Origin

of Threshold Voltage Instabilities in Indium Oxide Transistors

Authors: Tzu-Jie Lin, Sheng-Chung Chen, Yung-Ting Lee, Sheng-Lun Cheng, Robert Tseng, Sung-Tsun Wang, Yu-Cheng Chang, Yi-Yu Pan, Chan-Yuen Chang, Tsung-Te Chou, Chia-Hsien Lin, Ching-Shun Ku, Chun-Liang Lin, Po-Tsun Liu, Hyungjin Kim, Der-Hsien Lien

Abstract

Oxide semiconductors have gained substantial interest for their low-temperature processability, allowing for their integration as functional add-on device layers for advanced monolithic 3D integrated circuits (ICs). However, reliability issues, particularly under thermal, environmental, and electrical stresses, remain critical issues and require immediate solutions. This study investigates the instability of ultrathin In2O3 transistors, revealing that threshold voltage (V T) drifts arise from interactions between surface-adsorbed oxygen and the In2O3 channels. We show that the oxygen in the ambient atmosphere attached to the In2O3 surface plays a crucial role in modulating In2O3 conductivity, thereby governing V T. External perturbations such as ultraviolet (UV)/X-ray illumination, thermal annealing, and bias stress could alter this interaction of surface oxygen with ultrathin In2O3, leading to a V T drift. Importantly, we propose a unified kinetic model that provides a generic physical description of V T instabilities induced by these commonly observed factors. By characterizing time-dependent V T instability, the model demonstrates that recovery dynamics exhibit identical behavior across all tested perturbations, indicating that the recovery process is independent of the initial stimulus. This study uncovers the surface oxygen as a critical factor affecting In2O3 transistor reliability, offering insights for designing oxide-based devices for advanced electronic and optoelectronic devices.

Introduction

Atomically thin In2O3 has emerged as a promising material for advanced electronics due to its high mobility, ease of synthesis, and compatibility with monolithic 3D integration, making it well-suited for next-generation semiconductor technologies. −

Its tolerance to lattice disorder arises from the isotropic transport of s-orbital electrons, which maintain overlap even without high crystallinity. , This unique feature enables In2O3 to achieve high mobility up to 100 cm^2^ V^–1^ s^–1^ in the amorphous phase, , thereby circumventing crystallinity constraints and allowing for low-temperature synthesis. This intrinsic advantage meets the stringent thermal budget requirements of back-end semiconductor processes, making In2O3 highly compatible with heterogeneous integration into CMOS technologies. −

Ultrathin In2O3 transistors also exhibit notable tunability in their electronic properties, with the threshold voltage (V T) largely tunable by environmental and processing conditions. For example, recent studies have shown that the electronic characteristics of 2-nm-thick In2O3 can be substantially modulated through ultraviolet (UV) exposure, enabling a wide range of carrier concentrations that directly modulate V T. , Similarly, annealing in nitrogen or oxygen environments can effectively shift the carrier density and electronic behavior of ultrathin films. , However, this sensitivity also raises critical reliability concerns, leading to undesired V T during the fabrication process; −

for example, pronounced V T shifts are observed when dielectric layers are deposited on top of the transistors. , It is also reflected in considerable reliability issues under electrical stress; for example, significant V T drifts are observed under positive and negative bias stresses (PBS and NBS). −

Those instabilities are often attributed to oxygen-related trap states in the amorphous structure, −

while others suggest surface-adsorbed molecules are responsible for the conductivity changes, , leaving the underlying mechanisms unresolved.

In this work, we investigate the mechanism of V T instability in ultrathin In2O3 transistors. By excluding the effects of other air molecules, our results identify surface oxygen as the origin of the V T drift, acting as an effective electron trap that modulates the conductivity of n-type In2O3. We show that external perturbations, such as high-energy illumination (UV and X-ray), thermal annealing, and gate bias stresses, induce V T shifts by modulating the equilibrium and interfacial interactions between surface oxygen and the In2O3 channel. A kinetic model is proposed to quantify the V T shifts and recovery dynamics. This study elucidates the surface phenomena in In2O3 transistors for further improving the device stability in next-generation electronic and optoelectronic applications.

Results and Discussion

We first examine the V T shifts of ultrathin In2O3 transistors under various external perturbations, including UV and X-ray exposure, high-temperature annealing, and electrical bias stress, followed by analysis of the subsequent V T drift after these stimuli are removed. 2 nm In2O3 films were deposited on the Si/SiO2 substrate using atomic layer deposition (ALD), with the height confirmed by high-resolution transmission electron microscopy (HRTEM) and atomic force microscopy (AFM), as shown in Figure a. The optical microscope images of as-deposited In2O3 are also shown in Figure a. The I D–V G transfer characteristics of In2O3 transistors exhibits on/off ratio of over 8 orders with V D = 1 V, as illustrated in Figure b. Enhancement-mode operation is achieved in In2O3 transistors with 2 nm thickness, highlighting a key advantage of ultrathin films, whose conduction can be effectively modulated by the gate. The as-made device is operated in enhancement mode and could vary between the enhancement and depletion modes depending on the perturbation conditions. The I D–V D output curves in Figure c exhibit clear saturation, with a maximum drain current of 125 μA/μm at V D = 15 V.

UV exposure can induce conductivity changes in oxide semiconductors and has previously been shown to cause V T shifts in In2O3 transistors. −

Figure a shows the transfer characteristics (I D–V G) of 2 nm In2O3 transistors exposed to UV irradiation (365 nm) from 1 to 600 s. The V T of the transistors is extracted from the I D–V G, as shown in Figure b (extraction of V T is described in Text S1). The In2O3 transistor shows a negative V T shift upon UV illumination. As the UV exposure time extends, V T shifts eventually reach a maximum ΔV T of −10 V (V T shift as a function of UV power density and exposure time in Figure S1). After the UV is turned off, the V T gradually returns to its original value after days. Similar behavior has been observed in ZnO and SnO2, where UV exposure increases conductivity as photogenerated holes neutralize physiosorbed oxygen on the oxide surface, , leading to its desorption and restoring free electrons to the channel. Upon cutting off the source, surface oxygen withdraws electrons from the n-type channel, reducing conductivity.

To further verify the light-matter interaction mechanism, In2O3 was illuminated by an X-ray source (Mg Kα; 1253.6 eV). With an increasing X-ray exposure time, 2 nm In2O3 transistors exhibit a negative V T shift, as shown in Figure c,d. Resembling the trend observed under UV exposure, a maximum negative V T shift (ΔV T = −29 V) was observed for X-ray irradiation after an exposure time of over 600 s (V T shift as a function of X-ray energy in Figure S2 and thickness dependence of In2O3 absorption spectra in Figure S3). Similar to the UV illumination, turning off the X-ray sources leads to a gradual recovery of V T to its original value over several days (Figure b,d). Notably, V T recovery occurs only when the devices are stored under ambient conditions, which will be discussed in detail later.

Under UV or X-ray exposure, the recovery of conductivity over several days is significantly longer than the minority carrier lifetime, suggesting that the effect is not governed by conventional recombination dynamics. Similar long-term light-induced conductivity changes have been previously observed in various metal-oxide semiconductors, including ZnO, TiO2, IGZO, , and SnO2 and are attributed to surface interactions involving the adsorption and desorption of ambient air molecules, particularly oxygen. These surface-related effects are further amplified in reduced-dimensional structures, such as nanowires, −

and nanorods, owing to their large surface-to-volume ratios. ,, Among nanostructures, ultrathin films, especially at the quasi-2D limit, are most sensitive to surface effects due to their high surface-to-volume ratios. Despite this, surface effects in ultrathin In2O3 channels have been insufficiently explored.

To further examine the role of gas molecules in governing the transport characteristics of In2O3 transistors, annealing was conducted in both oxygen-rich and oxygen-scarce environments to regulate the total amount of oxygen on the In2O3 surface, as shown in Figure . To create oxygen-scarce environments, In2O3 transistors were annealed at 150 °C in nitrogen. The I D–V G in Figure a shows that the curves shift negatively with increasing annealing time in a nitrogen environment, with a maximum negative ΔV T of −22 V. The V T recovers to the original value when the device is cooled to room temperature in normal ambient, as illustrated in Figure b. The result indicates that annealing in a N2 environment reduces surface oxygen concentration, suppressing electron trapping effects and consequently increasing conductivity, leading to a negative V T shift. In contrast, oxygen-rich annealing produces opposite results. When annealed in oxygen at 150 °C, In2O3 transistors exhibit a positive V T shift with an increasing annealing time, as shown in Figure c. The I D–V G shows a maximum positive ΔV T of 1 V, along with the subsequent recovery of V T after cooled down in the ambient, as illustrated in Figure d. In this case, overall surface oxygen concentration is increased, leading to a higher charged surface oxygen concentration and a positive V T shift.

To clarify the role of electrons and holes in In2O3, gate bias was applied to modulate the carrier density in In2O3. , When a negative gate bias (V G = −15 V) was applied, V T shows a negative shift with a maximum negative ΔV T of −4.8 V, as shown in Figure a,b. Since In2O3 is an n-type semiconductor, a negative gate bias increases the minority carrier density (holes). Although hole-branch current is not observed, likely due to low mobility or high contact resistance for holes, the increased hole density under negative gate bias facilitates charged surface oxygen neutralization. This process leads to a negative V T shift, similar to the effect of UV/X-ray exposure and oxygen-scarce annealing. Conversely, applying a positive gate bias of V G = 15 V resulted in a positive V T shift, as shown in Figure c,d. This positive V T shift is due to a reduction in hole density caused by the gate bias, which promotes electron transfer from In2O3 to surface oxygen and lowers electrical conductivity. Since holes are minority carriers in n-type In2O3, this reduction has a limited effect on the balance of the surface oxygen, resulting in a slight positive V T shift.

While oxygen is widely recognized as the primary ambient species affecting the conductivity of oxide materials, the effects of other air molecules remain unclear. To investigate their impact on ultrathin In2O3, we examine the V T drift by exposing a degenerated In2O3 channel to N2, O2, H2O, three major constituents of ambient air. Among these, only oxygen exposure leads to V T recovery, while no significant change is observed under nitrogen or water vapor, as shown in Figure a,b, confirming the dominant role of oxygen. Density functional theory (DFT) simulations also support the strong charge exchange between surface oxygen and In2O3 (Figures S4–S6). The results show that with the adsorption of oxygen on the In2O3 surface (either in the form of mono oxygen or oxygen molecules), the Fermi level drops to the bottom of the bandgap, indicating a transfer of electrons from the In2O3 to oxygen. −

Here, a kinetic model is developed, where the whole charge exchange kinetics associated with the surface oxygen is governed by three (i) transport of oxygen molecules to the In2O3 surface, (ii) adsorption and desorption of oxygen at the In2O3 surface, and (iii) charge exchange between oxygen and In2O3. The diffusion rate of oxygen from the atmosphere to the surface of In2O3 is determined by Henry’s law F = h g(C g – C s), where h g is the gas phase mass-transfer coefficient, C g is the oxygen concentration of air, and C s is the oxygen concentration on the surface. In ambient conditions, transport of oxygen from the atmosphere to the In2O3 surface occurs within milliseconds (Text S2). While oxygen arrives at the surface, the physiosorbed rate is determined by R = FS, where F is the flux of oxygen and S is the sticking coefficient (fraction of gas molecules that adsorb upon hitting the surface). In ambient conditions, oxygen adsorption and desorption reach an equilibrium within seconds (Text S3). Once the adsorption and desorption dynamics of oxygen reach equilibrium, a stable charge exchange pathway is established between the oxygen species and the In2O3 channel. The total oxygen concentration is defined as [Osur ^–^]tot = [Osur] + [Osur ^–^], where [Osur] represents the concentration of neutral oxygen species, [Osur ^–^] refers to the concentration of oxygen that withdraws electrons from the In2O3.

As the observed V T shift dynamics are much slower than the first two processes, the kinetics is predominant by the charge exchange that occurs between the oxygen and In2O3, which can be described by the following first-order reversible reactions1Osur−+h+k1⇌k2Osur 2Osur+e−k3⇌k4Osur−where the constant k 1 and k 2 are the forward and reverse rate constants of eq , respectively, while k 3 and k 4 are the forward and reverse rate constants of eq , respectively. For eq , electrons exchange from oxygen to In2O3, and the oxygen restores the neutral state. For eq , electrons transfer to the oxygen from the In2O3 to oxygen, so the oxygen is charged. According to the two equations, we obtain the general solution for [Osur ^–^] as a function of time (details of derivation in Text S4)­3Osur−=[Osur−]0e−(k1δp+k2+k3δn+k4)t+(k2+k3δn)[Osur]totk1δp+k2+k3δn+k4(1−e−(k1δp+k2+k3δn+k4)t)where δn and δp are excess electron and hole densities in the nonequilibrium state, which is determined by the external perturbations. Note that all perturbations operate under high-level injection conditions, where carrier density is governed by δn and δp, as they surpass the equilibrium electron (n 0) and hole (p 0) densities. [Osur ^–^]0 is the initial surface-charged oxygen concentration under thermal equilibrium in the ambient environment. Equation shows that [Osur ^–^]­(t) decays exponentially over time when the first term dominates, as observed under UV illumination, X-ray illumination, N2 annealing, and negative bias stressing, as illustrated in Figure c,d (Fourier-transform infrared (FTIR) spectra of surface-adsorbed oxygen on In2O3 films in Figure S7). The equation shows that [Osur ^–^] increases with an exponential saturation trend when the second term dominates, as seen in the cases of O2 annealing and positive bias stressing. We then use the kinetic model to examine the dynamics of the V T shift under various perturbations shown in this study. Each stimulus drives δn and δp toward a new equilibrium through the respective mechanisms described above. The altered carrier densities in the nonequilibrium state shift the [Osur ^–^] concentration toward a new equilibrium, with the rebalancing time dictated by both δn and δp, and the reaction rate constants. As oxygen extracts an electron from In2O3, it effectively acts as an acceptor dopant. Consequently, changes in [Osur ^–^] modulate the carrier concentration of In2O3. According to the Drude model, variations in the carrier concentration induced by [Osur ^–^] directly correspond to a linear shift in V T. By fitting the V T-time curves using eq , the time constants required to achieve the rebalanced [Osur ^–^] under different stimuli can be obtained, as shown in Figure a–d (The parameters associated with each specific perturbation are summarized in Figure S8 and Table S1. Post-perturbation mobility change is also shown in Figure S9). ,

The same kinetic model can also be applied to the recovery dynamics. The stimuli rebalance the [Osur ^–^], which reaches the saturation value [Osur ^–^]sat 4[Osur−]sat=(k2+k3δn)k1δp+k2+k3δn+k4[Osur]tot

Upon removal of the perturbations, [Osur ^–^] begins to restore the initial value from [Osur ^–^]sat. Meanwhile, with the removal of the perturbations, electron and hole densities return to their thermal equilibrium values n 0 and p 0, respectively. The variance of [Osur ^–^] can be obtained from the same kinetic model5Osur−=[Osur−]sate−(k1p0+k2+k3n0+k4)t+(k2+k3n0)[Osur]totk1p0+k2+k3n0+k4(1−e−(k1p0+k2+k3n0+k4)t)

The equation indicates that when the recovery time is sufficiently long, [Osur ^–^]­(t) reverts to its initial value [Osur ^–^]0, as all cases eventually restore their initial equilibrium under identical ambient conditions (Text S5). Therefore, when overlaying the recovery curves, they closely align, as shown in Figure e (The parameters associated with each recovery are summarized in Figure S8 and Table S1. For simplicity, the parameters are lumped into coefficients α, β, τ1, and τ2 in eqs and .). Additionally, fitting these curves with eq yields similar recovery time constants (∼2 days) across different cases, confirming a consistent recovery mechanism. Notably, the intrinsic minority carrier lifetime of In2O3 typically ranges from tens to hundreds of nanoseconds and is expected to be even shorter in amorphous films due to higher trap densities. −

In contrast, the observed conductivity recovery in this study occurs over days, far exceeding the minority carrier lifetime. This difference arises because the extended perturbation time provides a sufficient nonequilibrium window to activate surface oxygen adsorption/desorption kinetics. Furthermore, the long recovery time could arise from two possible charge exchange between physiosorbed oxygen and the In2O3 channel, or diffusion of surface oxygen into the bulk, accompanied by vacancy filling, as illustrated in Figure c (Thickness dependence of In2O3 devices postperturbation in Figures S10–S12). Both mechanisms could explain the observed recovery behavior; however, the exact mechanism of the underlying charge exchange process remains to be clarified.

To further examine the pressure effect, the transistors were placed in a vacuum environment following external perturbations, as shown in Figure f. Different from previous results in the ambient, the negative V T shift caused by the UV exposure and N2 annealing treatment shows no recovery phenomenon when the device is stored in a vacuum. The V T shift remained after 14 days. The results indicate that the neutralization of oxygen [Osur ^–^ → Osur] could cause a desorption process of surface oxygen. ,,,,, When the In2O3 surface is exposed to a vacuum, in the absence of an additional oxygen supply, the charge exchange process described in eq is inhibited. Consequently, this leads to a significant reduction in [Osur ^–^]tot, weakening the contribution of the second term of the exponential in eq , which governs the recovery mechanism. This controllable recovery phenomenon offers a unique opportunity to emulate synaptic plasticity and multi-scale temporal information processing in neuromorphic systems, as explored in recent studies. ,

Conclusions

In summary, this research elucidates the underlying mechanism of electrical instabilities of ultrathin In2O3 devices caused by external perturbations. We show that the modulation of V T is primarily attributed to the charge exchange between surface oxygen and In2O3. A key contribution of this study is the development of a unified kinetic model that captures the V T recovery dynamics following illumination, annealing, and electrical perturbations within a generic model. The V T recovery at room temperature indicates that the charge exchange of surface oxygen, along with the adsorption and desorption processes, is the primary reason influencing carrier concentration, rather than the formation of oxygen vacancies or defects. The results indicate that adding a caping layer on In2O3 could effectively prohibit the exchange of surface oxygen, thus improving the reliability, explaining the result reported previously. , This work also supports that surface functionalization could potentially passivate the surface and suppress surface interaction, thereby improving device stability. This passivation mechanism is expected to be universally applicable across various oxide semiconductors, offering critical insights into developing advanced strategies to improve the reliability of oxide-based electronics and optoelectronics.

Experimental Section

In2O3 Growth and Transistor Fabrication

2 nm-thick In2O3 thin films were deposited on SiO2/p^+2^Si substrates by ALD, forming a basic metal–semiconductor-oxide (MOS) structure. The 30 nm-thick SiO2 layer serves as the dielectric, with heavily phosphorus-doped Si (p^+2^Si) as the back gate, and the ultrathin In2O3 layer exposed in air. The active area was then defined by photolithography, followed by etching of In2O3 using a diluted solution of HCl. 30 nm-thick nickel was deposited on In2O3 to form the source-drain metal electrodes. The thickness of the as-deposited In2O3 was measured using an ellipsometer, AFM, and HRTEM.

Transistor

Characterizations

The In2O3 devices were exposed to UV and X-ray illumination. The 365 nm UV light source is LED, and the X-ray light source is 1253.6 eV with Mg Kα, and the other one is 20–80 keV with a Teresa nano-CT system (Excillum NanoTube N3). The devices were also exposed for 600 s. The In2O3 devices were placed in a customized chamber with two gas inlets. Then, the devices were annealed in O2 and N2 with a 1 L/min gas flow at 150 °C. The pressure was kept at about 1 atm. The negative and positive gate bias were generated by Keysight Agilent B2902B source at room temperature. The devices were stressed by ±15 V for 1000 s under atmospheric conditions. The electrical characteristics were measured using Keysight Agilent B2902B source at room temperature, in the absence of light, under atmospheric conditions.

DFT Calculation

First-principles calculations of In2O3 materials were performed using OpenMX (v3.9.9) within the framework of density functional theory (DFT). The computations utilized the generalized gradient approximation (GGA), norm-conserving pseudopotentials, and optimized pseudoatomic basis functions. −

The exchange-correlation interactions were treated using the Perdew–Burke–Ernzerhof (PBE) functional within GGA. The optimized radial functions used were In-s3p2d2 for indium and O-s2p2d1 for oxygen, respectively. The criterion for force convergence was set to 1 × 10^–4^ hartree/bohr, and the electronic self-consistent field convergence criterion was set to 1 × 10^–9^ hartree. The energy cutoff was 400 Ry for all DFT calculations. The DFT model comprises a corundum-type In2O3 layer, representing an amorphous In2O3 layer. The optimized In2O3 film with an adsorbed oxygen atom [Osur] consisted of 31 atoms, including 12 indium and 19 oxygen atoms. The k-point grid was 6 × 6 × 1 for the In2O3 film.

The In2O3 Annealing Environment

The oxygen-rich environment refers to pure ambient (99.999% purity) O2 at 1 atm. The oxygen-scarce environment refers to a high-purity N2 ambient (99.999% purity), where the oxygen partial pressure is minimized. Quantitatively, the oxygen level in the N2 environment is estimated to be less than 10 ppm, which is several orders of magnitude lower than that in the O2 environment.

Supplementary Material