Minority Carrier Recombination Properties of Crystalline Defect on Silicon Surface Induced by Plasma Enhanced Chemical Vapor Deposition

This research investigates the carrier recombination properties of a crystalline defect layer introduced by the plasma enhanced chemical vapor deposition (PECVD) process of amorphous hydrogenated silicon nitride (SiN x ) passivation ﬁlms. A direct PECVD technique was used for SiN x ﬁlms deposition. A crystalline defect layer existed on the surface of the silicon substrate and is under the SiN x passivation ﬁlm. The recombination lifetime in this defect layer was obtained by focusing on the thickness of the defect layer and the effective lifetime before and after the defect layer etching. After etching a few nanometer thickness, effective lifetime drastically increased. On the other hands, the carrier recombination center could be electrically inactivated by 600 ◦ C annealing after SiN x deposition. According to the depth proﬁle of effective lifetime, it was clariﬁed the high carrier recombination region were concentrated near the surface of silicon substrate. a photo-conductance lifetime transient The maximum recombination S the effective lifetime τ bulk silicon inﬁnity. S

The future of the crystalline silicon solar cell industry depends on being able to reduce the production cost and increase the conversion efficiency. In order to realize the high conversion efficiency, the carrier lifetime needs to be increased by reducing the number of crystalline defects and/or inactivating the recombination centers in the bulk and on the surface of the silicon substrates. Crystalline defects, such as grain boundaries, dislocations, and impurities, act as recombination centers. Therefore, a lot of studies have been carried out to clarify the crystalline defect properties and to investigate the suppression techniques. [1][2][3][4] Surface passivation techniques are used on the surfaces of substrates to inactivate the surface recombination centers or dangling bonds. Many materials have been investigated for use in high quality passivation films. [5][6][7][8][9] Passivation can be achieved by two mechanisms, one is chemical passivation and the other is field-effect passivation. [10][11][12][13] Amorphous hydrogenated silicon nitride (SiN x ) passivation films are mainly used as surface coating for crystalline silicon solar cells because of their excellent surface passivation properties as well as their anti-reflection effect. SiN x films also act as a source of hydrogen for surface and bulk defects passivation. An annealing process can be used to diffuse the hydrogen contained in these films. [14][15][16] Plasma-enhanced chemical vapor deposition (PECVD) technique is used to form most of the SiN x passivation films because it allows for the fabrication of SiN x films at a low temperature. [16][17][18] However, the plasma process causes surface damage as crystalline defects due to the plasma and/or ion bombardment. [19][20][21][22] A case of crystalline defects introduced by PECVD is shown in Fig. 1. The crystalline defects exist at the surface of the silicon wafer, especially at the top of the texture structure.
There are some reports about the relationship between the passivation properties and surface damage. 23,24 However, in these reports, the main topics are the passivation properties and not the carrier recombination at the defect layer in the silicon substrate. It is necessary to separate components, such as lifetime degradation by the defect layer and carrier recombination suppression by the passivation.
In this paper, the carrier recombination activity of crystalline defects introduced by PECVD in SiN x passivation films was investigated. The thickness of the crystalline defect layer was observed by transmission electron microscopy (TEM). The depth profile of carrier recombination properties were studied using mirror polished silicon z E-mail: t-tachibana@aist.go.jp wafers. SiN x passivation properties were determined under several annealing condition.

Thickness and carrier recombination properties.-4 inch
Czochralski silicon (phosphorus doped n-type, 7-13 cm, double side mirror polished) substrates were used for this study. The wafers were 500 μm thick. A 10:1 diluted hydrogen fluoride (DHF) solution was used to remove the native silicon oxide layer prior to the SiN x passivation films deposition. Then, a direct PECVD technique using monosilane (SiH 4 ) and ammonia (NH 3 ) gasses (SiH 4 :NH 3 = 40:120 sccm) was used to form the SiN x passivation film. The deposition temperature was 450 • C, the pressure was 67 Pa, and the radio frequency power was 150 W. SiN x films were deposited on both sides in order to determine the recombination properties. The deposition thickness and refractive index of the SiN x films are approximately 80 nm and 2.1, which were obtained by ellipsometry measurement. After the SiN x film deposition, some samples were annealed by rapid thermal annealing (RTA) from 400 to 600 • C in a N 2 ambient for 30 sec. The effective minority carrier lifetimes were measured using a photo-conductance lifetime tester (Sinton Instruments, WCT-120) operated in transient mode at room temperature. The maximum surface recombination velocity (S max ) was estimated from the effective lifetime (τ eff ) by assuming that the bulk lifetime of the silicon substrate is infinity. S max was calculated from the following equation, where W is the wafer thickness. The thickness of the crystalline defect layer was determined by TEM (JEOL JEM-2100) at 200 kV. Electron beam deposition was used to deposit thin carbon films on the SiN x passivation films before the TEM sample preparation in order to protect the SiN x films from ion beam damage on the SiN x films. A focused ion beam (FIB) system was used to prepare the TEM samples, which were approximately 100 nm thick.
For the evaluation of carrier recombination properties in the crystalline defect layer, the SiN x passivation films were removed by DHF. Then, Iodine/Ethanol solution was used to passivate the substrates prior to the effective lifetime measurement. After lifetime measurement, those substrates were cleaned by DHF. Then, the crystalline defect layers were etched with a NHO 3 , HF, and CH 3 COOH mixture at room temperature (NHO 3 :HF:CH 3 COOH = 40:1:15). After etching away a few nanometers, the effective lifetimes were measured again. The etching and effective lifetime measurement were done again and again. The etched thicknesses were determined based on the weight difference before and after etching.

Recombination lifetime calculation.-
The recombination lifetime in the crystalline defect layer was estimated by considering the following. The recombination rate (U total ) is expressed as 25 where W is the thickness of silicon substrate, S f ront and S rear are the surface recombination velocity (S) of the front and rear sides, n Ave. is the average excess carrier density, and τ bulk is the bulk recombination lifetime. Assuming that n(x) = cont. and S f ront = S rear = S, Equation 2 is rewritten as where d is the thickness of the crystalline defect layer and τ de f ects layer is the recombination lifetime in the crystalline defect layer. Additionally, the thickness of the defect layer is thinner than that of the silicon substrate: d W . Thus, the defect layer could be neglected: W − 2d ≈ W , and thus, Equation 3 is expressed as Using Equation 4, τ ef f before the etching (τ ef f −1 ) and after the etching (τ ef f −2 ) of the crystalline defect layer can be described as Using Equations 5 and 6, the τ de f ects layer is assumed to be [7]

Results and Discussion
The measured effective lifetime as a function of annealing temperature is shown in Fig. 2. In the cell fabrication process, there is post annealing effect during the firing process for printed electrode. For the confirming this effects, the SiN x deposited substrates were annealed. In the as-deposition samples, the effective lifetime was approximately 1000 μsec. After 400 and 500 • C annealing, the effective lifetime slightly decreased or remained constant compared to the as-deposition samples. The dispersion is smaller than that in as-deposition samples. In the 600 • C annealed samples, the effective lifetime increased to approximately 5000 μsec. The corresponding S max values are approximately 25 cm/s (as-deposition substrates) and 5 cm/s (600 • C annealed substrates). We investigated the recombination properties in the crystalline defect layer underneath the SiN x layer in the as-deposition substrates and 600 • C annealed substrates.
Thickness and carrier recombination properties.-The cross sectional TEM images at the interface between the SiN x film and silicon in the as-deposition substrates and 600 • C annealed substrates are shown in Fig. 3. The black layer corresponds to the crystalline defect layer near the surface of the silicon substrates underneath the SiN x films. The contrast of defect layer looks uniform at the surface of the silicon substrates. The thickness of the crystalline defect layer is approximately 40-50 nm in both substrates. There is no large difference between as-deposition and annealed substrates. These defect layer could also affect the carrier recombination properties. On the other hand, it is unclear whether the reason of effective lifetime increased in 600 • C annealed substrates is improving the interface state between SiN x and silicon substrates or decreasing these defect layer effect. However, there are no large structural variation between the as-deposition substrate and 600 • C annealed substrates.
For the separate the effect of carrier recombination at the defect layer from the carrier recombination at the interface between SiN x passivation and silicon substrates, SiN x passivation films were removed and effective lifetime were determined with Iodine/Ethanol solution. The effective lifetimes in dependence of excess carrier concentration before and after damaged layer etching were shown in Fig.  4. The circle and square symbols show as-deposition substrates and 600 • C annealed substrates, respectively. Open symbols correspond to before etching, and filled ones to after etching. The effective lifetimes were approximately 100 μsec after the removal of SiN x films in as-deposition substrates. This value increased by etching a few nanometer thickness of damaged layer, the effective lifetime drastically increased. After the damaged layer was completely removed, the effective lifetime were saturated at approximately 1000 μsec, which corresponds to the bulk lifetime without the damaged layer. This result suggests that the damaged layer act as recombination centers of minority carriers, resulting that the effective lifetime was decreased. On the other hand, in the 600 • C annealed substrates without SiN x film, the effective lifetimes were approximately 1400 μsec. This lifetime is almost same as bulk lifetime, although there is damaged layer. Because of 600 • C annealing, the defects in the damaged layer were electrically inactivated. The origin of the difference between the effective lifetimes in as-deposition and the 600 • C annealed substrates might be hydrogen (H atoms) diffused from SiN x film due to the annealing. Which terminate the defects. From these results, the damaged layer is one of the origin for the difference of effective lifetime in the as-deposition substrates and 600 • C annealed substrates.
For the investigation of the carrier recombination distribution in the damaged layer, the depth profile of recombination properties were evaluated. The recombination at the crystalline defect layer was an- Figure 4. The carrier recombination properties in as-deposition and 600 • C annealed substrates. Effective lifetime in dependence of the excess carrier concentration before and after etching. Open symbols correspond to before etching, and filled ones to after etching. alyzed using the S max . Since the damaged layer thickness is thin as compared with the wafer thickness, this recombination could be assumed to be the surface recombination. The relationship between S max and the etched thickness of the crystalline defect layer is shown in Fig. 5. In the as-deposition substrates, S max was approximately 200 cm/s before etching. It dramatically decreased as increasing the etching layer thickness and becomes to less than half this value after a few nanometer etching. After over 100 nm etching, S max was saturated approximately 20 cm/sec. This S max corresponds to the recombination at the passivated surface and bulk. The damaged layer thickness determined by TEM image was around 50 nm. This is thinner than the thickness of the layer which act as recombination centers. On the other hand, S max obtained from the 600 • C annealed wafer was almost same as that without the damaged layer. There is no recombination center which determines the effective lifetime in this substrate.
Considering that S max drastically decreased after a few nanometer etching in as-deposition substrates, it can be assumed that high density recombination center exist at the near-surface region of the silicon substrates. These results indicate that not the complete crystalline defect layer obtained by TEM acts as carrier recombination center. The carrier recombination centers introduced by plasma damage are concentrated at the near-surface region of the silicon substrates. The origin of the defect layer might be connected to hydrogen and/or nitrogen atoms induced during the PECVD process. The dark contrast in this study is similar to the results found in damage formation promoted by nitrogen atoms during electron beam irradiation. 26,27 Nitrogen atoms were introduced into the silicon substrates during PECVD process and the crystalline defect might be formed by electron irradiation. On the other hand, hydrogen atoms could be introduced into the near-surface region of silicon substrates during the plasma process. 28,29 Since hydrogen atoms can form high-density carrier recombination centers at the silicon surface, they could explain the carrier recombination properties found in this study. Additionally, the presence of nitrogen could explain the dark TEM contrast.  Fig. 4. Thus, the recombination lifetime in the crystalline defect layer is approximately 20-40 nsec. In case of the crystalline defect layer thickness is thinner than that obtained with TEM, from few nanometer to 10 nm, τ de f ects layer decrease to 2-7 nsec.
As a possible explanation for the suppression of surface recombination after passivation, we considered that the defect layer was inactivated by hydrogen diffusion from the SiN x film. Therefore, for fabrication processes that do not include high temperature annealing, the introduction of this defect layer need to be suppressed since no hydrogen diffusion can occur during the fabrication. We reported one technique to suppress the introduction is fabricating the thin oxide layer before PECVD processes. 30 From the view point of system development, technique to do less plasma damage to silicon surface would be required.

Conclusions
Carrier recombination properties of the crystalline defect layer introduced by PECVD were investigated using lifetime measurement. The thickness of the defect layer was approximately 40-50 nm as obtained by TEM observation. The carrier recombination properties were obtained using lifetime measurement and nanometer scale etching. Assuming the recombination at the crystalline defect layer is the surface recombination, S max was obtained. In as-deposition substrates, the defect layer acted as a carrier recombination center. After few nanometer defect etching, S max drastically decreased. Therefore, high density recombination centers can be assumed to exist at the nearsurface region of the silicon substrates. After the damaged layer was completely removed, the effective lifetime were saturated the value which corresponds to the bulk lifetime without the damaged layer.
In 600 • C annealed substrates, S max was almost unchanged before and after etching. The defects in the damaged layer were electrically inactivated due to the 600 • C annealing. The origin of the crystalline defect might be connected to hydrogen and/or nitrogen atoms induced during the PECVD process. One of the reason for increasing the effective lifetime in 600 • C annealed substrates is that the carrier recombination centers at the damaged layer are electrically inactivated.