Hydrogen passivation for reduction of SiO2/Si interface state density using hydrocarbon-molecular-ion-implanted silicon wafers

The reduction in the density of SiO2/Si interface state (Dit) in the isolation region and transfer transistor gate oxide is necessary to improve the performance of complementary metal-oxide semiconductor (CMOS) image sensors. In this study, we demonstrated that a hydrocarbon-molecular-ion-implanted epitaxial silicon wafer can reduce the Dit and Pb0 center density in SiO2/Si interface regions analyzed by quasi-static capacitance–voltage and electron spin resonance measurements, respectively. The Dit and Pb0 center density of wafers without hydrocarbon molecular ions increased after annealing at 700 °C. On the other hand, the Dit and Pb0 center density of wafers implanted with hydrocarbon molecular ions decreased after annealing at 700 °C. We also estimated the activation energy to be 1.67 eV for the hydrogen termination reactions with hydrogen molecules and Si dangling bonds at the SiO2/Si interface. The termination effects of the hydrocarbon-molecular-ion-implanted epitaxial silicon wafers can contribute to the high electrical performance of CMOS image sensors.


Introduction
Complementary metal-oxide semiconductor (CMOS) image sensors have been widely used in ubiquitous devices such as smartphones and smartwatches. Consumer markets strongly demand for high sensitivity and speed image data processing for fabricating high-performance CMOS image sensors. [1][2][3][4][5][6] However, there are some serious technical issues in manufacturing high-performance CMOS image sensors, namely, dark current noise, fixed-pattern noise, and random telegraph noise generated from CMOS image sensor circuits.
The origins of noise are SiO 2 /Si interface states at transfer gate oxides, local oxidation of silicon, shallow trench isolation, and deep trench isolation. [7][8][9] The interface states form deep energy levels in the silicon bandgap, which can act as generation-recombination (G-R) centers. The G-R centers strongly affect the electrical performance of CMOS image sensors. Therefore, CMOS image sensor manufacturers try to reduce noise from various parts of sensor circuits using optimum circuit designs and device fabrication processes. Previously, some circuit design researchers proposed a new circuit design concept for noise reduction from CMOS image sensor circuits. 10,11) However, it is insufficient to improve the noise reduction performance. Another solution is low-temperature hydrogen annealing after the back-end-of-line (BEOL) process. 9,12,13) We call this forming gas annealing (FGA) treatment. However, recently, multifilm formation by atomic layer deposition has been widely applied to the BEOL process for the formation of interconnect layers. Hydrogen does not reach the SiO 2 /Si interface because the deposited interconnect layers act as a diffusion barrier. FGA is not effective for solving the above-mentioned technical issue. A previous study has shown the reduction in interface state density (D it ) at the SiO 2 /Si interface by forming a SiO 2 layer after monomer hydrogen ion implantation. 14,15) However, since high-temperature annealing is repeatedly performed after forming the SiO 2 layer, a concern is that D it increases when hydrogen at the SiO 2 /Si interface is released again during high-temperature annealing. In addition, hydrogen does not remain in the silicon wafer after annealing owing to its high diffusion velocity. It is necessary to repeat the step of implanting hydrogen ions, which increases the device fabrication cost. Thus, the reduction in D it by monomer hydrogen ion implantation is difficult to apply to any device process. Therefore, we have developed an alternative solution to the technical issue encountered in CMOS image sensor fabrication by hydrocarbon molecular ion implantation. [16][17][18][19][20][21][22][23][24] We have demonstrated the fabrication of a novel silicon wafer with three unique characteristics, namely, the high gettering capability for metallic impurities, the oxygen out-diffusion barrier effect and the hydrogen storage effect in the projection range of hydrocarbon molecular ions. We have already reported that our novel silicon wafer leads to a marked improvement of CMOS device key parameters, such as low dark current and white spot defect density when using the CMOS image sensor manufacturing line. 18) In addition, in a previous study, the diffusion behavior of two types of hydrogen was investigated in the hydrocarbonmolecular-ion-implanted region during isochronal annealing. 24) One is molecular hydrogen (H 2 ) that outdiffused to the silicon lattice from the hydrocarbon-ionimplanted region during annealing below 800°C. The other is atomic hydrogen (H) that out-diffused to the silicon lattice from the hydrocarbon-molecular-ion-implanted region during annealing above 800°C. The study suggests that these two types of hydrogen with different diffusion behaviors can reduce D it at the SiO 2 /Si interface. However, it did not clearly show the hydrogen termination mechanism at the SiO 2 /Si interface after isochronal annealing.
Therefore, in this study, we focus on the mechanism of hydrogen termination of the SiO 2 /Si interface state after isochronal annealing analyzed by capacitance-voltage (C-V ) and electron spin resonance (ESR) measurements. Previous studies have shown that D it increases owing to the dissociation of hydrogen from the SiO 2 /Si interface, which is predominant during annealing at 500°C or higher. Therefore, annealing above 500°C could not be performed after FGA. On the other hand, it has been reported that hydrogen dissociates from the hydrocarbon-molecular-ionimplanted region by annealing at 500°C or higher. In other words, hydrogen will be supplied to the SiO 2 /Si interface from the hydrocarbon-molecular-ion-implanted region even during annealing at 500°C or higher. Therefore, we aimed to clarify the hydrogen termination effect of the hydrocarbonmolecular-ion-implanted wafer at annealing temperatures higher than 500°C. Figure 1 shows the schematics of the flow of sample preparation and the sample structure. A p-type (100) wafer with a resistivity of 10 ohm cm was implanted with C 3 H 5 ions at a dose of 3.3 × 10 14 molecular ions cm −2 at room temperature. The C 3 H 5 implantation energy was 80 keV. The projection range of C 3 H 5 under this condition was approximately 80 nm from the surface. Then, 5 μm thick p-type silicon epitaxial layers were grown with a resistivity of 10 ohm cm at 1100°C. The same epitaxially grown samples without C 3 H 5 implantation were prepared to compare the hydrogen termination effect of C 3 H 5 . Dry oxidation for 100 min at 900°C was performed for the growth of SiO 2 layers of 25 nm thickness, as shown in Fig. 1. Then, the samples were irradiated with an electron beam of 800 keV at a dose of 2.0 MGy to increase D it at the SiO 2 /Si interface, because D it obtained by C-V and ESR measurements was below the detection limit with or without C 3 H 5 implantation immediately after the SiO 2 layer growth. The samples irradiated with the electron beam were isochronally annealed at 500°C, 600°C, 700°C, 800°C, and 900°C for 10, 30, 60, and 120 min in nitrogen ambient. In addition, the MOS structure for C-V measurement was prepared by depositing 100 nm thick Al contacts on the oxide surface by electron beam evaporation using a mechanical mask. The Al electrode area was 6.4 × 10 −3 cm 2 . Post metallization annealing was performed at 400°C for 30 min in 1 atm nitrogen ambient. D it at the SiO 2 /Si interface was determined by standard quasistatic CV measurement. 25,26) ESR analysis of the samples was conducted to determine the density of Pb centers. [27][28][29][30] 3. Results and discussion Figure 2 shows the high frequency (100 kHz) and quasi-static C-V curves before electron beam irradiation (black dotted line) without C 3 H 5 implantation and after electron beam irradiation without (blue dashed line) and with (red solid line) C 3 H 5 implantation. There was a concern that an inversion layer would be formed in this structure without a channel stop layer. However, the high-frequency C-V curves show almost the same value of C min . Thus, it was considered that the inversion layer did not affect the D it evaluation. On the other hand, the measured quasi-static C-V curves of wafers with and without C 3 H 5 implantation were found to be different in Fig. 2. These results indicate that D it at the SiO 2 /Si interface can be calculated by the standard quasistatic method. 25,26) Figure 3 shows the D it of wafers after annealing at (a) 500°C and (b) 700°C without (blue squares) and with (red diamonds) C 3 H 5 implantation and before annealing without (black circles) C 3 H 5 implantation. D it at  the midgap of the wafer with C 3 H 5 implantation was lower than that without C 3 H 5 implantation after annealing at 700°C

Experimental procedure
. A previous study demonstrated that the D it of the sample with hydrogen FGA increases with annealing above 600°C because hydrogen dissociates from the SiO 2 /Si interface during high-temperature annealing. 8) However, in this study, the D it of the C 3 H 5 -implanted wafer decreased after annealing at 700°C. The result demonstrated that the D it of the C 3 H 5 -implanted wafer may be decreased by hydrogen outdiffusion from the C 3 H 5 -implanted region after high-temperature annealing. Then, the dependence of D it on annealing temperature was analyzed by the quasi-static C-V method. Figure 4 shows D it at the midgap of the wafers without and with C 3 H 5 implantation after annealing. The D it of the wafers without C 3 H 5 implantation decreased initially after annealing at 500°C, but increased after annealing at 900°C. In contrast, the D it of the C 3 H 5 -implanted wafer decreased with increasing annealing temperature. A previous study showed that the concentration of hydrogen that out-diffused from the C 3 H 5 -implanted region increases with annealing temperature. Thus, the D it of the C 3 H 5 -implanted wafer was lower after annealing at 900°C than after annealing at 500°C. We consider that the increase in the concentration of hydrogen that out-diffused from the C 3 H 5 -implanted region was larger than the increase in D it during annealing at 900°C. Consequently, the reduction in D it at the SiO 2 /Si interface after annealing at 900°C is considered to be the hydrogen termination effect of the C 3 H 5 -implanted epitaxial silicon wafer.
Then, we analyzed the C-V-measured samples by ESR measurement to identify the defects at the SiO 2 /Si interface and determine their relationship with the reduction in D it . ESR measurement is used to analyze the density of Pb centers, which are also known as SiO 2 /Si interface state defects. [27][28][29] Most ESR studies of the SiO 2 /Si interface have dealt with Pb center defects. 29,30) One of the Pb center signals was the Pb 0 center at g values of 2.0060-2.0062 for each annealed sample. A Pb 0 center has a structure of three silicon bonds: ·Si≡Si 3 . 31) In addition, to verify the relationship between D it and the density of Pb 0 centers, we plotted D it at the midgap versus density of Pb 0 centers at each annealing temperature in Fig. 5. The results indicate that the density of Pb 0 centers determined by ESR measurement is proportional to D it obtained by CV measurement. In this study, electron beam irradiation was performed to increase D it and clarify the hydrogen termination effect on the C 3 H 5 -implanted epitaxial wafer. As shown in Fig. 5, D it and Pb 0 center density were increased by electron beam irradiation, which can be reduced by using a C 3 H 5 -implanted silicon epitaxial wafer. Figure 5 also shows that the D it in the wafer without C 3 H 5 implantation is also proportional to the density of Pb 0 centers.
Then, we discuss the increases or decreases in D it and the density of Pb 0 centers in wafers without and with C 3 H 5 implantation after isochronal annealing. Firstly, we consider the change in the D it of wafers without C 3 H 5 implantation. A previous study showed that D it decreases in nitrogen ambient at 500°C. 32) Here, the initial decrease in the D it of wafers without C 3 H 5 implantation at 500°C may be a passivation effect of the thermal strain induced by thermal stress in nitrogen ambient. 33) However, it has been reported that D it is   lower after hydrogen FGA than after annealing in nitrogen ambient. The increase in the D it of wafers without C 3 H 5 implantation at 700°C can be explained by two possibilities. One is D it generation due to Si-O bond breaking induced by the compressive stress during annealing. 33) The other is the dissociation of hydrogen or nitrogen at the SiO 2 /Si interface during annealing. 32) On the other hand, the D it of the C 3 H 5 -implanted wafers decreased with increasing annealing temperature. The results suggest that the hydrogen that outdiffused from the C 3 H 5 -implanted region can terminate dangling bonds of silicon at the SiO 2 /Si interface during annealing at 700°C. Probably, the hydrogen that terminated interface state defects should also dissociate from the SiO 2 /silicon interface of the C 3 H 5 -implanted wafer. However, it is considered that the concentration of hydrogen that out-diffused from the C 3 H 5 -implanted region is higher than that of the dissociated hydrogen during annealing at 700°C. Indeed, a previous study has demonstrated that the concentration of hydrogen that out-diffused from the C 3 H 5 -implanted region at 700°C is sufficient compared with D it at the SiO 2 /Si interface. 24) Therefore, we consider that D it at the SiO 2 /Si interface obtained by C-V measurement was reduced owing to the termination of the Si dangling bonds (·Si≡Si 3 ) at the SiO 2 /silicon interface with hydrogen outdiffused from the C 3 H 5 -implanted region. We then consider how the out-diffused hydrogen terminated the Si dangling bonds. Previous studies have shown that two types of hydrogen, molecular and atomic, dissociate from the C 3 H 5 -implanted region. In this study, a SiO 2 layer was formed at 900°C for 100 min and then irradiated with an electron beam for annealing at 500°C and 700°C for 30 min. A previous study showed that molecular hydrogen first dissociates from the C 3 H 5 -implanted region. 24) In addition, it has been demonstrated that a considerable amount of atomic hydrogen diffuses during annealing at a temperature higher than 800°C. For this reason, molecular hydrogen diffusion predominantly occurs during the subsequent annealing at 500°C and 700°C.
The mechanism of the hydrogen termination effect in the hydrocarbon-molecular-ion-implanted wafer was analyzed on the basis of the assumed termination reaction at the SiO 2 /Si interface. The dependence of the D it of C 3 H 5 -implanted wafers on annealing time and temperature was evaluated by quasi-static C-V measurement. Figure 6 shows D it at the midgap of the C 3 H 5 -implanted wafers after 500°C, 600°C, 700°C, 800°C and 900°C annealing. The D it of C 3 H 5 -implanted wafers decreased with increasing annealing temperature and time. In the wafers with high-temperature annealing, D it rapidly decreased significantly. A previous study showed the effect of hydrogen passivation by FGA below 500°C. 32) Previous studies have never shown a decrease in D it with annealing time at temperatures above 500°C. Understanding the reduction mechanism of the D it of C 3 H 5 -implanted wafers is extremely important for the application of advanced CMOS image sensors to device processing.
Therefore, we analyzed a termination reaction by reaction kinetic analysis in order to explore the reduction mechanism of the D it of C 3 H 5 -implanted wafers. The reaction equation for the time change in D it obtained from the assumed reaction model was derived and fitted with the experimental results for the annealing time dependence of D it . Furthermore, the activation energy of hydrogen termination at the SiO 2 /Si interface was derived by obtaining the reaction rate constant of the assumed reaction model from the fitting results. A reaction model of the hydrogen termination effect at the SiO 2 /Si interface was assumed. The reaction between hydrogen and a silicon dangling bond at the SiO 2 /Si interface is expressed by reaction Eq. (1). This is a reaction model in which hydrogen binds to silicon dangling bonds (Si·) to form Si-H. The reaction equation obtained from reaction Eq. (1) is shown in Eq. (2) Si H .
This reaction model is derived from the experimental results of hydrogen termination after low-temperature FGA. In addition, Eq. (5) is derived from the above Eqs. (2) and (4) to consider the hydrogen passivation mechanism at the interface during high-temperature annealing k B is Boltzmann constant and T is temperature. Using k 1 in Eq. (5) as a parameter, we derived k 1 by fitting the model with the experimental results of the annealing time dependence of the D it of the C 3 H 5 -implanted wafers. Figure 7 shows the fitting plots of D it at the midgap of wafers with C 3 H 5 implantation by Eq. (5). In addition, Fig. 8  We discuss the obtained activation energy of 1.67 eV. Stesmans reported that the activation energy in the reaction where hydrogen molecules terminate the Pb 0 centers at the SiO 2 /Si interface formed by dry oxidation on Si (100) is 1.51 eV. 34) The following is a model of the reaction of molecular hydrogen at the SiO 2 /Si interface: A previous study demonstrated that the molecular hydrogen out-diffused from the C 3 H 5 -implanted region at annealing temperatures below 900°C. 24) In addition, we clarified that the change in D it at the SiO 2 /Si interface was enhanced by electron beam irradiation and proportional to that in the density of Pb 0 centers obtained by ESR measurement. Therefore, we found that hydrogen molecules out-diffusing from the C 3 H 5 -implanted region can reduce D it at the SiO 2 /Si interface. The diffusion rate of hydrogen molecules in silicon wafers is high, and the molecules can easily reach the SiO 2 /Si interface. As shown previously, hydrogen easily dissociates from the SiO 2 /Si interface during annealing at temperatures higher than 500°C. The analysis of hydrogen termination was limited to low-temperature annealing below 500°C. However, the activation energy of 1.67 eV for hydrogen termination in high-temperature annealing above 500°C has been derived assuming two reaction models for the SiO 2 /Si interface and hydrogen. In this study, the hydrogen termination effect of the C 3 H 5 -implanted silicon epitaxial wafer, which has not yet been reported, reduced D it at the SiO 2 /Si interface. It was found that the activation energy of the hydrogen termination reaction for the silicon dangling bonds at the SiO 2 /Si interface even in high-temperature annealing is close to those in previous studies. The hydrogen termination effect in terms of D it reduction under annealing conditions from 500°C to 900°C is compatible with device processes, which has recently been progressing to lower temperatures.
On the other hand, the C 3 H 5 -implanted region also contains carbon atoms, which form carbon complexes that act as the gettering and trapping sites for metal impurities and hydrogen, respectively. [19][20][21][22][23] If carbon diffuses from the carbon complexes to the device active region, carbon can form the deep energy levels that cause the dark currents in CMOS image sensors. 35) However, since carbon diffuses more slowly in silicon than in hydrogen, it is possible to reduce the effect of carbon depending on the annealing conditions. Therefore, we believe that the results of this study are meaningful for device processes that tend to use lower temperatures.
The results suggest that hydrogen molecules out-diffusing from the projection range of hydrocarbon molecular ion can efficiently terminate D it at the SiO 2 /Si interface. This is an important characteristic that contributes to the reduction in D it at the SiO 2 /Si interface, which is required for highperformance CMOS image sensors.

Conclusions
We demonstrated that C 3 H 5 -implanted epitaxial silicon wafers can reduce D it and Pb center density determined by CV and ESR measurements, respectively. The D it and Pb center density of wafers without C 3 H 5 implantation increased after annealing at 700°C. On the other hand, the D it and Pb center density of C 3 H 5 -implanted wafers decreased after annealing at 700°C. These results indicate that C 3 H 5 -implanted silicon epitaxial wafers have the passivation effect through hydrogen termination, which decreases D it at the SiO 2 /Si interface. In addition, we derived the activation energy of 1.67 eV for the hydrogen termination reaction with hydrogen molecules and Si dangling bonds at the SiO 2 /Si interface. We clarified that hydrogen molecules out-diffusing  from the hydrocarbon-molecular-ion-implanted region can reduce D it at the SiO 2 /Si interface. We conclude that the hydrogen termination effect of the hydrocarbon-molecularion-implanted epitaxial silicon wafer can contribute to high electrical device performance, such as low noise or dark current owing to the interface state at the SiO 2 /Si interface of CMOS image sensors.