The existence of band-tail states (BTS) in indium-gallium-zinc oxide (IGZO) results in TFTs with electrical characteristics that are not well represented by conventional device models. Common parameters such as threshold voltage and channel mobility that are extracted from measured electrical characteristics can be misrepresentative due to discrepancies between the chosen operational model and the underlying device physics. There have been several reports of analytical solutions for the electrostatic operation of IGZO transistors that have been motivated by model accuracy and efficiency in circuit simulation. While developed compact models may be physics based, the addition of fitting parameters to exceedingly complex solutions results in a loss of physical connection and limits the applications to circuit design. Similarly fitting parameters can be added to conventional SPICE models used by device researchers; however the modifications will only have merit if the physical connection to device operation is maintained. This is especially important as channel lengths are scaled and device structures exhibit short channel effects (SCE).

A device model for the on-state operation of accumulation-mode IGZO TFTs which maintains consistency with the gradual channel approximation was recently developed and presented as an adaptation of a Level 2 SPICE model. The model accounts for the ionization and deionization of acceptor-like BTS, as controlled by both the gate and drain bias conditions. Long-channel TFT I-V characteristics are well represented by the physical channel length and width along with five operational parameters shown in equations (1-4). The model fit for a representative long-channel device is virtually indistinguishable from measured data, as shown in Fig. 1. The threshold voltage (V_T) and field-independent channel mobility (μ_o) have traditional meaning, whereas the introduced BTS parameters (θ_BTS, V_BTS, α) regulate the level of free channel charge and account for spreading of the output conductance. Model parameters were extracted using regression analysis on output characteristics with fine gate voltage increments. Parameter values for the representative device are shown in Table 1, with the resulting threshold voltage and channel mobility values consistent with TCAD analogs. The η_G and η_D model elements preserve the distinction between the gate-impressed and drain-impressed response, respectively, with dissociation from an effective channel mobility. This distinction is important to avoid confounding with SCE as devices are scaled and E-fields are increased.

This work extends the device model to represent long-channel and scaled devices (i.e. L ≤ 3 μm) with bottom-gate (BG) and double-gate (DG) electrode configurations. BG and DG TFTs have different levels of gate control and influence over BTS, which becomes more pronounced as the channel length is decreased. For scaled devices, channel length modulation as well as conventional field-effect mobility parameters that account for normal-field degradation and velocity saturation have been added. TCAD simulation is used to discriminate between the influence of BTS and SCE. The physical correlation of the model to device operation is demonstrated through comparisons with measured characteristics and TCAD simulation.