thickness was 9.0 µm. For the nickel gettering test, the sample wafers were contaminated with nickel metallic impurities (2.5 × 10 13 atoms=cm 2 ) by spin coating with a metallic-impurity-contaminated acid solution. The surface initial nickel concentration was measured by inductively coupled plasma mass spectrometry (ICP-MS) after spin coating. After the metallic impurity contamination, the wafers were heated at 900°C for 1800 s for nickel diffuse treatment.

Sample preparation
The distributions of carbon, hydrogen, oxygen, and nickel were analyzed by secondary ion mass spectrometry (SIMS). The defects in the multielement molecular cluster ion projection range were analyzed by transmission electron microscopy (TEM).

Results and discussion
3.1 SIMS depth profile of carbon, hydrogen, and oxygen after epitaxial growth Figure 3 shows the carbon, hydrogen, and oxygen distribution-depth profiles of CH 3 O-ion-implanted silicon epitaxial wafers. The implantation energy was 80 keV and the implantation dose was 1.0 × 10 15 carbon atoms=cm 2 . The epitaxial layer thickness was 9 µm. From these results, the CH 3 O ion projection range had concentration peaks of carbon, hydrogen, and oxygen.
The peak concentration of the carbon impurity in the CH 3 O ion projection range was higher than the solid solubility of carbon in the silicon single-crystal wafers, which was the same as in the carbon cluster ion projection range. 18) Kurita and coworkers reported that a high concentration of carbon is observed after epitaxial growth in the carbon cluster ion projection range, and an aggregate of carbon is formed in this range as a gettering sink. 14,16) Therefore, it is speculated that the CH 3 O ion projection range has almost the same characteristics as the carbon cluster ion projection range. In the CH 3 O ion projection range, carbon and oxygen concentration peaks were approximately at the same concentration. Such a trend was not observed in the carbon cluster ion projection range. Figure 4 shows the distribution-depth profiles of oxygen in CH 3 O-and C 2 H 3 -ion-implanted silicon epitaxial wafers. The carbon dose was the same in both CH 3 O and C 2 H 3 ions, i.e., 1.0 × 10 15 carbon atoms=cm 2 . As shown in Fig. 4, the oxygen peak concentration of the CH 3 O-ion-implanted sample was  The blue line is carbon, the black line is hydrogen, and the red line is oxygen.
6.5 × 10 19 atoms=cm 3 . The oxygen peak concentration of the C 2 H 3 -ion-implanted sample was 3.1 × 10 18 atoms=cm 3 . Thus, the oxygen peak concentration of the CH 3 O-ion-implanted sample was up to more than 10-fold higher than that of the C 2 H 3 -ion-implanted sample. The oxygen concentration of the CH 3 O-ion-implanted sample in the epitaxial layer was up to approximately 2-fold higher than that of the C 2 H 3 -ionimplanted sample. These results suggest that almost all the oxygen implanted by CH 3 O ions is fixed in the CH 3 O ion projection range. The detailed mechanism of the implanted oxygen behavior is not clear. However, it is assumed that there are two mechanisms.
First, the VO complex forms in the CH 3 O ion projection range. Shirasawa et al. calculated, by DFT calculation, the possibility of VO complex formation in the carbon cluster ion projection range. 26,27) The multielement molecular ion implantation forms high-concentration vacancy by collision with the silicon crystal. Thus, the multielement molecular ion projection range is considered to have a high VO concentration. Therefore, owing to VO complex formation, it is assumed that implanted oxygen is fixed in this projection range.
Second, oxygen is gettered in the carbon projection range. A previous study reported that the carbon cluster ion projection range can getter oxygen impurity. 14) A carbon cluster ion contains carbon and hydrogen, and a CH 3 O ion contains carbon and hydrogen. Thus, it is assumed that there is a possibility that oxygen is gettered in the CH 3 O ion projection range. However, the oxygen concentration in the CH 3 O ion projection range is more than 10-fold higher than that in the C 2 H 3 ion projection range. Thus, we speculate that a new complex different from that in the carbon cluster ion projection range is formed after epitaxial growth in the CH 3 O ion projection range. Figure 5 shows the cross-sectional TEM images of the CH 3 O ion projection range. The implantation energy was 80 keV, and the implantation dose was 5.0 × 10 14 -2.0 × 10 15 carbon atoms=cm 2 . In these images, approximately 50 nm defects were observed in the samples with an implantation dose of 7.5 × 10 14 -2.0 × 10 15 carbon atoms=cm 2 in the CH 3 O ion projection range. However, there were no such large defects at the dose of 5.0 × 10 14 carbon atoms=cm 2 . These results suggest that approximately 50 nm defect formation has dose dependence. This tendency is thought to be related to ionimplantation damage. We previously reported that the critical dose of CH 3 O ion implantation is approximately 1.0 × 10 15 ions=cm 2 . 23) Thus, it is considered that excessive damage given to the silicon crystal by CH 3 O ion implantation is one of the causes of secondary defects. Figure 6 shows a high-resolution cross-sectional TEM image of the CH 3 O ion projection range. The implantation energy was 80 keV=ion and the implantation dose was 1.0 × 10 15 carbon atoms=cm 2 . The interface of the epitaxial layer and Si substrate is indicated by the black dashed line in Fig. 6. As shown in Fig. 6, the projection range of CH 3 O ions was observed at a depth of approximately 120 nm from the interface of the epitaxial layer and Si substrate. Furthermore, two types of defects (5 and 50 nm defects) were observed in the CH 3 O ion projection range. The 5 nm defect was observed at a depth of approximately 80 nm from the interface of the epitaxial layer and Si substrate. On the other hand, the 50 nm defect was observed at a depth of approxi-   mately 160 nm from the interface of the epitaxial layer and Si substrate. Therefore, the 50 nm defect formation region was deeper than the 5 nm defect formation region in the CH 3 O ion projection range. Defects similar to the 5 nm defect were observed in the carbon cluster ion projection range. 16,18) The formation of such defects is due to carbon implantation and the defects act as gettering sinks. A CH 3 O ion contains carbon atoms. Hence, the 5 nm defect in the CH 3 O ion projection range is considered to be due to carbon implantation and to act also as the gettering sink. From this result, it is assumed that the CH 3 O ion projection range has at least the same degree of gettering capability as the carbon cluster ion projection range. Defects different from those were observed in the CH 3 O ion projection range. The 50 nm defect is the new type of defect. Okuyama et al. reported the defect formation in the carbon cluster ion projection range. 16) In the sample with high carbon cluster ion dose, the carbon cluster ion projection range forms a 40 nm defect after epitaxial growth. This defect is due to the recrystallization of the amorphous region formed by ion implantation and is located on the surface layer side of the carbon-cluster-implanted region. However, as shown in Fig. 6, the 50 nm defect is located at the back side of the CH 3 O ion projection range. Thus, it is considered that 50 nm defects are peculiar to the CH 3 O ion projection range. The formation of 50 nm defects is considered to be the effect of adding oxygen to the ions. To analyze the origin of 50 nm defects, we conducted high-resolution TEM observation and fast Fourier transform (FFT) analysis. Figure 7(a) shows a high-resolution cross-sectional TEM image of 50 nm defects in the CH 3 O ion projection range. There seem to be two types of 50 nm defects, i.e., round and linear in the silicon {111} direction. However, round defects are assumed to be silicon {111} stacking faults in the direction perpendicular to the TEM analysis direction. Therefore, it is assumed that round and linear defects are the same. Therefore, we conducted a detailed analysis of linear defects.

TEM observation of CH 3 O ion projection range
For the detailed analysis of this type of defect, we examine an FFT pattern at the edge of a defect. An enlarged TEM image of the defect edge and FFT pattern is shown in Fig. 7(b). Figure 7(c) shows a defect image obtained from the FFT pattern by extracting only the silicon {111} direction component. From Fig. 7(c), the 50 nm defects are the stacking faults inserted with one atomic layer in the silicon {111} direction. Okuyama et al. reported that the 40 nm defects in the carbon cluster ion implantation region are silicon single crystals from the electron diffraction pattern. 16) From the difference in diffraction pattern, the defect in the CH 3 O ion implantation region is thought to be also different from that formed in the carbon cluster ion implantation region. A 50 nm defect is a defect peculiar to the CH 3 O ion implantation region.
To consider the origin of 50 nm defects, we calculated and compared the damage distribution profiles after C 2 H 3 and CH 3 O ion implantations by technology computer aided design (TCAD) simulation with the Sentaurus process simulator from Synopsys. 24) Figure 8 shows the distribution profiles of carbon, hydrogen, oxygen, and damage calculated by TCAD simulation. The implanted energy was 80 keV. CH 3 O and C 2 H 3 were implanted at a dose of 5.0 × 10 14 ions=cm 2 . Tilt and twist were 0 degrees. The damage concentration was calculated using the Frenkel pair model that defines interstitial silicon (I) and vacancy (V) concentrations generated by the ion implantation. At a high implantation dose, the damage concentration saturated owing to the supersaturated generation of I and V. In this situation, it was not possible to compare the damage profiles of CH 3 O and C 2 H 3 . Therefore, the distribution profiles were calculated at an implantation dose of 5.0 × 10 14 ions=cm 2 to compare the damage profiles of the ion implantation regions. As shown in Fig. 8, the damage concentration increased in CH 3 O ion implantation, even at the same ion dose. Since CH 3 O is an ion in which one carbon atom of C 2 H 3 is replaced with oxygen, the impact of oxygen in the ion implantation damage is extremely large.  From Fig. 8, it was found that CH 3 O ion implantation has large damage in the region deeper than the carbon concentration peak of 160 nm depth as compared with C 2 H 3 implantation. The damage profile indicated that CH 3 O ion implantation results in higher I and V concentrations than C 2 H 3 implantation. The increase in V concentration seems to form a VO complex with O of CH 3 O. Therefore, it is considered that an excessive amount of I remained without recombination with V during heat treatment in epitaxial growth. A previous study indicated that an excessive amount of I induced by ion implantation forms the defect called the "endof-range defect". 25) Therefore, the origin of 50 nm defects is considered to be aggregates of interstitial silicon formed in deep regions by additional oxygen implantation in CH 3 O ion implantation. This inference is consistent with the finding that the 50 nm defect is an intrinsic stacking fault as shown in Fig. 7. Furthermore, it is in agreement with the finding that 50 nm defects are located behind carbon complex defects as shown in Fig. 6.
It is expected that the gettering capability of metallic impurity will be improved by the formation of 50 nm defects and a VO complex using CH 3 O ion implantation. Figure 9 shows the SIMS depth profiles measured on the CH 3 O-ion-implanted epitaxially grown samples (CH 3 O ions were implanted at 80 keV at a carbon dose of 7.5 × 10 14 carbon atoms=cm 2 ) after the nickel metallic impurity contamination.

Nickel gettering capability improvement using CH 3 O ion implantation
As shown in Fig. 9, the carbon peak concentration was 6 × 10 19 atoms=cm 3 in the CH 3 O ion projection range. Moreover, the oxygen peak concentration in this projection range was 3 × 10 19 atoms=cm 3 . These results suggest that carbon and oxygen impurities remain at concentrations exceeding the solid solubility in the silicon single-crystal wafers after nickel metal impurity contamination. We also confirmed that nickel impurities are gettered in the CH 3 O ion projection range. Therefore, the CH 3 O ion projection range can getter nickel metallic impurities. Figure 10 shows the SIMS depth profiles measured on CH 3 O-and C 2 H 3 -ion-implanted epitaxially grown samples (CH 3 O and C 2 H 3 ions were implanted at 80 keV at a carbon dose of 7.5 × 10 14 carbon atoms=cm 2 ) after nickel metallic impurity contamination. The nickel contamination surface concentration measured by ICP-MS was 2.5 × 10 13 atoms=cm 2 . As shown in Fig. 10, nickel concentration peaks were observed in the CH 3 O and C 2 H 3 ion projection ranges. From these results, both CH 3 O and C 2 H 3 ion implantations could getter nickel contaminants. The nickel concentration in the CH 3 O ion projection range was 2.0 × 10 13 atoms=cm 2 and that in the C 2 H 3 ion projection range was 1.8 × 10 13 atoms=cm 2 . Figure 11 shows the nickel gettering efficiencies of the CH 3 O-and C 2 H 3 -ion-implanted samples after nickel contamination. The nickel gettering efficiency was calculated using the following equation, where η eff is the gettering efficiency, C ni is the nickel concentration in the ion projection range, and C init is the initial surface nickel concentration.
As shown in Fig. 11, the nickel gettering efficiency of the CH 3 O-ion-implanted sample improved by 5.7% compared with that of the C 2 H 3 -ion implanted sample. Thus, CH 3

O ion
Carbon Oxygen ---Nickel  implantation can import a higher nickel gettering capability than C 2 H 3 ion implantation. The details of the gettering mechanism in the CH 3 O ion projection range are not clear, although this improvement in gettering efficiency is thought to be the effect of the three types of estimated gettering sinks below. First, the carbon projection range has a gettering capability on nickel. Figure 12 shows carbon and oxygen distributiondepth profiles after nickel metallic contamination. As shown in Fig. 12, the carbon peak concentration in the CH 3 O ion projection range was 5.5 × 10 19 atoms=cm 3 and that in the C 2 H 3 ion projection range was 4.7 × 10 19 atoms=cm 3 . These results suggest that implanted carbon remains in the C 2 H 3 and CH 3 O ion projection ranges after nickel metallic contamination. Kurita et al. reported that the high-concentration carbon region, which was formed by carbon cluster ion implantation, has a high gettering capability for various metals. 14) As shown in Fig. 12, the CH 3 O ion projection range had the same ordered carbon concentration as in the carboncluster (C 2 H 3 ) ion projection range. Furthermore, as shown in Fig. 6, the CH 3 O ion projection range had the same defects as the carbon cluster ion projection range. Thus, it is conceivable that the CH 3 O ion projection range has at least the same function as the carbon cluster ion projection range.
Second, the silicon {111} defects in the CH 3 O ion projection range function as gettering sinks. As shown in Figs. 6 and 7, the CH 3 O ion projection range had 50 nm defects. Kuroi et al. reported that secondary defects formed by highenergy ion implantation function as gettering sinks. 19) The CH 3 O ion implantation condition is not high energy. However, it is sufficiently conceivable that defects, as shown in Fig. 9, function as new gettering sinks, and the nickel gettering capability improves.
Third, the VO-complex point defects in the CH 3 O ion projection range function as gettering sinks. As shown in Fig. 12, carbon concentrations in the CH 3 O and C 2 H 3 ion projection range were almost the same. However, the oxygen peak concentration in the CH 3 O ion projection ranges was approximately 10-fold higher than that in the C 2 H 3 ion projection range (oxygen concentration in the CH 3 O ion projection range was 3.0 × 10 19 atoms=cm 3 and that in the C 2 H 3 ion projection range was 1.5 × 10 18 atoms=cm 3 ). Shirasawa et al. used DFT to determine the binding energy between metallic impurity and point defects. 26,27) In the calculation results, the carbon-interstitial silicon (C-I), vacancy-hydrogen (VH n , n = 1 to 3), or VO complex has a high binding energy with metallic impurities. In the case of CH 3 O ion implantation, oxygen was implanted at a high concentration, and vacancy was formed by CH 3 O ion implantation. Thus, it is expected that the CH 3 O ion projection range will have a high-concentration VO complex and will improve the nickel gettering capability through VO-complex point defect formation.
From these types of gettering sinks, it is considered that the nickel gettering capability in the CH 3 O ion projection range is improved by the coexistence of the carbon cluster ion projection range and silicon {111} or VO-complex point defects.

Conclusions
We succeeded in epitaxial growth on the top surface of a CH 3 O-ion-implanted silicon wafer. From SIMS analysis results, the oxygen concentration peak in the CH 3 O ion projection range was 10-fold higher than that in the C 2 H 3 -ion projection range. These results also suggest that the CH 3 O ion projection range forms different chemical bonding states form the carbon cluster ion projection range.
From TEM observation results, the CH 3 O ion projection range had two types of defects. The first type is an approximately 5 nm defect, which is the same as that observed in the carbon cluster ion projection range. These results indicate that the CH 3 O ion projection range has at least the same gettering capability for metallic impurity as the carbon cluster ion projection range. In the CH 3 O ion projection range, approximately 50 nm defects, which were not observed in the carbon cluster ion projection range, were observed. From high-resolution TEM observation and FFT analysis results, it was found that the defects are silicon {111} one-layer stacking faults. From the nickel gettering test results, the CH 3 O ion projection range was found to improve the nickel gettering capability compared with the C 2 H 3 -ion projection range. We speculated that the CH 3 O ion projection range improves the gettering capability through the formation of silicon {111} stacking faults as new gettering sinks.