Stability Improvement of Nitrogen Doping on IGO TFTs under Positive Gate Bias Stress and Hysteresis Test

Thin-ﬁlm transistors (TFTs) using indium-gallium-oxide (IGO) semiconductor materials as channel layers were fabricated. In this study, nitrogen was introduced in the process of channel deposition to investigate its effect on device performance. The experimental results showed that moderate nitrogen doping can signiﬁcantly improve the device stability under positive gate bias stress owing to the reduction of oxygen vacancies. Furthermore, for the purpose of understanding the inﬂuence of different doping levels, the nitrogen doping ratio was modulated in ascending order from 0 sccm to 5 sccm. Among the fabricated c-IGO TFTs, the one with 2 sccm nitrogen doping exhibited the least threshold voltage shift. In addition, the hysteresis measurement further conﬁrmed that the interface traps between the channel and the dielectric were signiﬁcantly passivated in nitrogen-doped TFT. In this regard, the method of in-situ nitrogen doping was certiﬁed to serve an efﬁcient way of fabricating a passivation-free TFT and improve the device stability simultaneously. © The

Recently, transparent oxide semiconductors (TOSs) have been the object of extensive research in various connected fields. Owing to the advantages of high mobility, good transparency and ideal uniformity, TOSs are more suitable for the application of thin-film transistors (TFTs) than conventional Si TFTs. 1,2 In addition, the features of low-temperature process and their compatibility with flexible electronic enable TOSs to become the mainstream channel materials in next generation flat panel displays, such as active-matrix liquid crystal displays (AMLCDs) 3 and active-matrix organic light-emitting diodes (AMOLEDs). As compared to most TOSs, In 2 O 3 (∼3.7 eV) and Ga 2 O 3 (∼4.9 eV) possess wide energy bandgap, and this merit makes IGO thin film demonstrate better transparency. By adjusting each stoichiometry, transparent IGO conductive film can be widely used for diverse optoelectronic devices. Besides, the advantage of having higher mobility also makes In 2 O 3 -based materials capable of providing larger driving current.
However, metal oxide semiconductors are very sensitive to the environment impact. As a general practice, the application of passivation layer contributes to the enhancement of device stability by isolating channel layer from the surrounding environment. 4 Unfortunately, the procedure of extra passivation layer deposition not only increases the fabrication complexity and production cost, but it also damages the active layer heavily. 5 Hence, doping an extra element in the process of channel deposition has become a popular gesture. Particularly, many researches have begun to study the effect of nitrogen doping on device characteristics. 6,7 It was found that nitrogen doping helps control the amount of oxygen vacancies and avoid excessive carriers remaining in bulk channel, ameliorating the device instability. Thin-film transistors with TOS channel like ZnO, 8 IGZO, 9 SnSiO, 10 and InSnO 11 had been reported to have better electrical performance due to in-situ nitrogen doping. Nevertheless, the investigation of stability concerning indium-gallium-oxide (IGO) TFTs under positive gate bias stress has not been studied yet. In this work, we researched the influences of nitrogen doping on the IGO TFT. The relevant material analysis would be discussed here. Figure 1 demonstrates the structure of inverted-staggered IGO TFTs. The devices with slightly nitrogen-doped active layers were prepared on quartz substrates in this work. In the beginning, the bottom gate electrode of 80-nm-thick aluminum (Al) was deposited by E-beam evaporation through the shadow mask and capped with a 200-nm-thick SiO 2 gate dielectric by plasma-enhanced chemical vapor deposition (PECVD) at 300 • C subsequently.

Experimental
The active layers of 10-nm-thick IGO films with different nitrogen doping ratios were sputtered on the dielectric by radio frequency (RF) magnetron sputtering using the IGO target (In 2 O 3 :Ga 2 O 3 = 90:10 wt%) at room temperature. In the process of channel deposition, the fluxes of argon (Ar) and oxygen gas (O 2 ) were fixed at 40 and 10 sccm, respectively. In order to investigate the influence of nitrogen doping in varying degrees on device performance, the nitrogen gas (N 2 ) with five different nitrogen doping fluxes (0, 1, 2, 3 and 5 sccm) was introduced into the chamber. The exact gas ratio of Ar:O 2 :N 2 were 40:10:0, 40:10:1, 40:10:2, 40:10:3 and 40:10:5 sccm. The total working pressure was kept at 5 mTorr with an RF discharge power of 80 W. Finally, 80-nm-thick Al films were evaporated onto the IGO channel, serving as source/drain (S/D) electrodes, via E-beam evaporation. Both the IGO channel and S/D electrodes were also patterned by specific shadow masks, where the channel length (L) and width (W) of the fabricated TFTs were 10 and 100 μm, individually. To con-  duct material analysis of the IGO thin films, X-ray diffraction (XRD) with a Cu Kα radiation source (λ = 1.54056 Å) and atomic force microscope (AFM) were used to characterize the crystallinities and the surface morphology. In addition, X-ray photoelectron spectroscopy (XPS) was applied to analyze the oxygen chemical state (O1s peaks) for IGO films with different percentages of nitrogen gas. The transfer characteristics of IGO TFTs were measured in darkness and in atmosphere environment at room temperature using an Agilent B1500 semiconductor parameter analyzer. Figure 2 shows the XRD spectra of the IGO films with different nitrogen doping ratios to observe their crystalline states. Apart from the broad halo peak of quartz substrate around 2θ = 23 • , a crystalline peak, located approximately at 31 • , was found, primarily due to the crystalline phase of In 2 O 3 based on the standard X-ray diffraction patterns of JCPDS. The fabricated IGO films had a specific orientation and appeared in polycrystalline phase when the proportion of Ga element was low. 12 It can be seen that, despite the increase in doping ratio, the introduction of nitrogen did not change the crystallinity of the IGO films. Figure 3 presents the transfer curves of IGO TFTs with 0, 1, 2, 3 and 5 sccm nitrogen-doped channel, respectively, where the drain current (I DS ) was a function of the gate voltage (V GS ), with a fixed drain voltage (V DS ) of 8 V. The corresponding calculated electrical parameters such as field effect mobility (μ FE ), subthreshold swing (SS), threshold voltage (V TH ), on-off current ratio (I ON /I OFF ), and density of trap states (N SS ) of these TFTs were summarized in Table I. The threshold voltage was characterized through the constant current method, where V TH was defined as the gate voltage when the drain current reached the level of 10 −8 A. 13 Figure 4 shows that the transfer curves of the IGO TFTs shifted to the right with the doping flow increasing from 0 to 5 sccm, where the threshold voltage also increased from −1.18 V to −1.10 V. Additionally, as shown in Table I, the on-current (I ON ), which was defined as the drain current with V GS = 20 V, was reduced from 6.1 × 10 −5 to 3.8 × 10 −6 A progressively. This result could be explained by the reduction of oxygen vacancies. According to relevant studies, 5,12 it is generally believed that the carrier concentration is strongly associated with the amount of oxygen vacancies in metal oxide films. It means that, during the growth of channel layer, the in-situ nitrogen doping filled the intrinsic oxygen vacancy and played the role of carrier suppressor. Therefore, we can presume that the incorporated nitrogen atom not only decreased the amount of free electron in channel layer but also made the device turn-on voltage more positive; this is coherent with earlier works. 14 Besides, the experimental result also revealed that the carrier mobility was decreased by nitrogen doping, which was in accordance with previous research as well. 4,15 From Table I, it is noticed that higher nitrogen concentration deteriorated the mobility more seriously. When the oxygen vacancies in bulk channel were filled with nitrogen atoms, they failed to provide more free electron carriers, and thus resulted in undesired mobility loss. The connection between mobility and carrier concentration could be manifested by the percolation conduction model mentioned in the related study. 16 Moreover, the subthreshold swing (SS), which represents the existence of defects in bulk channel and/or at the interface between channel and gate dielectric, degraded from 0.23 to 1.80 V/dec. The degradation of SS was driven by the elevation of the number of undesirable traps originating from nitrogen doping process. Hoffman et al. discovered that the deep defects in metal oxide materials, including undercoordinated oxygen atoms and dangling bonds, etc., 17,18 could affect subthreshold swing extremely. Siddiqui et al. also reported defect creation is the fundamental cause for the degradation of mobility and subthreshold swing of ZnO TFTs under NBTS. 19 With regard to the strong correlation between subthreshold swing and deep trap states, the trap density (N ss ) was calculated using the following equation: 20

Results and Discussion
where k is Boltzmann constant, T is absolute temperature, q is electron charge and C i is the capacitance of gate dielectric per unit area.   It increased from 4.10 × 10 11 to 3.21 × 10 12 cm −2 with the nitrogen doping up to 5 sccm. From a certain perspective, excessive nitrogen atoms, which were considered as doping impurities, not only induced the strained bonds but disturbed the electron transportation. 21 However, the incorporation of nitrogen had advantageous effect on device stability. The time evolution of the transfer curves under positive gate bias stress (PBS) are shown in Figs. 4a-4e, where the positive gate bias stress was performed with the gate bias fixed at +20 V, and with the source/drain electrodes grounded for a stress time of 1000 seconds. The positive V TH shifts of the IGO TFTs with nitrogen doping of 0, 1, 2, 3 and 5 sccm were 7.89 V, 7.56 V, 4.39 V, 7.04 V, and 7.30 V, respectively. The nitrogen doping hereby contributed to a significant reduction in V TH . Fig. 4f clearly reveals that 2 sccm nitrogen-doped IGO TFT possessed the optimized stability in the PBS test among all the fabricated devices, where its V TH was 4.39 V. Normally, the bias-stress-induced V TH was mainly concerned with the generation of defects in the bulk channel and/or at the interface between channel and dielectric. 22 By means of nitrogen doping process, the oxygen vacancy related defects could be effectively diminished, and restrained the threshold voltage shift. Another advantage is that some researchers confirmed that, from the aspect of bonding energy, the metal-nitrogen bonding strengthened the lattice structure and was more stable as compared with the metal-oxygen bonding. 23,24 .
However, with increasing the nitrogen doping concentration, the IGO TFT with 3 and 5 sccm doping exhibited inferior PBS stability than the one doped with 2 sccm. This tendency, where the V TH shift declined at first but became larger after the doping ratio rose over 2 sccm, was the result of too much intrinsic deep traps induced by excessive nitrogen doping in the bulk channel, and it also led to the degradation of subthreshold swing. Figure 5 indicates that the application of positive bias stress degraded the subthreshold swing, which meant the increment of interface trap and/or defect trap creation after the bias stress, as further shown in Table II. Additionally, the similar situation also occurred in Fig. 6, making the carrier mobility decrease. These results showed that the degradation for 2 sccm device was least severe, implying only moderate nitrogen doping showed the best electrical stability.
To further clarify the associated mechanism, XPS measurement was conducted to characterize the chemical bonding states of the channel layer with varied nitrogen doping ratio. The deconvolution of XPS spectrum of O1s were analyzed in Figs. 7a-7e, with the three sub-peaks fitted by Gaussian fitting method. 25 The O1s could be fitted by three principal component peaks: O I was centered at 530.1 ± 0.2 eV, O II was centered at 531.5 ± 0.2 eV and O III was centered at 532.5 ± 0.2 eV. The three major sub-peaks, which are O I , O II , and O III , represent the oxygen ions combined with metal ions without oxygen deficiencies, the oxygen ions in the oxygen-deficient region, and the adsorbed oxygen from the environment, involving absorbed O 2 , −CO 3 , or −OH, 26,27 respectively. With the raise of nitrogen doping ratio, the proportional areas of sub-peaks were affected clearly, and demonstrated the alteration of O1s bonding states. O II , which stands for the amount of oxygen vacancy exiting in channel film peak, is especially important. The proportional area of O II /O total is usually calculated to characterize the level of oxygen vacancy related defects, where O total was defined by the sum of O I , O II and O III . 28 As seen from the comparison chart in Fig. 7f, O II /O total declined from 39.6% to 29.6% when nitrogen doping increased from 0 to 5 sccm. The measurement results revealed that the nitrogen doping reduced the amount of oxygen vacancies, as was consistent with the previous report for nitrogen-doped a-IGZO TFT. 15,29 Meanwhile, the decrement of O I /O total also proved that the nitrogen doping directly reduced the probability of the chemical combination between metal atom and oxygen atom. It should be noticed that O III /O total increased with nitrogen doping. This was because the nitrogen atoms, which are considered as doping impurities, induced strained bonds and hence affected crystalline structure of c-IGO film, leading to additional adsorption of ambient molecules from surrounding environment.
For a more detailed study of the influence of nitrogen doping, the surface morphologies of the IGO films doped in varying degrees were implemented by AFM. Figure 8 illustrates the variation of the surface root-mean-square (RMS) roughness, which decreased progressively in the case of higher doping concentration. It was also reflected in above-mentioned XRD results. We can reasonably presume that the smaller grain size suggested that more grain boundaries existed in the bulk channel and retained more defect traps. These defect traps interfered the carrier transport and brought about the worse carrier mobility, as agreed with the previous research. 30 To realize whether the method of nitrogen doping decreased the trap density, the IGO TFTs with the five N-doping ratios were prepared for the I-V hysteresis to estimate the number of defect traps. The application of hysteresis measurement further helped differentiate the interface traps between channel and dielectric from deep traps in bulk channel. 31 The gate voltage here was first swept by a forward sweep (F) from −15 V to +20 V and followed by a reverse sweep (R) from +20 V to −15 V subsequently. The clockwise hysteresis in transfer curves from 0 to 5 sccm nitrogen-doped TFTs are displayed in Figs. 9a-9e, separately. It is worth mentioning that the hysteresis window, encircled by both the forward sweep and the reverse sweep, was able to deduce the amount of interface traps between the dielectric and the channel layer. As shown in Fig. 9a, the undoped IGO TFT exhibited an obviously wider window with slight degradation in subthreshold swing, presenting an inferior hysteresis behavior. This was attributed to the phenomenon of charge trapping happening at the interface. 32 On the contrary, the transfer curve of the IGO TFT with 2 sccm nitrogen doping in Fig. 9c was shifted only slightly, and no subthreshold swing decay was measured, which directly accounted for the fact that no defect creation took place at the dielectric/channel interface after I-V hysteresis. 33 It showed that the nitrogen atoms repaired the dangling bonds and avoided a worse V TH shift during the positive gate bias stress. It should be noticed that the hysteresis window decreased after the PBS, showing that the accumulation of electron under positive gate bias reduced the interface traps between the channel layer and dielectric layer, as was confirmed in the previous literature. 34 Um et al. confirmed   moisture, 15,35 despite lacking extra passivation. The back-channel of TOSs TFT was susceptible to the surrounding gas during its operation. Notably, the accumulation of adsorptive oxygen molecules, complying with the reaction of O 2 (g) + e − → O 2 − (s), would capture free electrons from bulk channel and lower the conductivity. This interaction between the active layer and the ambient environment also accelerated the positive V TH shift given the positive bias stress. Therefore, it can be concluded that oxygen adsorbed by the positive bias stress was responsible for the degradation of device instability, but the introduction of moderate amounts of nitrogen into IGO active layer enhanced the device stability by well control of oxygen vacancies and defect traps at the interface.

Conclusions
In summary, the electrical characteristics of IGO TFTs with nitrogen doping at different levels were investigated. Though the introduction of excessive nitrogen imposed serious impact on both mobility and subthreshold swing, this study proved that moderate nitrogen doping not only lowered the number of oxygen vacancy related defects, but also passivated the interface trap states effectively. Owning to the reduction of oxygen vacancy related defects, IGO TFT with 2 sccm nitrogen doping exhibited superior positive bias stress stability relative to that of undoped TFT. Moreover, the application of I-V hysteresis confirmed the interface trap states were reduced by nitrogen doping as well. For these reasons, it can be seen that the in-situ nitro-   gen doping process achieved the effect of improving device stability, and it served as channel passivation from environmental impact at the same time.