Analysis of Nitrogen-Doping Effect on Sub-Gap Density of States in a-IGZO TFTs by TCAD Simulation

In this work, the impact of nitrogen doping (N-doping) on the distribution of sub-gap states in amorphous InGaZnO (a-IGZO) thin-film transistors (TFTs) is qualitatively analyzed by technology computer-aided design (TCAD) simulation. According to the experimental characteristics, the numerical simulation results reveal that the interface trap states, bulk tail states, and deep-level sub-gap defect states originating from oxygen-vacancy- (Vo) related defects can be suppressed by an appropriate amount of N dopant. Correspondingly, the electrical properties and reliability of the a-IGZO TFTs are dramatically enhanced. In contrast, it is observed that the interfacial and deep-level sub-gap defects are increased when the a-IGZO TFT is doped with excess nitrogen, which results in the degeneration of the device’s performance and reliability. Moreover, it is found that tail-distributed acceptor-like N-related defects have been induced by excess N-doping, which is supported by the additional subthreshold slope degradation in the a-IGZO TFT.


Introduction
Currently, the backplane technology of amorphous InGaZnO (a-IGZO) thin-film transistors (TFTs) is attracting great attention for its use in pixel switching and driving units for next-generation display applications. The competitive advantage of a-IGZO TFT technology is that it can offer high current-driving capacity, high optical transparency, low power consumption and low process temperature compared with traditional Si-based TFTs [1][2][3]. Although a-IGZO TFT technology has made remarkable progress since it was first proposed by Nomura et al. in 2004, these devices still cannot achieve the desired performance and reliability due to high-density sub-gap states existing in the bandgap of a-IGZO [4,5]. It has been demonstrated that the sub-gap defects mainly originate from oxygen-vacancyrelated (V o -related) defects induced by the structural disorder in a-IGZO [5][6][7], which degrades the electrical properties and reliability of TFTs by trapping electrons or holes in the channel layer and interfacial region under bias, light, and thermal stress [8][9][10][11]. To enhance the device performance and reliability, an in-situ nitrogen-doping (N-doping) approach during the a-IGZO active layer deposition has been proposed to suppress V o defect generation [12][13][14]. For example, it has been demonstrated that N-doping can significantly improve the reliability of a-IGZO TFTs under positive gate-bias stress (PBS) and PBS with light illumination, since the N incorporated into a-IGZO will occupy the V o sites and suppress V o -related defect generation [14,15]. Moreover, it has also been reported that the device performance and reliability are simultaneously enhanced by N and H co-doping, which is ascribed to the passivation of the V o distributed at the active layer and interface region by forming N-H and Zn-N bonds [16]. Although V o -related defects can be efficiently passivated by N-doping, a fundamental physical understanding of the impact of N-doping on the distribution of sub-gap states in a-IGZO TFTs is lacking. Since the device performance and reliability basically depend upon the nature and density of sub-gap defect states [4,17], an in-depth systematic study of the impact of N-doping on the sub-gap density of states (DOS) in a-IGZO TFTs is the key to future process improvement and optimization.
In this work, the influence of N-doping on the sub-gap V o -related defects in a-IGZO TFTs is qualitatively analyzed using technology computer-aided design (TCAD) simulation [18]. It is found that the density of the interface V o trap states, bulk V o -related tail states and deep-level V o -related defect states of a-IGZO TFTs are significantly decreased by moderate N-doping, which is validated by the improvement in electrical properties and stability during PBS and sub-band illumination. In contrast, the DOS of a-IGZO TFT is increased when the a-IGZO TFTs are doped with excessive nitrogen atoms, which causes degeneration of the device performance and reliability. Meanwhile, it is confirmed that the tail-distributed acceptor-like N-related defects are formed by excessive N-doping, which leads to the degradation of subthreshold slope (SS) in a-IGZO TFT.

Experiments and Modeling Scheme
The a-IGZO TFT structure used for numerical simulation is shown in Figure 1a. The devices in this work were fabricated on n-type Si substrate. First, the gate insulator was composed of a 200 nm SiO 2 thin film grown by plasma-enhanced chemical vapor deposition (PECVD) with a rate of~50 nm/min at 350 • C. The 45-nm-thick a-IGZO thin films were then deposited by direct-current (DC) sputtering system with a various gas mixture of N 2 /(O 2 + N 2 ) = 0%, 20%, and 40% at a fixed Ar flow rate of 30 sccm. The composition of the ceramic target used was In:Ga:Zn = 2:2:1 in atomic ratio. Subsequently, the device active region was patterned by conventional photolithography and wet chemical etching. Next, the Ti/Au (30/70 nm) bi-layer drain/source contact electrodes were evaporated by e-beam evaporation, and is the active region was further patterned using lift-off technique, which resulted in the final device dimensions of W/L = 100 µm/20 µm. Finally, a 100-nm-thick SiO 2 passivation layer was deposited by PECVD. The fabricated devices were annealed in ambient air for 1 h at 300 • C.
where the and denote the density of donor-like tail and acceptor-like tail states. The and represent the Gaussian-distributed acceptor-like and donor-like states. The and are the effective density at the conduction band minimum (EC) and valence band maximum (EV), respectively. The and are the characteristic slope energy of valence-band tail states and conduction-band tail states. The and are the total density of Gaussian donor and acceptor states, respectively. The and are the corresponding peak energy. and are the corresponding characteristic decay energy.
In addition, the interface trap density ( ) at the a-IGZO/dielectric interfacial region can be described as [23]: where and represent the donor-like and acceptor-like interface trap density, respectively. The and denote the corresponding slope energy. The NTD and NTA represent the donor-like and acceptor-like tail states, respectively. The three gaussian curves represent the deep Vo-related states (NGD(Vo-related)), oxygen interstitials (NGA(Oi)), and shallow donor states (NGD(Vo + /Vo 2+ )), respectively. The inset is the schematic illustration of the interface trap density (Dit(E)). Figure 2 shows the simulated and experimental transfer characteristics of the a-IGZO TFTs under various N-doping conditions at VDS = 5 V. The simulation results, in consistency with the experimental data, were achieved by calibrating the DitA, NTA, NGA(Oi) and NGD(Vo + /Vo 2+ ), and the simulation parameters are extracted and summarized in Table  1. The total trap density of the 20% N-doping ratio a-IGZO TFT was significantly decreased compared to the undoped a-IGZO TFT. For example, the DitA is decreased from 2.5 × 10 13 eV −1 cm −2 to 8.0 × 10 12 eV −1 cm −2 , and the NTA was reduced from 8.0 × 10 19 eV −1 cm −3 to 1.0 × 10 19 eV −1 cm −3 . Correspondingly, the subthreshold slope (SS) was decreased from 0.8 V/dec to 0.6 V/dec, and the threshold voltage (Vth) was reduced from 5.0 V to 3.8 V. It has been demonstrated that the interface states and bulk traps in a-IGZO TFTs mainly

Results and Discussion
respectively. The inset is the schematic illustration of the interface trap density (D it (E)).
In the Silvaco TCAD Simulation tool, ATLAS, a physics-based device simulator, is used to perform the electrical characterization, which can reduce the cost and time needed for experimentation [19]. It is also a powerful tool to predict the electrical behavior of specified semiconductor structures by using the Poisson and the continuity equations, which describe the electronic phenomena and electrical transport mechanism [19]. Based on the DOSs model of the a-IGZO TFTs, the types of the sub-gap states in the TFT channel region and interface region are illustrated in Figure 1b. In the a-IGZO material, the sub-gap states are mainly classified as acceptor-like and donor-like states, which can be depicted by Gaussian distribution states and exponentially decaying band-tail states. The specific mathematical model is expressed as follows [20][21][22]: where the g TD (E) and g TA (E) denote the density of donor-like tail and acceptor-like tail states. The g GA (E) and g GD (E) represent the Gaussian-distributed acceptor-like and donorlike states. The N TA and N TD are the effective density at the conduction band minimum (E C ) and valence band maximum (E V ), respectively. The W TD and W TA are the characteristic slope energy of valence-band tail states and conduction-band tail states. The N GD and N GA are the total density of Gaussian donor and acceptor states, respectively. The E GA and E GD are the corresponding peak energy. W GA and W GD are the corresponding characteristic decay energy. In addition, the interface trap density (D it (E)) at the a-IGZO/dielectric interfacial region can be described as [23]: where D itD and D itA represent the donor-like and acceptor-like interface trap density, respectively. The W itA and W itD denote the corresponding slope energy.  Table 1. The total trap density of the 20% N-doping ratio a-IGZO TFT was significantly decreased compared to the undoped a-IGZO TFT. For example, the D itA is decreased from 2.5 × 10 13 eV −1 cm −2 to 8.0 × 10 12 eV −1 cm −2 , and the N TA was reduced from 8.0 × 10 19 eV −1 cm −3 to 1.0 × 10 19 eV −1 cm −3 . Correspondingly, the subthreshold slope (SS) was decreased from 0.8 V/dec to 0.6 V/dec, and the threshold voltage (V th ) was reduced from 5.0 V to 3.8 V. It has been demonstrated that the interface states and bulk traps in a-IGZO TFTs mainly originate from V o -related defects [5,15,24]. Therefore, the simulation results confirm that the improved electrical properties of a-IGZO TFTs can be ascribed to the suppression of the generation of V o -related defects in the device channel and interface region by N-doping. In contrast, when the N-doping ratio was increased to 40%, the number of total trap states was increased compared to the 20% N-doping ratio TFT, as shown in Table 1. Meanwhile, it was observed that the SS and V th of the 40% N-doping ratio device were increased to 0.9 V/dec and 7 V, respectively, which indicates that V o -related defects generate when the device is subjected to excessive N-doping. This result can be explained by the fact that the formation of N-Ga bonds is facilitated by heavy N-doping, which then suppresses the bonding of Ga-O in the a-IGZO thin films [14,25].

Results and Discussion
In addition, based on the simulation results, it was found that although the sub- ) existing in the 40% N-doping ratio TFT was significantly higher than that of the 20% N-doping ratio TFT, the sub-gap DOS other than N TA was lower than that of the undoped TFT. Because the sub-band-gap density in a-IGZO film mainly originates from V o -related defects, the amount of V o in annealed a-IGZO thin films with various N-doping ratios was analyzed by X-ray photoelectron spectroscopy (XPS). Figure 3a-c show the O 1 s XPS spectra of the a-IGZO films grown using different N-doped ratios. The binding energies were calibrated by taking the C 1s as reference at 284.6 eV. Gaussian fitting was applied to decompose the combined O 1s peak. The sub-peaks centered at binding energies of 529.7 eV, 530.5 eV, and 531.5 eV were attributed to O 2− ions surrounded by metal atoms (In, Ga and Zn), oxygen vacancies (V o ), and OH − impurities, respectively [26][27][28]. The relative level of V o in a-IGZO film can be estimated by the proportion of peak area V o to the whole O 1s (O whole ). It was found that the area proportion of V o /O whole was decreased from 35% for N-free a-IGZO film to 25% for a-IGZO film with the 20% N-doped ratio, as shown in Figure 3a,b, indicating that V o decreases when the N is incorporated into the a-IGZO film. However, as shown in Figure 3c, it was found that the V o increased to 31% for the a-IGZO film deposited with the 40% N-doped ratio, which means that the additional V o was created when excess N atoms were doped into the a-IGZO film. Furthermore, the N 1s spectra XPS of the annealed a-IGZO film with 20% N-doped ratio was also analyzed, as shown in Figure 3d. The N 1 s spectrum is decomposed into two peaks at 395.7 and 397.3 eV, which are associated with the Ga Auger and N-Ga bonding [29], respectively. Therefore, the XPS results reveal that moderate N doping in a-IGZO film can suppress the generation of V o , and excess N incorporation into a-IGZO film leads to an increase in V o . Because the V o existing in the 40% N-doping ratio a-IGZO film was lower than that of undoped a-IGZO film, the increased N TA in 40% N-doping ratio a-IGZO TFT should be the result of the generation of N-related defects by excess N-doping [16,30], which agrees well with the increased SS from 0.8 V/dec to 0.9 V/dec compared to undoped a-IGZO TFT. Meanwhile, to quantitatively estimate the N concentration in the a-IGZO active layer, the actual level of N-doping in annealed a-IGZO films is characterized by secondary ion mass spectrometry (SIMS) measurement [31][32][33]. Figure 4 shows the depth profile of N concentration in the a-IGZO film deposited with the 20% and 40% N-doped ratios. N is clearly detectable, and there is a considerable amount of incorporated nitrogen (~10 20 cm −3 ) in the a-IGZO film. It has been reported that the value of the sub-gap density of states (DOSs) near the VBMs is about 5-9 × 10 20 cm −3 in a-IGZO film [5,34]. In this work, it is found that when the concentration of N doping in the channel region of a-IGZO TFT with a 20% N doping ratio is~1.0 × 10 20 cm −3 , the electrical performance and stability of the device are dramatically improved. But when the concentration of N-doping in the channel region of a-IGZO TFT with a 40% N doping ratio is increased to~1.2 × 10 20 cm −3 , the electrical performance and stability of a-IGZO TFT are degraded.
Micromachines 2022, 13, 617 4 the formation of N-Ga bonds is facilitated by heavy N-doping, which then suppresse bonding of Ga-O in the a-IGZO thin films [14,25]. In addition, based on the simulation results, it was found that although the sub DOS (NTA, NTD, NGD(Vo-related), NGA(Oi), and NGD(Vo + /Vo 2+ )) existing in the 40% N-do ratio TFT was significantly higher than that of the 20% N-doping ratio TFT, the sub DOS other than NTA was lower than that of the undoped TFT. Because the sub-band      According to the distribution of the sub-gap DOS fitted in a-IGZO TFTs with various N-doping ratios, a comprehensive quantitative study on the device stability under positive bias stress (PBS) was carried out. During the PBS process, the TFTs were applied at a V GS of 15 V for the stress duration of 5000 s. Figure 5a-c show the experimental and simulated evolution of transfer characteristics as a function of PBS time for the a-IGZO TFTs with different N-doping ratios. It was found that the shift in threshold voltage (∆V th ) induced by PBS was 2.06 V, 0.8 V, and 1.68 V for undoped a-IGZO TFT, 20% N-doping a-IGZO TFT, and 40% N-doping a-IGZO TFT, respectively. It has been reported that the shift in V th (∆V th ) of a-IGZO TFTs under PBS basically originates from the interfacial V o -related defects trapping electrons at the device interfacial region [35,36]. In the simulation results, it was clearly seen that the D it (E) for the 20% N-doping ratio a-IGZO TFT was lower than of the undoped a-IGZO TFT and 40% N-doping ratio a-IGZO TFT. For example, the D itA for undoped a-IGZO TFT, 20% N-doped a-IGZO TFT, and 40% N-doped a-IGZO TFT is 2.5 × 10 13 eV −1 cm −2 , 8.0 × 10 12 eV −1 cm −2 , and 1.5 × 10 13 eV −1 cm −2 , respectively, suggesting that interfacial V o -related defects can be suppressed by moderate N-doping.

Parameters
Undoping 20% N-Doping Ratio   According to the distribution of the sub-gap DOS fitted in a-IGZO TFTs with various N-doping ratios, a comprehensive quantitative study on the device stability under positive bias stress (PBS) was carried out. During the PBS process, the TFTs were applied at a VGS of 15 V for the stress duration of 5000 s. Figure 5a-c show the experimental and simulated evolution of transfer characteristics as a function of PBS time for the a-IGZO TFTs with different N-doping ratios. It was found that the shift in threshold voltage (∆Vth) induced by PBS was 2.06 V, 0.8 V, and 1.68 V for undoped a-IGZO TFT, 20% N-doping a-IGZO TFT, and 40% N-doping a-IGZO TFT, respectively. It has been reported that the shift in Vth (∆Vth) of a-IGZO TFTs under PBS basically originates from the interfacial Vorelated defects trapping electrons at the device interfacial region [35,36]. In the simulation results, it was clearly seen that the Dit(E) for the 20% N-doping ratio a-IGZO TFT was lower than of the undoped a-IGZO TFT and 40% N-doping ratio a-IGZO TFT. For example, the DitA for undoped a-IGZO TFT, 20% N-doped a-IGZO TFT, and 40% N-doped a-IGZO TFT is 2.5 × 10 13 eV −1 cm −2 , 8.0 × 10 12 eV −1 cm −2 , and 1.5 × 10 13 eV −1 cm −2 , respectively, suggesting that interfacial Vo-related defects can be suppressed by moderate N-doping. In addition, it has been reported that weak oxygen ions originating from structural disorder in a-IGZO TFTs cause ∆Vth under PBS [37]. During the PBS process, the weak oxygen ions are ionized to generate oxygen interstitials (Oi) because of their low formation energies [4,38,39]. Meanwhile, according to the first-principle studies, the generated Oi during PBS forms an octahedral configuration [Oi(oct)] and is electrically active. Correspondingly, the introduced Oi(oct)-related defect states are distributed above the mid-gap (Ei) in the a-IGZO TFTs [40]. When the Fermi level moves up under PBS, the Oi(oct)-related states are filled by trapping electrons and thus negatively charged to generate Oi 2− oct) [41]. Figure 6a-c show the forming process of Oi 2− (oct)-related charged states in a-IGZO TFTs undergoing PBS. Because of the structural relaxation effect, the Oi 2− (oct)-related charged states are transformed into deep-level negative-U states, which are located below the mid-gap in the a-IGZO TFTs [37,42]. As a result, although new Oi 2− (oct)-related In addition, it has been reported that weak oxygen ions originating from structural disorder in a-IGZO TFTs cause ∆V th under PBS [37]. During the PBS process, the weak oxygen ions are ionized to generate oxygen interstitials (O i ) because of their low formation energies [4,38,39]. Meanwhile, according to the first-principle studies, the generated O i during PBS forms an octahedral configuration [O i (oct)] and is electrically active. Correspondingly, the introduced O i (oct)-related defect states are distributed above the mid-gap (E i ) in the a-IGZO TFTs [40]. When the Fermi level moves up under PBS, the O i (oct)-related states are filled by trapping electrons and thus negatively charged to generate O i 2− oct) [41].  According to the simulation results, NGA(Oi) exhibited an increasing trend for a-IGZO TFTs with various N-doping ratios under the PBS process, as shown in Table 2. For example, the NGA(Oi) in the N-free a-IGZO TFT continuously increased from 2.6 × 10 17 eV −1 cm −3 to 3.6 × 10 17 eV −1 cm −3 after 5000 s PBS. This result shows that the Oi(oct)-related defects were generated in the a-IGZO TFT upon PBS, originating from weak oxygen ions in the device channel region. Compared to the undoped a-IGZO TFT, the generated Oi(oct)-related trap states under PBS in the 20% N-doping ratio a-IGZO TFT decreased from 3.6 × 10 17 eV −1 cm −3 to 2.5 × 10 17 eV −1 cm −3 after 5000 s PBS, due to the suppression of Vo-related traps by N-doping. Correspondingly, the device had superior electrical reliability under PBS. In contrast, the generated Oi(oct)-related trap states under PBS in the 40% N-doping ratio a-IGZO TFT (3.2 × 10 17 eV −1 cm −3 ) were higher than that of the 20% N-doping ratio a-IGZO TFT, due to the Vo-related traps generated by heavy N-doping, resulting in devicestability degeneration under PBS. Finally, to reveal the impact of N-doping on the distributions of deep-level sub-gap states, the transfer characteristics of a-IGZO TFTs with various N-doping conditions under sub-band-gap light illumination were simulated. The simulation results are shown in Figure 7a-c, and the simulation parameters are extracted in Table 3. It was found that the I-V curves of the undoped a-IGZO TFT exhibited an overall shift in a negative direction with the decrease in incident light wavelength from 650 nm to 500 nm, as shown in Figure  7a. The negative ΔVth is attributable to the photorelease of occupied electrons in the interface states and deep-level sub-gap states. It has been demonstrated that the deep-level defects in a-IGZO mainly originate from neutral Vo, which would be entirely occupied above EV with an energy width of ~1.5 eV [5,24,43]. As a result, the occupied interfacial and deep-level Vo-related defects would be ionized into Vo + /Vo 2+ under the corresponding photon energy illumination, which agrees well with the simulation result that the NGD(Vo + /Vo 2+ ) exhibits continuously increase as the illumination wavelength decreases  Finally, to reveal the impact of N-doping on the distributions of deep-level sub-gap states, the transfer characteristics of a-IGZO TFTs with various N-doping conditions under sub-band-gap light illumination were simulated. The simulation results are shown in Figure 7a-c, and the simulation parameters are extracted in Table 3. It was found that the I-V curves of the undoped a-IGZO TFT exhibited an overall shift in a negative direction with the decrease in incident light wavelength from 650 nm to 500 nm, as shown in Figure 7a. The negative ∆V th is attributable to the photorelease of occupied electrons in the interface states and deep-level sub-gap states. It has been demonstrated that the deep-level defects in a-IGZO mainly originate from neutral V o , which would be entirely occupied above E V with an energy width of~1.5 eV [5,24,43]. As a result, the occupied interfacial and deeplevel V o -related defects would be ionized into V o + /V o 2+ under the corresponding photon energy illumination, which agrees well with the simulation result that the N GD (V o + /V o 2+ ) exhibits continuously increase as the illumination wavelength decreases [44,45], as shown in Table 3. It is clear that the N GD (V o + /V o 2+ ) of the undoping a-IGZO TFT is increased from 1.5 × 10 17 eV −1 cm −3 to 3.0 × 10 17 eV −1 cm −3 with the decrease of incident light wavelength from 650 nm to 500 nm. In addition, based on the first-principle calculation and experimental observation, the activation energy (E a ) of the photoexcited ionization process from occupied deep-level V o to V o + and V o 2+ is required to be~2.0 eV and~2.3 eV, respectively [37,45]. Meanwhile, these photo-induced transitions (both V o to V o + and V o to V o 2+ ) could cause the outward relaxation in the vicinity of metal atoms, which lead to the generation of new defect level near the E i and E c edge [24,45]. The formation process of V o + and V o 2+ states induced by sub-band-gap illumination is illustrated in Figure 8. In the simulation, it is observed that the generated N GA (V o + -related) near the mid-gap is 9.0 × 10 16 eV −1 cm −3 at λ = 600 nm (~2.0 eV), and the generated N GD (V o 2+ -related) near bottom of the conduction band is 1.2 × 10 17 eV −1 cm −3 at λ = 500 nm (~2.3 eV). It has been reported that the variation of SS (∆SS) is in connection with the amount of created trap states (∆N t ) in the TFTs channel and interface region, which is expressed by using the following equation [45]: where k is the Boltzmann's constant, T is the absolute temperature, C i is the capacitance of the gate dielectric. Therefore, the SS degradation for the undoping a-IGZO TFT induced by incident illumination of λ ≤ 600 nm is caused by the new defects creation near the E i and E c edge.
tion and experimental observation, the activation energy (Ea) of the photoexcited ionization process from occupied deep-level Vo to Vo + and Vo 2+ is required to be ~2.0 eV and ~2.3 eV, respectively [37,45]. Meanwhile, these photo-induced transitions (both Vo to Vo + and Vo to Vo 2+ ) could cause the outward relaxation in the vicinity of metal atoms, which lead to the generation of new defect level near the Ei and Ec edge [24,45]. The formation process of Vo + and Vo 2+ states induced by sub-band-gap illumination is illustrated in Figure 8. In the simulation, it is observed that the generated NGA(Vo + -related) near the mid-gap is 9.0 × 10 16 eV −1 cm −3 at λ = 600 nm (~2.0 eV), and the generated NGD(Vo 2+ -related) near bottom of the conduction band is 1.2 × 10 17 eV −1 cm −3 at λ = 500 nm (~2.3 eV). It has been reported that the variation of SS (∆SS) is in connection with the amount of created trap states (∆Nt) in the TFTs channel and interface region, which is expressed by using the following equation [45]: where k is the Boltzmann's constant, T is the absolute temperature, Ci is the capacitance of the gate dielectric. Therefore, the SS degradation for the undoping a-IGZO TFT induced by incident illumination of λ ≤ 600 nm is caused by the new defects creation near the Ei and Ec edge.    Comparatively, it is found that the density of deep-level V o -related traps is suppressed by moderate N-doping into the a-IGZO TFT. As shown in Table 3, the N GD (V o -related) in the 20% N-doping ratio a-IGZO TFT is decreased from 8.0 × 10 20 eV −1 cm −3 to 5.0 × 10 20 eV −1 cm −3 compared with the undoping a-IGZO TFT, and the generated N GD (V o 2+ -related) is reduced from 1.2 × 10 17 eV −1 cm −3 to 7.0 × 10 16 eV −1 cm −3 under λ= 500 nm. Correspondingly, the electrical stability of the 20% N-doping ratio a-IGZO TFT under light illumination is significantly improved. It is found that the ∆SS and ∆V th in the 20% N-doping ratio a-IGZO TFTs (0.58 V/dec; −0.8 V) are lower than that of undoping a-IGZO TFT (1.95 V/dec; −1.6 V) at λ = 500 nm, as shown in Figure 7b. Therefore, it can be concluded that the improved device reliability under sub-band light illumination is owing to the passivation of V o -related traps in the device channel region by moderate N-doping. However, the deep-level V o -related defect in the 40% N-doping ratio a-IGZO TFT is increased to 6.5 × 10 20 eV −1 cm −3 compared with the 20% N-doping ratio a-IGZO TFT, and the generated N GD (V o 2+ -related) is increased to 9.0 × 10 16 eV −1 cm −3 under λ = 500 nm. Meanwhile, the degradation of SS and V th (1.5 V/dec; −1.3 V) are observed in the 40% N-doping ratio a-IGZO TFT. This result means that superfluous N-doping into the a-IGZO TFTs will result in the increase of deep-level V o traps [25], which degrades the device stability under sub-band-gap illumination. Comparatively, it is found that the density of deep-level Vo-related traps is suppressed by moderate N-doping into the a-IGZO TFT. As shown in Table 3, the NGD(Vorelated) in the 20% N-doping ratio a-IGZO TFT is decreased from 8.0 × 10 20 eV −1 cm −3 to 5.0 × 10 20 eV −1 cm −3 compared with the undoping a-IGZO TFT, and the generated NGD(Vo 2+related) is reduced from 1.2 × 10 17 eV −1 cm −3 to 7.0 × 10 16 eV −1 cm −3 under λ= 500 nm. Correspondingly, the electrical stability of the 20% N-doping ratio a-IGZO TFT under light illumination is significantly improved. It is found that the ∆SS and ∆Vth in the 20% N-doping ratio a-IGZO TFTs (0.58 V/dec; −0.8 V) are lower than that of undoping a-IGZO TFT (1.95 V/dec; −1.6 V) at λ = 500 nm, as shown in Figure 7b. Therefore, it can be concluded that the improved device reliability under sub-band light illumination is owing to the passivation of Vo-related traps in the device channel region by moderate N-doping. However, the deep-level Vo-related defect in the 40% N-doping ratio a-IGZO TFT is increased to 6.5 × 10 20 eV −1 cm −3 compared with the 20% N-doping ratio a-IGZO TFT, and the generated NGD(Vo 2+ -related) is increased to 9.0 × 10 16 eV −1 cm −3 under λ = 500 nm. Meanwhile, the degradation of SS and Vth (1.5 V/dec; −1.3 V) are observed in the 40% N-doping ratio a-IGZO TFT. This result means that superfluous N-doping into the a-IGZO TFTs will result in the increase of deep-level Vo traps [25], which degrades the device stability under sub-bandgap illumination.

Conclusions
In this work, the fundamental physical understanding of the N-doping on DOS over the whole sub-band-gap range has been analyzed by Silvaco TCAD simulation. It is found that the improved electrical performances for the 20% N-doping ratio a-IGZO TFT are owing to the suppression of interface Vo trap states and bulk tail states (Vo-related) by Ndoping. Meanwhile, the Oi and deep-level Vo-related traps are suppressed by an appropriate amount of N dopant, which causes the improvement of device stability during PBS and sub-band illumination processes by suppressing the formation of Oi and the photoexcited ionization from occupied deep-level Vo to Vo + and Vo 2+ , respectively. In contrast, the excessive N-doping will cause the generation of acceptor-like N-related defects and the increase of Vo-related traps in the channel and interface region of a-IGZO TFTs, which leads to the degeneration of the device performance and reliability.

Conclusions
In this work, the fundamental physical understanding of the N-doping on DOS over the whole sub-band-gap range has been analyzed by Silvaco TCAD simulation. It is found that the improved electrical performances for the 20% N-doping ratio a-IGZO TFT are owing to the suppression of interface V o trap states and bulk tail states (V o -related) by N-doping. Meanwhile, the O i and deep-level V o -related traps are suppressed by an appropriate amount of N dopant, which causes the improvement of device stability during PBS and sub-band illumination processes by suppressing the formation of O i and the photoexcited ionization from occupied deep-level V o to V o + and V o 2+ , respectively. In contrast, the excessive N-doping will cause the generation of acceptor-like N-related defects and the increase of V o -related traps in the channel and interface region of a-IGZO TFTs, which leads to the degeneration of the device performance and reliability.