A review of plasma-induced defects: detection, kinetics and advanced management

Plasma-induced defects are often recognized in state-of-the-art semiconductors, high-efficiency solar cells and high-sensitivity image sensors. These defects are in the form of a dangling bond, bond deformation, or impurity/residual, which impacts on the device performance and reliability. The defects are introduced via plasma-material interactions during manufacturing processes such as deposition, etching and implantation. So, the management of defects throughout the manufacturing is important for high-performance device fabrication. In this review, we overview the generation and recovery of plasma-induced defects in order to develop the defect-managed advanced plasma processing for further improving the device performances. The defect generation and recovery are described, based on the recent results of in-situ and real-time detection of plasma-induced defects. Two examples are presented: the growth of hydrogenated amorphous silicon and the surface passivation of crystalline silicon for high-efficiency solar cell applications.


Introduction
The defects in semiconductors impact on the device performances and reliability [1,2]. For example, in field-effect transistors, shown in figure 1(a), defects located at the channel-gate interface induce the deterioration of the carrier transport and the current leakage [3][4][5][6]. For another example, in solar cells Original Content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. based on silicon heterojunction (SHJ), shown in figure 1(b), defects formed at the heterointerface result in a decrease in the conversion efficiency via reductions in both the photocurrent and the output voltage (see figure 1(c)) [7][8][9][10][11]. Therefore, the suppression of these defects is important for improving the device performances. The suppression of defects becomes further important for the next-generation three-dimensional devices such as stacked nanosheet gate-all-around transistors [12,13], since they have many heterointerfaces, where defects are concentrated due to the heterointerface nature, e.g. lattice mismatch and/or stress accumulation.
The defects in silicon are well investigated [14][15][16][17] and highly controlled, because silicon is the most common and Reproduced with permission from [33]. The solar cell photovoltaic performance, e.g. photocurrent vs. output voltage, is shown in (c) for defect-rich and less-defect solar cells. The defects result in a reduction in the conversion efficiency of solar cells. Copyright (2021) The Japan Society of Applied Physics.  [33]. Vacancies and interstitials are intrinsic defects. Impurities such as metals, N, C, O atoms, and dopants of B, P and As atom are extrinsic. The plasma induced-defects are extrinsic, and those are in the form of dangling bond (DB), bond deformation, or residual/impurity originating from the processing gases. (b) Detection methods and detection range [33]. There are a variety of well-established methods. TEM/EDS, Raman, XPS, FTIR, XRF, SIMS, ESR, PL and CL are widely used to characterize various types of defects. Capacitance-voltage (CV) measurement is widely used to characterize the trap energy and density. DLTS and PCD are the most sensitive methods to detect the defects, as low as ⪅10 11 cm −3 . The pump-probe photocurrent (PP-PC) measurement is aimed for in-situ and real-time detection of plasma-induced defects during plasma processing. Copyright (2021) The Japan Society of Applied Physics.
important semiconductor for various electronic devices. In general, defects are classified either intrinsic or extrinsic. As shown in figure 2(a), vacancies [18,19] and interstitials [20,21] are intrinsic defects. Impurities such as metals [22,23], N, C, and O atoms [24][25][26] and dopants of B, P and As atoms [27,28] are extrinsic defects. The density of these defects varies in a very wide range from ⪅ 10 10 cm −3 for metals to ≈ 10 19 cm −3 for dopants. For silicon, metals are extremely reduced by well-established manufacturing processes, and thereby high-quality silicon is obtained, which is one of the advantages for silicon-based electronic devices in terms of the device performance and reliability, with respect to other materials.
In addition to the above-mentioned defects, there is another type of defects, so called plasma-induced defects [29][30][31][32][33]. The plasma-induced defects are often introduced unintentionally by plasma processing for thin-film deposition [34,35], nanoscale etching [36][37][38] and dopant implantation [39][40][41], which are commonly used for semiconductor manufacturing [42]. The plasma-induced defects are extrinsic, and they are in the form of a dangling bond [43][44][45], bond deformation [46], or impurity/residual [47,48] associated with processing/ source gases. The accumulation of plasma-induced defects often causes a serious damage in the material, which is observed as the disordering of lattice, i.e. amorphization [49][50][51][52][53]. Most of the plasma-induced defects are recovered or passivated by forming gas annealing [54] (e.g. annealing under a nitrogen containing hydrogen gas (a few-%) at 450 • C for 30 min). Nevertheless, some of the defects remain in the material and become a bottleneck of the device performance and reliability. So, it is necessary to suppress these residual defects, by understanding the details of defect kinetics during plasma processing as well as annealing.
In this review, we begin with plasma-material interactions in terms of the generation of plasma-induced defects near the material surface (section 2). The defect structures, energy levels, and carrier dynamics are briefly described (section 3). We then describe the detection methods of defects (section 4), and introduce in-situ and real-time detection of plasma-induced defects, based on the pumpprobe photocurrent measurement (section 5). Two examples of plasma-induced defects in actual plasma processing are presented. One is for the growth of hydrogenated amorphous silicon (a-Si:H) (section 6) and the other is for the surface passivation of crystalline silicon for high-efficiency SHJ solar cells (section 7). The defect generation and recovery are discussed and interpreted, according to the recent experimental results. The improvement of defect detection and future prospective are presented (section 8). Finally, we summarize this review article (section 9).

Plasma-material interactions
The plasma-induced defects are created as a result of plasmamaterial interactions [29][30][31][32]. An example is shown in figure 3, in which various elementary processes associated with the generation of plasma-induced defects are illustrated for silicon under a hydrogen (H 2 ) or argon (Ar) plasma [32]. As known, plasma is a collection of electrons, ions, radicals, and photons that are energetic and/or reactive species [55,56]. So, once these species arrive at the material surface, they can generate defects, interacting with the material atoms. Among them, energetic ions are known to efficiently generate defects. The ions are accelerated in the plasma sheath over the material surface, and collide with the material atoms at a high kinetic energy of a few tens to hundreds of eV [57,58] (shown in the upper left). By the collisions, the structure of material, i.e. lattice, near the surface is deformed. If the deformation is large, the bonds in lattice are broken, yielding dangling bonds (DBs) [59][60][61]. Associated with this bond breaking, point defects such as vacancies and interstitials can be created if the lattice atom moves from the original site (see the lower). When the deformation and point defects are accumulated in the material, a structural transition may occur from an ordered to disordered state, i.e. amorphization [49][50][51][52][53]. The amorphization is often observed after an intense plasma processing [62][63][64][65]. In such an intense plasma processing, some high-energy ions penetrate into the material, where they stay and become impurities/residues (see the middle left) [62][63][64].
As for photons, they can generate DB defects via electronic excitation of weak bonds in material [66][67][68] (see the upper middle). The generation of DB defects depends on the photon energy; the vacuum ultraviolet photons is known to efficiently generate DB defects. Nevertheless, such DB defects are mostly recovered by annealing. For radicals, i.e. reactive atoms or molecules having DBs, they easily react with weak bonds in material, which results in DB defects. There are many species of radicals in processing plasmas, according to processing/ source gases. Among them, hydrogen (H) [65,69] and fluorine (F) atoms [70] are known to be highly reactive and diffuse easily into the material (see the upper right), so they can generate defects not only on the material surface, but also inside the material in a depth of several nanometers or even deeper. Therefore, the control of these reactive species in addition to the energetic ions is a key for an advanced management of plasma-induced defects.  [33]. An example is shown for silicon exposed into a hydrogen (H 2 ) or argon (Ar) plasma. The bombardments of energetic ions to the material surface cause the bond deformation and breaking, leading to the generation of DB defects. Energetic ions penetrate into the material, where they stay and become impurities. Ultraviolet (UV) or vacuum UV (VUV) photons are absorbed in the material, where weak bonds are dissociated, generating DBs. Reactive radicals such as H atoms react with material atoms, which generates DB defects. Besides, the deposition of polymers and/or particulates is also considered to introduce defects in the form of void. These defects are schematically illustrated. Copyright (2021) The Japan Society of Applied Physics.

Defects and carrier dynamics
The defects in semiconductors influence the carrier dynamics in two ways. One is the mobility reduction and the other is recombination/generation [1,2]. The mobility reduction is widely known for amorphous semiconductors [71] (figure 4(a)), and the recombination/generation is induced by metals and DB defects [2,71] (figure 4(c)). The mobility reduction and recombination are intuitively explained, using the energy diagram, as shown in figures 4(b) and (d). When defects are formed in lattice, the corresponding localized states are created, which are often located in the electrical bandgap. The electrical bandgap, also called mobility gap, is defined by an energy gap between the conduction band (CB) minimum and the valence band (VB) maximum. It is slightly larger than the optical bandgap, which is often obtained from a Tauc plot, where (αhν) 1/2 = A(hν − E g ) for amorphous semiconductors [72][73][74]. Here, α is the absorption coefficient, hν is the photon energy, A is the proportional constant and E g is the optical bandgap. Hereafter, the bandgap refers the electrical bandgap except for particular description.
The localized states in the bandgap behave as traps for free carriers (electrons or holes) by capturing them from the CB or VB, depending on the n-or p-type. Once they are trapped, they can be emitted or not, depending on the energy level of the localized states. For example, if the localized states are located close to the CB or VB edge, defined as shallow levels, trapped carriers are emitted thermally in a short period. So, this type of shallow-level localized states, also called shallow-level defects, mainly affect the mobility of carriers ( figure 4(b)). On the other hand, if the localized states are located far from the CB and VB edges, defined as deep levels, trapped carriers are not able to be emitted thermally in a short period. Instead, the trapped carriers recombine with the other type of carriers if they meet. So, this type of deep-level localized states, also called deep-level defects, primarily limit the lifetime of carriers (figure 4(d)). Therefore, for optoelectronic devices such as solar cells and image sensors, the suppression of deep-level defects is particularly important. In fact, in the SHJ solar cells described in introduction, the deep-level defects are reduced down to ⪅10 10 cm −2 [11], which deserves an excellent photovoltaic performance of a conversion efficiency of ⪆25% [9][10][11].

Methods of defect detection
There are a variety of well-established methods to detect defects in materials [2]. Some of those methods are listed in figure 2(b), where typical detection ranges are denoted by color bars. Details are summarized in table 1, in which the spatial resolution, chemical sensitivity, mapping capability, detection limit and some limitations are described. In the following paragraphs, each method is briefly explained.
Transmission electron microscope (TEM) [75] is commonly used to observe the crystal structures, lattice dislocation, lattice mismatch, interface roughness [76] and amorphization [62][63][64][65]. Energy dispersive x-ray spectroscopy (EDS) [77] is often performed together with TEM for the elemental analysis or chemical characterization. Mapping over a TEM image gives useful information on spatial distribution of elements in addition to nanostructure. Raman spectroscopy [78] is used for structural analysis, characterizing the crystal structure, crystallinity, stress migration and amorphization. The structure of chemical bonds is also obtained. X-ray photoelectron spectroscopy (XPS) [79] is highly sensitive to the surface states, and thus the surface impurities are detected. The elements of impurities and their chemical states are determined. Fourier-transform infrared spectroscopy [80] is employed to determine the configuration of chemical bonds. X-ray fluorescence (XRF) [81] is a convenient technique for elemental and chemical analysis.
Secondary ion mass spectrometry (SIMS) [82] is widely used for characterizing the depth distributions of impurities and dopants [83,84]. Positron annihilation spectroscopy (PAS) [85] is useful for the detection of vacancy-type defects in solids and surfaces. The density of such defects and pore size are obtained. Electron spin resonance (ESR) [86] is commonly used to detect DBs as well as radicals. The density of DBs, related to deep-level defects, are determined. However, the charged states are not detected. The photoluminescence [87,88] and cathodoluminescence (CL) [89,90] are widely used for analyzing the energy levels of localized states near the surface. The deep-level transient spectroscopy (DLTS) [91] and photoconductivity decay (PCD) [92] are known to be the two most sensitive methods. With these methods, defects can be detected as low as ⪅10 11 cm −3 , as listed in table 1. The energy levels of defects are determined with DLTS, whereas the lifetime of minority carriers is obtained with PCD. Besides, capacitance-voltage (CV) measurement [2] is often used to characterize the defect/trap energy and density. For CV and DLTS, the contact formation and its isolation are required, which may limit the material and structure. In PCD, the lifetime is obtained from a decay of the photoconductivity after the illumination of a light pulse.
All the methods described above are powerful and reliable for characterizing defects in materials. Many useful and important information are obtained with them. Nevertheless, those ex-situ based methods are often not able to apply to a plasma processing environment in order to study the defect generation and recovery kinetics, because the measurement  112] is disturbed or interfered with a processing plasma, electrically, magnetically or instrumentally. For instance, TEM, EDS, XPS, SIMS, and CL are not feasible under a plasma processing environment, because those methods use an electron or ion beam as a probe or signal, and thus the beam is distorted or shielded by a processing plasma as a conductor, containing electrons and ions. ESR is also not feasible since it uses a strong magnetic field, and thus it significantly distorts a plasma. For DLTS and CV, the capacitance measurement is modified by charging and electric field induced by the plasma sheath. For other methods such as Raman, XRF and PAS, the measurement may interfere instrumentally with a plasma chamber/reactor. These limitations are also listed in table 1.
For in-situ detection of plasma-induced defects, there are two sophisticated methods developed so far. One is timeresolved microwave conductivity (TRMC) [93,94] and the other is cavity ringdown spectroscopy (CRDS) for subgap DB defect absorption [95,96]. With TRMC, the transient conductivity of an a-Si:H film during the growth is measured, and a defect-rich surface layer is recognized. With CRDS, the subgap DB defect absorption is measured during the growth of a-Si:H, and the density of the surface DB defects is obtained. These are two pioneer works, yielding important information of plasma-induced defects. Nevertheless, the equipments of those methods seem to be relatively large and complicated. So, a simple and compact in-situ detection method is required for studying more details of plasma-induced defects under various plasma processing environments.

In-situ detection of plasma-induced defects
In this section, we explain our unique and simple in-situ detection method for plasma-induced defects [97], based on the photocurrent measurement under a pump-probe operation [98].

Sample structure
The sample structure is shown in figure 5(a). The sample consists of a semiconductor thin film prepared on a substrate with a pair of contacts for photocurrent measurement [99]. The substrate is usually an insulating material such as glass or quartz in order to minimize the current leakage to the substrate. The contact materials are selected to have ohmic contact with the semiconductor [100]. For instance, an ITO/Ag/ITO stack is adapted for a-Si:H [101,102]. The gap between contacts, i.e. the channel length, is properly selected, taking into account of the conductivity of the semiconductor thin film. In the following section, we use an insulator as a substrate, however it is also possible to use a semiconductor such as crystalline silicon. In that case, the plasma-induced defects are detected in crystalline silicon. This is explained in section 7.

Pump-probe photocurrent measurement
The pump-probe photocurrent is measured to detect the plasma-induced defects. The sample is illuminated with the pump and probe light, while a DC voltage is applied between the contacts. The pump light is used to generate free carriers in a semiconductor film, whereas the probe light is used to emit the carriers trapped at defect-mediated localized states, i.e. trapped carriers, into CB (for electrons) or VB (for holes), as shown figure 5(b). For the pump, the photon energy is higher than the semiconductor bandgap. In contrast, the photon energy of the probe light is selected to be smaller than the semiconductor bandgap. For an example of a-Si:H, a visible laser is employed for the pump, while a near-infrared laser is employed for the probe [97]. The generated carriers travel towards one of the contacts, where they are collected, yielding currents ( figure 5(a)). There are two kinds of currents, associated with the pump and probe. Here, we define the former as a photocurrent and the later as a trap current. These two currents are required to distinguish in order to characterize the defects and trapped carriers. To do that, in the experiments, the pump and probe light intensities are modulated at different frequencies, and then the currents synchronized at these frequencies are detected using a lock-in technique. The modulation frequencies are usually in a range of several hundred Hz to several kHz, taking into account of the carrier transport across the channel, the frequency responses of light sources and a lock-in amplifier. This modulation is a key for the photo-and trap current measurements during plasma processing. Because, in plasma processing, the plasma-related currents such as the ion current [103,104], plasma emission [105,106]-induced current and leakage currents, are superimposed over the photo-and trap currents.

Defects and trapped carriers
The photocurrent and trap current are closely related to the defects and trapped carriers. The photocurrent, I p , is proportional to the carrier mobility, µ, and carrier lifetime, τ , which is expressed as [107] I p ∝ µτ. (1) Here, the carrier lifetime is limited by various recombination processes [108]. For silicon, the recombination is dominated by deep-level defects in a case of a low-injection level of carriers [109]. The recombination through deep-level defects is described by the Shockley-Read-Hall (SRH) model [110,111] and thus the lifetime is inversely proportional to the density of deep-level defects, n d .
Substituting equation (2) into equation (1), the photocurrent is shown to be inversely proportional to the density of deep-level defects, i.e.
So, a change in I p reflects either an increase or decrease of the deep-level defects. Specifically, a decrease in I p reflects an increase of defects, i.e. the defect generation. On the other hand, an increase in I p reflects a decrease of defects, i.e. the defect annihiration/recovery. As for the trap current, the density of trap carriers, n d , can be determined by the following equation (details in a [112]).
where Γ pump and Γ probe are the photon fluxes of pump and probe lights, σ p and σ t are the optical absorption cross-sections of the carriers for the valence and trapped carriers, and n v is the valence electron density. The optical absorption cross-section depends on the nature of defects, i.e. electronic structures and charging. The cross-section is given by an optical matrix [113], and it is not simply expressed. So, we assume it to be constant for simplicity, and set at σ p /σ t = 1. Then, the density of trapped carriers (electrons) is simply obtained from the ratio of I t /I p and the known values of Γ pump and Γ probe .

Pump-probe operation
The pump-probe operation is demonstrated for a solar-grade thin film of a-Si:H, as shown in figure 6 [114,115]. In demonstration, a 220 nm-thick a-Si:H film with a bandgap of 1.7 eV is prepared on a glass substrate by means of plasma-enhanced chemical vapor deposition [116][117][118]. The a-Si:H film is then illuminated with the pump light of a visible laser (532 nm in wavelength, a photon energy of 2.33 eV) and the probe light of a near-infrared laser (1342 nm in wavelength, a photon energy of 0.92 eV).   proportional to the pump light intensity, whereas the trap current shows a tendency to be gradually saturated. The density of trapped carriers is calculated with these two currents, using equation (4), and shown in figure 6(b). It shows a tendency of saturation similar to the trap current. This means that the occupation of localized states, i.e. traps, increases with increasing the pump light intensity, and gradually saturated. Figure 6(c) shows the dependence of the trap current on the probe light intensity, P probe . The trap current responds linearly to the probe light intensity. It indicates that the occupation is not influenced by the emission of trapped carriers; the emission of the trapped carriers is relatively small compared with the carriers trapped at the localized states.

Trap occupation and origin
Here, the origin of the trapped carriers is briefly described. Figure 6(d) shows a schematic diagram of the density of states (DOS) of a-Si:H [71,119,120]. There are many localized states in the bandgap, consisting of shallow levels associated with the structural disorder and deep levels originated from DBs. The density of shallow levels is known to be at least 10 18 cm −3 [71,121,122], while the density of deep levels is usually 10 15−16 cm −3 [122,123]. Comparing to the density of trapped carriers of 10 17−18 cm −3 , given by equation (4), the origins of trapped carriers is considered to be mainly the shallow levels, not the deep levels. Because the density of the deep levels is much smaller than that of the observed trapped carriers.

In-situ detection of defects in a-Si:H growth
To demonstrate in-situ detection of plasma-induced defects, the above-mentioned method is applied to an a-Si:H growth process of solar cells [97]. The demonstration shows the depth distribution of defects in a growing film of a-Si:H. The defect generation and recovery are discussed, based on the results.

Experimental part
A capacitively coupled high-frequency discharge system is used for the growth of an a-Si:H film [124][125][126]. The source gas is a gas mixture of H 2 and silane (SiH 4 ). The growth temperature is set at 200 • C. The other conditions for discharge are following: a parallel-plate electrode gap of 22 mm, the discharge frequency of 60 MHz, a gas pressure of 0.3 Torr, gas flow rates of 53 sccm for H 2 and 7 sccm for SiH 4 , respectively. The details of experimental apparatus and conditions are described in a [97].

Optical properties of a-Si:H
The experimental results on the optical and electrical properties of an a-Si:H film during growth are shown in figure 7. The time evolutions of optical properties are shown in figures 7(a) and (b), which are obtained from real-time spectroscopic ellipsometry [127]. Here, a Tauc-Lorentz model [128] is assumed for an a-Si:H film to determine the film thickness, d, the optical bandgap, E g , the refractive index, n, and the extinction coefficient, k. As expected, the thickness is proportional to the discharge time, t, (i.e. the growth time) and the growth rate is constant. For the optical bandgap, refractive index, and extinction coefficient, those optical parameters are nearly constant throughout the discharge, i.e. regardless of the film thickness. It indicates that the growing a-Si:H is optically homogeneous from the bottom to the top of the film.

Electrical properties of a-Si:H
On the other hand, the electrical properties of a-Si:H are highly dependent on the growth time, i.e. the film thickness. The time evolutions of the photocurrent and trap current are shown in figure 7(c). The photocurrent is not observed in the initial growth period for a thin film (t ⪅ 120 s, d ⪅ 20 nm). The photocurrent is observed once the film thickness exceeds an critical value of d ≈ 20 nm. Then, it rapidly increases with the discharge time, i.e. the film thickness (t ⪆ 120 s, d ⪆ 20 nm). For the trap current, it shows a tendency similar to that of the photocurrent. Figure 7(d) shows the photoconductivity, σ ph , obtained from the photocurrent divided by the thickness. The photoconductivity increases rapidly in the initial growth period, and then approaches to a certain value in a later period for a thick film. The density of trapped carriers, obtained from the photocurrent and trap current, using equation (4), is also shown in figure 7(d). It indicates that the density of trap carriers is roughly constant during growth, except for the initial growth period, in which the trap current is below the detection limit.

Annealing effects
The optical and electrical properties are also measured after the growth, i.e. during annealing, as shown in figure 7, for t ≧ 1280 s. After the growth, the optical properties are confirmed to be unchanged. In contrast, the electrical properties are drastically changed after the growth. As shown in figure 7(c), the photocurrent is rapidly increased immediately after the growth, i.e. the termination of discharge. Then, it is slowly increased by annealing. The density of trapped carriers is decreased immediately after the discharge, and then it is slowly decreased by annealing.

Depth distribution and kinetics of defects
The above-mentioned time evolution of photocurrent suggests the depth distribution of plasma-induced defects, as in figure 8. As shown in figure 7(c), the photocurrent is initially not observed for a thin film growth (t ⪅ 120 s, d ⪅ 20 nm), which indicates the presence of a large number of deep-level defects in such a thin film. As explained in the introductory part (section 2), the plasma-induced defects are distributed near the surface, in a range of the penetration depth of those species of ions, photons and radicals. The photons and radicals such as H atoms are known to penetrate in a depth of several to tens of nm, so the defects are distributed in such a thin film from the surface to the bottom, as shown in figure 8(a). Because of that, photocarriers generated by the pump light recombine efficiently through those deep-level defects (e.g. associated with DBs) before being collected at the contacts. As a result, the photocurrent is not observed for such a thin film.
Once the thickness exceeds a critical value of d ≈ 20 nm, the photocurrent is observed, and then it increases with the thickness, as shown in figure 7(c). These results suggest the following two points: the presence of a defect-rich surface layer and the growth of a bulk layer, as shown in figure 8(b). The defect-rich surface layer is a surface layer, in which the plasma-induced defects are concentrated, and its thickness is d ≈ 20 nm for the present case. The bulk layer is a high-quality, low-density-of-defects layer, which is formed underneath the defect-rich surface layer, and it becomes thicker as the film grows (t ⪆ 120 s, d ⪆ 20 nm). In this bulk layer, the photocarriers travel towards the contacts, yielding a photocurrent. The photocurrent increases as the bulk layer becomes thick, i.e. the growth of the film. As for the trapped carriers, shown in figure 7(d), the density is roughly constant during the growth. It suggests that the shallow-level defects, associated with the bond deformation in amorphous network, are uniformly distributed in the bulk layer.
After the growth, the photocurrent is significantly increased, as shown in figure 7(c), for t ≧ 1280 s, which indicates the annihiration/recovery of the plasma-induced defects at deep levels, as shown in figure 8(c). The annihiration/recovery time of those deep-level defects depends on the annealing temperature, and it ranges from a few minutes to several hours. From the experiments performed at different annealing temperatures, the activation energy is determined to be ≈ 0.53±0.06 eV, which corresponds to the diffusion barrier (typically 0.5 eV) [129,130] of mobile hydrogen (H) atoms [131,132]. This indicates that the defect annihilation/recovery is associated with the diffusion of mobile H. Specifically, the deep-level defects such as DBs are considered to be terminated by mobile H during annealing [69,133,134]. As for the trapped carriers, their density is slightly reduced during annealing, as in figure 7(d), for t ≧ 1280 s. It implies the recovery of the shallow-level defects, i.e. the improvement of the structural disorder of amorphous network by annealing.

In-situ detection of defects in crystalline silicon
In this section, we present in-situ detection of plasma-induced defects in crystalline silicon under a silicon surface passivation process [135,136]. To do that, crystalline silicon is used as a substrate, instead of an insulator such as glass used in the previous section. The defect kinetics, associated with H atoms, is also discussed, based on the observed results.

Experimental part
In the experiment, silicon on insulator (SOI) [137] is used as a substrate. There are two beneficial points to adapt a SOI substrate, with respect to a neat silicon wafer. One is the suppression of current leakage, and the other is the improvement of the sensitivity of defect detection, owing to the thin active layer of SOI.
Two experiments are performed to demonstrate the generation of defects in silicon and the surface passivation of silicon. For the former, a H 2 plasma treatment [138] is simply performed on SOI. For the later, the growth experiment of a low-density-of-defects a-Si:H is performed on SOI [135,136]. The surface passivation obtained by such high-quality a-Si:H is crucial for high-efficiency SHJ solar cells [9][10][11], described in section 1 and shown in figure 1(b).
The experimental procedure is following. An SOI is cleaned with a diluted HF solution to remove the surface native oxide [139]. The silicon surface is thus terminated with H atoms [140,141]. The SOI is then placed in a capacitively coupled high-frequency discharge apparatus for a H 2 plasma treatment or an a-Si:H growth. The processing temperature is fixed at 147 • C for both experiments. Real-time spectroscopic ellipsometry is performed in addition to the photocurrent measurement. From spectroscopic ellipsometry [127], the etch amount of silicon is determined for a H 2 plasma treatment. For the a-Si:H growth, the film thickness is obtained [128]. The details of experimental conditions are listed in table 2 [135]. Figure 9(a) shows the experimental results on the time evolution of the photocurrent in SOI during H 2 plasma treatments [135]. The H 2 plasma treatments are performed six times, where the period of each treatment is varied from 1 ms to 100 s, denoted by arrows. The interval between the treatments is fixed at 5 min, in which the SOI is annealed at 147 • C in a H 2 gas, which is the same temperature of the plasma treatments.

Defect-rich surface layer and amorphization by H 2 plasma
As shown, the photocurrent clearly decreases with a H 2 plasma treatment [65], indicating the generation of deep-level defects. A decrease in photocurrent is enhanced for the treatment period; the photocurrent is decreased by one order of magnitude for a treatment period of 100 s, compared with that of initial, i.e. without treatments. This decrease in photocurrent reflects that the density of defects is increased by one order of magnitude, according to the relation in equation (3). The density of defects is examined with a neat silicon wafer by the lifetime measurement of minority carriers using a quasi-steadystate photoconductance technique [142,143], which indicates the density of the deep-level defects is increased to the order of 10 12 cm −2 after a H 2 plasma treatment of 100 s. Figure 10(b) shows a cross-sectional TEM image of the silicon surface, treated with a H 2 plasma for 100 s [135]. The structure near the silicon surface is disordered, and thus the amorphization is observed. This disordered layer or amorphized layer is recognized as a defect-rich surface layer, in which a large number of defects are concentrated. The layer is formed when the treatment time exceeds ≈ 10 s for the present conditions. The thickness of this layer monotonically increases with the treatment time. For instance, the thickness is increased to be 1.3 nm for an treatment time of 100 s. The amorphization phenomena are explained by the penetration of H atoms and their reactions with silicon lattice bonds. As described in section 2, H atoms supplied from a H 2 plasma [144] easily react with the silicon bonds in lattice, causing the bond deformation and breaking, generating to DBs. When those deformation and DBs are accumulated, the lattice structure is broken, inducing a transition from ordered to disordered, i.e. amorphization [49][50][51][52][53]. After such amorphization takes place, the recovery is no longer possible by low-temperature annealing, and thus the photocurrent does not return to the initial level, observed in the present experiment, as shown in figure 9(a) for t ⪆ 32 min.   Figure 9(b) shows the time evolution of the photocurrent in SOI during the growth of a-Si:H [135]. The growth is performed six times, growing ≈3 nm-thick each by a discharge of 10 s, denoted by arrows. The thickness of the a-Si:H film is indicated nearby. The film is annealed at a temperature of 147 • C, i.e. the same temperature of growth, for 5 min until the next growth, i.e. discharge. As shown, the photocurrent is initially decreased by the growth of an ultrathin-layer (d ≈ 3 nm) of a-Si:H, formed by the first discharge. However, it is increased with the growth of a relatively thick layer (d ⪆ 6 nm) of a-Si:H, formed by the second and following discharges. This increase in the photocurrent means that the deep-level defects on the silicon surface are decreased, and the recombinations of carriers in silicon are suppressed. The silicon surface is thus passivated with the growth of an a-Si:H layer (d ⪆ 6 nm). The increase in the photocurrent is enhanced for a thick layer of a-Si:H, which indicates that the surface passivation is improved by a thick layer (e.g. d ≈ 18 nm). Figure 10(a) shows the interface structure of an a-Si:H/SOI stack, prepared by the above-mentioned processes (see also table 2). As shown, an a-Si:H film is uniformly grown on the silicon surface, and the a-Si:H/SOI interface is sharp. The disordered layer or amorphization is not observed. Such sharp interface formation without amorphization is crucial particularly for high-efficiency SHJ solar cells with an excellent surface passivation [101,102].

Defect generation and recovery kinetics
The time evolution of the photocurrent observed in the experiments indicates the following defect generation and recovery kinetics, according to the relation in equation (3). Firstly, a reduction in the photocurrent during the discharge for a H 2 treatment, as shown in figure 9(a), denoted by (i)-(v), indicates the generation of plasma-induced defects in silicon near the surface ( figure 11(a)). Secondly, an increase in the photocurrent after the discharge, i.e. during annealing (see between the arrows), indicates the recovery of these plasma-induced defects.
Such a defect generation and recovery is more clearly recognized in a case of growth of a-Si:H on a glass substrate, which is shown in figure 9(c). The growth of a-Si:H is performed six times, growing ≈3 nm-thick each by a 10 s discharge, denoted by arrows. The photocurrent is reduced during discharge for the growth, however, it is increased after the discharge, i.e. annealing. Thus, the defects are generated by plasma treatment and recovered by annealing. By looking data carefully, the decrease in the photocurrent during discharge is smaller for a thicker film (see (vi) in figure 9(c)), compared with that of a thin film (see (ii)). It supports that the defects are concentrated near the film surface, as described in section 6.5. Besides, the amplitude of the photocurrent for a ultrathin film (d ≈ 3 nm) is recognized to be rather small, with respect to a relatively thick film (d ⪆ 6 nm), by one to two order of magnitude. This suggests that defects are concentrated more in an ultrathin film (d ≈ 3 nm) [115].
Based on the above experimental results, the details of defect kinetics in surface passivation are explained as follows. As shown in figure 9(b), the photocurrent is initially reduced by the first discharge for an ultrathin-layer growth, and it stays low even in annealing, denoted by (i), which indicates the presence of a large number of defects in an ultrathin a-Si:H/SOI stack. There are two reasons for that. One is the defect generation in silicon near the surface via H-atom in-diffusion and reactions. The H atoms generated by the decomposition of a source gas molecule of SiH 4 in plasma penetrate into the a-Si:H layer and arrive at the silicon surface if the a-Si:H layer is extremely thin (d ⪅ 3 nm). The H atoms may also penetrate into silicon and react with the lattice bonds, generating defects, as demonstrated clearly by a H 2 plasma treatment described in section 7.2. This effect is prominent for the initial growth under a high-H 2 dilution condition [135,136]. The other reason for a reduction in the photocurrent for an ultrathin-layer growth is a high concentration of defects in such an ultrathin film, as described in the previous paragraph, possibly due to the interface stress (this point is described more in the following paragraph). Because of these two reasons, the defects are distributed near the a-Si:H/SOI interface, and thereby the photocurrent is reduced for an ultrathin a-Si:H/SOI stack.
For the growth of a relatively thick layer of a-Si:H (d ⪆ 6 nm), the penetration of reactive H atoms as well as energetic ions into silicon are reduced; a thick layer behaves as a barrier layer for those species, as in figure 11(b). Because of that, a further generation of defects in silicon is suppressed. Instead, the defects are generated near the a-Si:H surface by those species. Nevertheless, the defects generated in a relatively thick layer of a-Si:H are recovered by annealing, indicated by the photocurrent increase (see in figure 9(c)). So, the surface passivation is obtained with a relatively thick layer of a-Si:H.
Another interesting feature related to the defect kinetics is suggested by slowly-decreasing behavior of the photocurrent observed in an a-Si:H/SOI stack during an annealing period, denoted by (ii)-(vi) in figure 9(b). Such a decrease in the photocurrent during annealing is not observed for the other two experiments described in figures 9(a) and (c). So, it suggests that for the a-Si:H/SOI stack, the interface defects are generated by annealing ( figure 11(c)), if the annealing temperature is not appropriate. The interface defects are considered to be generated by the relaxation of interface stress [145,146], which involves the bond breaking of weak bonds and reorganization of the interfacial structure.

Advanced processing for surface passivation improvement
According to the above-mentioned defect kinetics, the surface passivation is improved via the suppression of defect generation in silicon at the beginning of the a-Si:H growth and the formation of a low-defect ultrathin layer. To do that, it is beneficial to use a low dilution of H 2 or pure SiH 4 as a source gas under a low-power discharge condition [135,136]. Reducing the H-atom supply to the silicon surface at the beginning of the a-Si:H growth is effective to suppress the defects in silicon [65]. The low-power discharge is also effective to suppress the defect generation by the bombardment of ions [57,58]. The low-power discharge, is also beneficial for reducing the polymerization and particulate generation [124,[147][148][149][150], which may contribute to the formation of a lowdefect ultrathin layer. It is also useful to develop the multi-step deposition and annealing at an appropriate temperature. With these process developments based on above-mentioned defect kinetics, an excellent surface passivation is obtained, which shows a high-efficiency solar cell performance, presented in recent reports [101,102].

Further improvement and future prospective
The in-situ defect detection described above, based on the pump-probe photocurrent measurement [97,99], is relatively simple and easy-to-use, with respect to the previously developed TRMC [93,94] and CRDS [95,96] (section 4 and  table 1). Besides, the measurement setup is not large, it is compact. This compact setup allows to assemble it into various plasma processing equipments to study the defect kinetics.
The measurement method is expected to further improve its sensitivity and spatial resolution. At present, the photoand trap current signals are detected by a lock-in technique [151], as described in section 5.2. The lock-in technique is a key to extract the signals from various noise/fluctuation associated with processing plasmas. This lock-in technique is a homodyne scheme [152,153], where the signal is detected with the reference that has the frequency same with that. On the other hand, a heterodyne scheme [153,154] is known to be robust against noise and fluctuation. The heterodyne scheme uses the reference that has a frequency different from that of the signal, and the difference frequency is employed to detect it. As a result, the detection frequency is lowered, and increasing the reference amplitude improves the detection sensitivity. So, the heterodyne scheme is one of the possible approaches to improve the sensitivity. Also, a pseudo heterodyne scheme [155,156], using the phase-controlled references, is considered to be another approach to improve the signal detection.
In the present setup, the spatial resolution is limited to be ≈100 µm or larger (table 1), by the channel length/width associated with the dimensions of interdigitated contacts for the photo-and trap current measurements. To improve it, the measurement combined with a concept of near-field spectroscopy [156][157][158] is one of the approaches. With nearfield effects, the intensity of light is localized and enhanced.
The spatial localization and intensity enhancement depend on the configuration of contacts. So, with an appropriate structure of contacts, the spatial localization as well as the detection sensitivity are improved. Scanning the localized spot of light over the channel may add the mapping capability.
In this article, the plasma-induced defects are focused on silicon-based materials, particulary for solar cell applications. However, the presented in-situ method is not limited by the materials of silicon. The method is able to apply to other semiconductor materials under a variety of plasma processing environments such as etching, implantation, sputtering, and surface treatments. By changing the sample structure and materials, the kinetics of plasma-induced defects can be studied in state-of-the-art logic semiconductors [12,13], high-sensitivity image sensors [159], high-frequency optoelectronic devices [160], and high-performance power devices [161].

Summary
In this review, the plasma-induced defects are overviewed, according to the recent research and progress. The generation of defects is concisely explained in terms of the plasmamaterial interactions. The defect structures, energy levels and carrier dynamics are described to explain the impact of defects on the device performances. The detection methods of defects are summarized, and our unique in-situ detection method is introduced. The method is based on the pump-probe photocurrent measurement, which is sensitive to the defects as well as trapped carriers. To demonstrate it, the pump-probe operation of carriers is presented for a device-grade thin film of a-Si:H. Two examples of plasma-induced defects are then presented in actual plasma processing for a-Si:H growth and crystalline silicon surface passivation. In these processing, the plasmainduced defects are successfully detected in-situ and real-time, confirming that the method is useful for studying the defect kinetics.
According to the experiments, the following conclusions are drown. (i) The plasma-induced defects are concentrated near the material surface in a depth of a few nanometers or more, depending on the penetration depth and reactions of ions, photons and radical species. (ii) The defects are generated by radical species of H atoms, in addition to the wellknown effects of ion bombardment and photon irradiation. The control of radical species, i.e. generation, diffusion and reactions, is important for advanced processing and defect management. (iii) The plasma-induced defects are annihirated/recovered by annealing. The annealing should be properly performed to obtain its effects as much as possible. (iv) The device performance is improved via the process developments, based on the knowledge of defect kinetics. As an example, high-efficiency solar cells are fabricated via advanced plasma processing and annealing. The advanced plasma processing and defect management become more important for the next-generation high-performance nanoscale device fabrication.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).