Microsecond Pulse I–V Approach to Understanding Defects in High Mobility Bi-layer Oxide Semiconductor Transistor

The carrier transport and device instability of amorphous oxide semiconductor devices are influenced by defects that are exponentially distributed in energy, because of amorphous phase channels and front/back interfaces with a large number of sub-gap states. Thus, understanding defects and charge trapping in oxide semiconductor transistors is required for being core device element in reliable production lines. In this paper, we present the transient charging effect, the charge trapping mechanism, and the dynamic charge transport of high-mobility bilayer oxide semiconductor transistors. To this end, we exploited microsecond ramps, pulse ID–VG, transient current, and discharge current analysis methods. The mobility enhancement rate of single HfInZnO (HIZO) and bilayer HfInZnO-InZnO (HIZO-IZO) were 173.8 and 28.8%, respectively, in the charge-trapping-free environment. Transient charge trapping can be classified to temperature insensitive fast charging and thermally activated slow charging with two different trap energies. Insignificant fast transient charging of a bilayer-oxide high-mobility thin film transistor(TFT) can be explained by the low density of sub-gap states in the oxide semiconductor. Understanding defects and transient charging in the oxide semiconductor helps to determine the origin of device instability of oxide TFTs, and finally, to solve this problem.

mobility in an environment with minimal charge trapping [42][43][44][45][46] . To this end, we employed microsecond fast ramp I-V (μs-FIV), pulse I-V (PIV), the transient current method, and discharge current analysis (DCA). The μs-FIV and PIV methods permit relatively fast sweeping and measurement times on the order of μs, unlike the conventional DC I-V method [47][48][49][50][51][52][53][54] . Thus, because we employ μs-FIV measurement, the charge trapping phenomenon induced by the defects can be minimized during measurement, enabling us to extract near-intrinsic transistor parameters. In addition, we extracted the sub-gap states of amorphous oxide TFTs with quantified parameters using the DCA method, which is a modified charge pumping method 55,56 . In the DCA measurement method, we applied a periodic pulse to the gate at a controlled frequency and measured the discharge current, thereby extracting the sub-gap states that are exponentially distributed in energy in the oxide TFT [57][58][59] . DCA result shows that the bilayer HIZO-IZO oxide TFT has lower sub-gap states in the oxide TFT, then in the single-layer HIZO oxide TFT. We also discuss the model of charging phenomena of single-layer HIZO and bilayer HIZO-IZO oxide TFTs using the transient current measurement method with varying temperature 36,51,52 . We observed fast and slow charge-trapping phenomena for both devices and provide a possible model following temperature-insensitive fast charging and thermally activated slow charging at two different trap energies. The various pulsed I-V techniques described in this paper are expected to help accurately extract the sub-gap density of states and the transistor parameters, and improve understanding of the impact of charging on oxide TFTs.

Results
As seen in Fig. 1(a), the cross-sectional high-resolution scanning transmission electron microscopy image shows the Mo bottom gate TFT with a top-contact Mo source/drain and passivation layer structure. An inverted staggered bottom-gate oxide TFTs were fabricated using a standard semiconductor fabrication process technology. For all of the electrode, Mo electrode was deposited by radio frequency (RF) sputtering using O 2 /Ar gas and subsequently growing a gate insulator of SiO 2 by the plasma-enhanced chemical vapor deposition (PECVD). As an active channel, four different type of semiconductor layer configurations, such as HIZO (40 nm), HIZO (20 nm)-IZO (20 nm), HIZO (20 nm)-IZO (40 nm) and IZO (60 nm) channel, were employed. Each active channel was deposited at room temperature by RF magnetron sputtering method. Subsequently, source/drain electrodes were sputtered at room temperature, followed by pattering process. Then PECVD SiO x passivation layer was formed and contact hole was patterned. The detail of fabrication can be found in the Methods (Device Fabrication). The transmission electron diffraction pattern images of the bilayer semiconductors, top IZO, and bottom HIZO, in the right part of Fig. 1(a), show tiny dots and characteristics lines in the hazy background image, implying that both semiconductor layers are composed of the nano-crystalline phase in an amorphous medium 4 . The energy dispersive spectroscopy data in Fig. 1(b) verifies that top and bottom layers are made of IZO and HIZO, respectively. For composition analysis, the concentrations of Hf, In, and Zn were determined from different HIZO and IZO samples using inductively coupled plasma atomic emission spectrometry (ICP-AES) on a Shimadzu ICPS-8100 sequential spectrometer. We found that for IZO, Zn/(In + Zn) and In/(In + Zn) are 0.51 and 0.49, respectively. For HIZO, Zn/(Hf + In + Zn), In/(Hf + In + Zn), and Hf/(Hf + In + Zn) are 0.4, 0.53, and 0.07. For electrical analysis on both samples, the Hall mobility and carrier concentration of HIZO and IZO were determined using Hall effect measurement (HIZO; μ: 17.2 cm 2 /V·s, n carrier : 2.71 × 10 15 cm −3 and IZO; μ: 48.4 cm 2 / V·s, n carrier : 9.93 × 10 15 cm −3 ), as shown in Fig. 1(c). In our study, the front HIZO semiconductor layer adjusts the threshold voltage and the back IZO semiconductor layer was designed as a high-mobility channel 10,60,61 . A schematic of the bilayer oxide TFT is presented in Fig. 2(a). For μs-FIV, PIV, and DCA measurements, we designed the device such that it minimized the overlap capacitances among the gate, source, and drain. Thus, all electrodes were patterned using conventional lithography and dry etching methods. A schematic of the μs-FIV/PIV measurement is shown in Fig. 2(b). For fast-pulse electrical testing, we used the waveform generator/fast measurement unit (WGFMU) module in the B1500A semiconductor parameter analyzer. For the DC I-V measurements, the voltage sweep rate was 1 V/s; for μs-FIV and PIV measurements, a single pulse (rise time and fall time 10 μs, pulse width 2 ms) was used. DCA was performed using a high-speed pulse generator (Agilent 81104 A) and a source meter (Keithley 2401) to obtain the quantitative parameters of the number of sub-gap states in the oxide TFT.  current level is increased because of both the high carrier concentration and the high mobility of IZO as the IZO thickness increases 60,61 . The DC I-V measurement method has a relatively long sweep/measurement time, and thus, the tested device experiences significant electrical stresses, thereby leading to significant charge trapping, mobility degradation, and device instability 42 . Therefore, we employed a fast I-V measurement method to minimize the fast trapping effect, thereby improving drain current and mobility. Properties were measured at a relatively rapid rise rate of several microseconds, as compared to conventional DC I-V. As shown in the inset in Fig. 2(c), the fast I-V measurement technique for the HIZO TFT compensates for the fast trapping effects, showing a higher drain current than the DC technique. Figure 2(d) shows the mobility versus voltage ramping time (i.e., 10 μs, 100 μs, 300 μs, and 500 μs). As expected, with shorter voltage ramping times, there is higher drain current because of minor charge trapping. Charge trapping depends on gate voltage ramping time. Slow voltage ramping results in high charge trapping. Such a microsecond short sweeping and measurement time can avoid and minimize some channel carriers to be trapped in the shallow-level defects placed in the semiconductor with a certain level of sub-gap states also gate insulator film or at the interface. Therefore, a fast I-V measurement method has been employed to minimize the effect of fast charging, leading to evaluate the near-intrinsic mobility.
Various mobility values of the four devices are presented Results show that the highest mobility was achieved with bilayer HIZO-IZO devices with relatively thick IZO layers. In addition, the mobility increase rates (μ fast /μ dc ) for HIZO (40 nm), HIZO (20 nm)-IZO (20 nm), HIZO (20 nm)-IZO (40 nm) and IZO (60 nm) TFTs were 173.8%, 38.5%, 28.8% and 23.8% respectively, indicating that bilayer devices with relatively thick IZO layers have fewer defects. As will be discussed below in detail, this phenomenon occurs because HIZO has a relatively large number of acceptor-like sub-gap states, causing charge trapping and device instability [62][63][64][65] . To further analyze the transient charging effect, we performed single-pulse I-V measurements. The voltage ramping profile for this measurement is shown in Fig. 3(a). In the single-pulse I-V measurement setup, we started to apply a relatively negative bias to the gate. During this stage, residual charges, which might be trapped in defects, are discharged. In the first stage, we ramp the gate voltage from 0 V to the transistor turn-on voltage during a rise time of 10 μs and simultaneously measure the drain current. In the second step, the transistor is turned on and some of the majority channel carriers might be trapped in the defects, causing a shift of the threshold voltage toward the positive gate bias direction and reducing the drain current level, because we applied the pulse for 2 ms and the drain current was gradually reduced during this time because of charge trapping. In the third phase of single-pulse measurement, we ramped down the gate voltage from the transistor turn-on voltage to 0 V during a fall time of 10 μs while measuring the drain current. For the three steps, the measured drain current versus gate voltage plots for four devices during rising and falling times (T R and T F , respectively) are presented in the left part of Fig. 3(b). During single-pulse measurement (composed of three phases), the transient drain current versus pulse time for four devices was measured; results are plotted in the right part of Fig. 3(b). When we performed singe-pulse measurement at various voltage rising times (i.e., 1 μs, 10 μs, 100 μs, 300 μs, and 500 μs), the transient drain current during the pulse time gradually becomes saturated after 0.5 ms, as seen in Fig. 3(c-f), following typical charge trapping behavior. The reduction of drain current during the pulse indicates charge trapping, and that transient charging behavior is significant for a single-layer HIZO TFT, but less so for a bilayer oxide TFT with relatively thin IZO, and insignificant for bilayer oxide TFT with relatively thick IZO, which is consistent with the result presented in Fig. 2. Transient charging behavior for four devices match the three-trap model with three trap time constants, as given by Eq. (1) 43 .
The above equation gives the number of events. The trapping time constants for the four oxide TFT devices are shown in Fig. 3(g). A short capture time constant (τ A ) and a long capture time constant (τ B and τ C ) were extracted from Fig. 3(c-f) by applying the above equation. This result reveals that with quantified constants, charge trapping phenomena during charge transport through the channel are significantly affected by the active structure 43,45 . Generally, carrier transport and charge trapping are thought to occur mainly between the channel and the insulator, or injected into the insulator from the channel [50][51][52][53][54][66][67][68] . However, in this study, we found that the effect of the IZO back channel is pronounced in TFTs with the same front channel (HIZO) and the same front interface (SiO 2 -HIZO). Because the back channel IZO layer has high mobility and high carrier concentration, this layer is attributed to the main charge transport layer in the HIZO-IZO stacked oxide TFTs. When charge transport occurs via an IZO layer with relatively low sub-gap states, charge trapping phenomena occur less often.
To extract the sub-gap states of the oxide TFT of four different channel structures, we exploited a DCA method modified by the charge pumping method [55][56][57][58][59] . Figure 4(a) shows the measurement diagram of the DCA method. The pulse generator (Agilent 81104 A) and the source meter (Keithley 2401) were controlled using the Lab-view program. We applied a periodic voltage pulse to the gate while the end nodes of source and drain were grounded, enabling us to modulate the carrier concentration. Figure 4(b) shows the gate pulse shape and corresponding discharge current versus the measurement frequency. When we ramp up the gate voltage, the interface and the bulk channel (with and without sub-gap states) are charged. Upon ramping down the gate voltage, the majority carriers in the channel are quickly discharged via source and drain. Thereafter, charges trapped in the interface and the channel with sub-gap states are slowly discharged. At a certain read time, we measured this discharge current at the drain using the source meter. Therefore, the tail part of the discharge current, δI, is related to the carriers discharged from the interface and the bulk channel with sub-gap states. The details of the measurement method and related physics are found in previous work [57][58][59] . As we measure the discharge current as a function of measurement frequency for the oxide TFT Fig. 4(b and c), the discharge current level varies with measurement frequency. At measurement frequency changes from 10 kHz to 1 MHz (from the 1 st to the 3 rd step), the discharge current linearly increases. Then, above 1 MHz, the value becomes saturated. For extracting the number of defects, we measured the discharge current of the device at a measurement frequency of 100-500 kHz. The number of charging sites is extracted from the slope, δI/δf, using Eq. (2), Defect sites 2 3 where N Defect sites is the density of defects (sub-gap states), k is the charge loss factor, and V is the active volume 57 . Additionally, we measure the measurement-frequency-dependent discharge current as a function of gate pulse amplitude (V G amp = V G − V FB ). As shown in Fig. 4(d-g), the discharge current and the slope, δI/δf, of the single-layer HIZO TFT is sharp, but is relatively lower for the bilayer oxide TFT with increasing IZO thickness. As shown in Fig. 4(h-j), from the slope, δI/δf, of the oxide TFTs, the density of the sub-gap states was extracted. The energy distribution of N Defect sites [N Defect sites (E)] is obtained by calculating the surface potential φ s (or E) as a function of V G using Eq. (3) 59 .  Table 1 summarizes the parameters of the sub-gap density profile from the model where N DA is the acceptor-like deep state density, kT DA is the acceptor-like deep state characteristic energy, N TA is the acceptor-like tail state density, and kT TA is the acceptor-like tail state characteristic energy [62][63][64][65]69 .
To further understand the transient charge trapping phenomena, the transient drain current versus pulse time plots for representative single-layer HIZO and bilayer HIZO-IZO TFTs were measured with different measurement temperature, as shown in Fig. 5. We used the model of G. Bersuker in which the existing sub-gap states or defects are filled via two processes, as shown in Fig. 5(a) 36,51,52 . As we apply a positive gate voltage, electrons drifting toward the front active channel are charged or trapped by acceptor-like sub-gap states. This step is called P C , which is a dominant process in the fast-transient charge trapping mechanism. Then, when electrons have sufficient thermal energy, the charged or trapped electrons are activated to overcome the barrier, then move via the trap (process P T ). On the other hand, P T is a secondary (and slow) transient electron trapping process based on electrons having sufficient thermal energy 36 . In this model, the equation for the kinetics of electron trapping can be expressed as follows.
Here, n is the number of the occupied defect-sites, N 0 is the total available trap number, p is the electron-trapping probability, and t is the time for the process. Figure 5(b and c) show the transient drain current of HIZO and HIZO-IZO layers versus measurement temperature. By fitting equation (4) to ΔI D during the short charging time (initial 50 μs), we analyzed fast electron trapping characteristics, which indicate that fast charging is not dependent on measurement temperature. For single-layer HIZO and bilayer HIZO-IZO TFTs, the extracted N 0 values are on the order of 8.4 × 10 13 and 2.2 × 10 12 cm −2 , respectively.
To probe the slow transient charge trapping phenomena, the transient drain current versus time was obtained during the long stress time (approximately 1000 ms), as depicted in Fig. 5(e,f), because slow transient charging follows a thermally activated process from trap to trap via activation energy. According to this model including the thermal activation process, the kinetics of the charging process can be expressed by Eq. (5).
where J is the current density, τ is the de-trapping time constant, E i is the trap energy, k is the Boltzmann constant, and T is temperature. Then, the slow transient charge trapping process with time is given by Eq. (6).  Table 1. Parameters of sub-gap density of states extracted using the discharging current analysis method.
HIZO-IZO), respectively, as shown in Fig. 5(i). This result indicates that the single-layer HIZO TFT has a larger number of defects and is more vulnerable to transient charge trapping, relative to the bilayer HIZO-IZO TFT.

Discussion
In summary, we evaluated the transient charge trapping effect, the charge trapping mechanism, and the dynamic charge transport of single-layer HIZO and high-mobility bilayer HIZO-IZO oxide TFTs.

Methods
Device Fabrication. We fabricated a bilayer oxide semiconductor TFT using an inverted staggered structure, as shown in Fig. 2(a). The integration of metal oxide semiconductor bilayer TFTs was done by sputter deposition a 100-nm thick Mo electrode at room temperature and subsequently growing a gate insulator of 100-nm thick SiO 2 by the plasma-enhanced chemical vapor deposition (PECVD). As an active channel, four different type of semiconductor layer configurations (such as HIZO (40 nm), HIZO (20 nm)-IZO (20 nm), HIZO (20 nm)-IZO (40 nm) and IZO (60 nm) channel) were employed. The films were deposited at room temperature by radio frequency (RF) magnetron sputtering method using separate 4inch diameter target of In 2 O 3 , ZnO, and HfO 2 . The RF power supplied to each of the targets was adjusted to control the composition of HIZO and IZO. The composition of the channel was analyzed by ICP-AES. The cation ratio was Hf/In/Zn = 0.07/0.53/0.4 for HIZO channel and In/Zn = 0.49/0.51 for the IZO channel. Subsequently, a 100-nm thick source/drain electrodes were sputtered at room temperature, followed by pattering process. Then, 200-nm thick PECVD SiO x passivation layer was formed and contact hole was patterned. After fabrication, the TFT devices were subjected to a post-annealing process in air for 2 h at 200 °C to cure any possible defects in the TFTs.
Device Characterization Device Characterization. The DC I-V, μs-FIV, and PIV transient current methods were applied using the Agilent waveform generator fast measurement unit (WGFMU) module of the B1500A semiconductor device analyzer. The WGFMU module creates pulses and arbitrary linear waveforms. These are applied to the device via the remote sensing unit (RSU) and current measurements are made using the RSU connected to the drain terminal. For the DC I-V measurements, the voltage sweep rate was 2 V/s, whereas for the μs-FIV and PIV measurements, the voltage scan rate was 1 V/µs. For DCA, square pulse bias in 100 kHz to 500 kHz range was applied to the gate electrode using a pulse generator (Agilent 81104 A). The pulse width was determined by the duty cycle of the pulse (50%) at a given frequency; the rising/falling time was fixed at 80 ns. The source is connected to ground within the source meter; the drain is connected to the parameter analyzer (Keithley 2401).