Engineering Amorphous IGZO Thin-Film Transistors: The Role of Composition and Channel Thickness in Mobility-Threshold Voltage Optimization

Abstract

Recently, it has been shown that the mobility of amorphous InGaZnO (a-IGZO) thin-film transistors (TFTs) depends strongly on channel thickness and metal composition (In/Ga/Zn), resulting in a mobility-threshold voltage (V _th) trade-off. To the best of our knowledge, this work provides the first comprehensive modeling study systematically integrating density functional theory (DFT) and machine learning potential (MLP) to capture structural disorder and thickness effects on mobility-V _th in amorphous IGZO. We establish the existence of a universal mobility-V _th trade-off across diverse IGZO compositions and channel thicknesses. To unveil the origin of the universal trend, we developed a mobility model that covers the full composition and thickness of a-IGZO with composition-resolved parameters using DFT and MLP, taking into account its stochastically and structurally driven variation of the amorphous material's properties. We found that the origin of the mobility-V _th trade-off is the strong dependence of both mobility and V _th on the carrier concentration. Despite the existence of the mobility-V _th trade-off, as a method for designing enhancement-mode devices with high mobility, we propose increasing the Zn content to reduce the structural disorder.

1

Schematic framework for mobility and V _th modeling, integrating atomistic simulations and the BTE-based scattering model across various a-IGZO compositions. (a–b) Iterative training process of MLP to span the entire composition space of a-IGZO. (c) Structure generation via MLP-interfaced MD for each composition. (d) Validation of generated structures: comparison of metal-oxygen radial distribution functions (RDF) obtained from DFT and MLP (left), and the inverse participation ratio of the generated 1:1:1 a-IGZO structure (right). RDFs for other compositions are presented in Figure S2. (e) Calculation of electronic parameters (highlighted in red) from DFT simulation, used for the mobility model. (f) Construction of a BTE mobility model incorporating quantum confinement effects for the V _th model.

2

A ternary map of (a) effective mass, (b) ρ_SD, (c) band gap, and (d) carrier concentration of a-IGZOs. The data presented on the ternary maps represent averages over ten different a-IGZO for each composition. The carrier concentration is calculated based on the band gap values, as detailed in the main text.

3

Three factors (E _F, ρ _SD, and m*) affecting mobility and their extent for various compositions. (a) The mobility value of compositions, contributed by composition-dependent factors, and (b) the corresponding mobility changes. E _F is the most dominant factor, but ρ_SD also has an impact on compositions containing Zn. The effect of the dielectric constant is negligible, which is not shown here.

4

Ternary maps of predicted mobility incorporating composition-dependent m _e and E _F values: (a) without ρ_SD dependence and (b) ρ_SD dependence included. (c) Comparison of the mobility values of five different compositions, In/Ga/Zn = 1:1:1, 1:1:6, 2:0:1, 4.6:1:1, and 1:0:0, between experimental measurements and model predictions. The composition of maximum mobility changes from In-rich to In₂Zn₁O₄ with 41.4 cm²/V·s.

5

(a) An illustration of the mobility change caused by the reduction of thickness and carrier concentration that makes the barrier relatively high. (b) Owing to the relation of mobility and E _F, a mobility and V _th trade-off arises, marked by the black dashed line. In the absence of any compositional change, the mobility values consistently lie on the curve, even with varying thickness and carrier concentration (or E _F). (c) For different values of ρ_SD, the curves split: an orange-colored line indicates Zn-rich composition with the lowest ρ_SD, while the purple corresponds to the In-rich composition with the highest ρ_SD. (d) Comparison between the experimental mobility-V _th values of TFTs ,,− and the predicted values.

6

(a) The analytical model is derived in the nondegenerate region and the degenerate region using approximations. This simple equation clarifies the mobility dependence on E _F, where η_F = (E _F – E _C) / (k _B T) and ${\mathcal{F}}_{\mathrm{i}}$ is the Fermi Dirac integral. The analytical model agrees well with the numerical model in the energy range where electrons can exist at room temperature. The analytical model also exhibits temperature dependency as well as E _F dependency. The coefficients M and α are presented in Supporting Information S9. (b) The mobility values, calculated by the analytical model, are plotted to compare with experimental data, at ρ_SD = 1.59 × 10²⁶ m^–3. The curve is matched well with the AOS TFT experiment data. The legend is the same as in Figure d.