2.1 Structural analysis by TEM

To figure out the possible causes, the first attempt was to check whether there were any abnormal things at the TFT structures or not. It was formed well of each part as BG electrode (Ti/TiN), Gate dielectric material (Al₂O₃ or SiO₂), IGZO and Source/drain electrode (Mo). In Fig. 5, TEM cross section of each experimental condition is shown. There was no irregular shape except for a dioxide crack at the sample (d), which could bring a direct electric short between BG and source node of TFT. However, the Vg-Id curve of the sample (d) did not show any direct shorts at each electric node.

Fig. 5. TEM cross section of each split sample (a) N2_anneal at Al2O3, (b) No_N2_anneal at Al2O3, (c) No_ N2_anneal at PE-Oxide, (d) N2_anneal at PE-Oxide.

2.2 Oxygen binding energy by XPS

To analyze oxygen vacancy change depending on N₂ anneal process, XPS measurement was fulfilled at the Al₂O₃ samples as shown in Fig. 6 and Table 2. The O1s peaks are classified as the low binding energy (LP), middle binding energy (MP), and high binding energy (HP) according to references ^{[3- 11]}. The LP is interpreted as an indicator of oxygen atom amounts, O^2- surrounded by In/Ga/Zn atoms ^{[4- 6]}. The MP is related with oxygen vacancies, O^2- in oxygen deficient ^{[7, 8]}. The HP is belonged to chemisorbed, dissociated oxygen or OH^- groups like as -CO₃, adsorbed H₂O or adsorbed O₂ ^{[9- 11]}.

At IGZO internal free electron generation mechanism, the metal cation (M) and the ionized oxygen vacancy (V''_o) behaves the following expression:

A finding is MP area decreasing from 8.3% to 5.9% by N₂_anneal process, which meant to the reducing of the amount of V''_o according to the expression (1). This is the reason why Ioff (< 1 nA) of N₂_anneal condition at Id-Vg curve was 100 times lower than that (< 100 nA) of No_N₂_anneal. Those vacancies normally act as shallow donor, supplying free electrons that make the channel easier to turn on. With fewer donors, we need a large positive bias (+0.4 V) to induce the same electron density in the channel, as can be seen in Fig. 4(a). Free carrier concentration changes the flat band and threshold voltage (Vth). If Vth follows a logarithm of donor density, $\Delta$Vth can be expressed by the following expression as a first approximate estimate:

It is well matched with the experimentally measured value. For many MOS-like models, a change in doping that changes channel charge produces a logarithmic shift in Vth, which is connected to off current that is dominated by subthreshold conduction or exponential dependence on gate-voltage relative to Vth. Therefore, Ioff can be represented by the following expression:

Finally combining (3) and (2) gives the following:

where SS is the subthreshold swing, Vo being the vacancy concentration.

In addition, from the fact of MP area reduction, we could interpret the IGZO layer changed to more semiconductor like state from semi-metal state of No_N₂_anneal condition.

Fig. 6. The normalized intensity vs. Oxygen binding energy by XPS, (a) No_N2_anneal, and (b) N2_anneal of the Al2O3 samples.

2.3 OH change observation by m-SIMS

We measured two different areas of TFT as channel and source/drain metal region by mass(m)-SIMS. No change found at channel region. However, at source/drain metal region, it is observed that the OH is decreasing at the underneath of source/drain metal (Mo) by N₂ anneal condition as shown in Figs. 7 and 8.

At the interface between source/drain and metal (Mo) or on the surface of IGZO, OH and electron generation mechanism is expressed:

Fig. 7. SIMS result at source/drain metal region.

Fig. 8. Enlargement of OH profile at source/drain metal (Mo).

Due to source/drain metal (Mo) thickness was 100nm, the OH amount and depth of SIMS data are exactly matched like 475 c/s (102 nm) of the No_N₂_anneal and 104 c/s (104 nm) of the N₂_anneal at the underneath of Mo metal.

It could be interpreted as SS of TFT Id-Vg curve's improvement caused by this OH decreasing of N₂_anneal condition. As early described in this paper, the SS of the N₂_anneal was much steeper than that of the No_N₂_anneal. The SIMS profiles of OH^- give density of interface trap charge (Nit, # of trap charges/cm²) with the help of standard sample by integrating the peak area, transforming the areal density into the interface trap density (Dit, # of trap charges/cm²/eV), resulting in interface traps (Cit, equal to q x Dit, q being the charge quantity). Electronically the reduction in Cit is directly correlated with the improvement in SS with the following expression:

If we assume that uniform bulk is depletion-limited in the channel and the Debye approximation is valid due to the fact that depletion width is smaller than film thickness (100 nm) of IGZO, then Cdep $\approx$ $\varepsilon_s/\lambda_D$, $\lambda_D = (\varepsilon_s\text{kT/q}^2\text{n})^{0.5}$ where Cdep is the capacitance of depleted channel, $\varepsilon_s$ is IGZO permittivity, $\lambda_D$ is Debye length, n being the free carrier density or concentration. Finaly SS can be expressed by the following again:

For the ultrathin IGZO TFTs where the entire film can fully deplete, the simple Debye capacitance gives a correct dependence $\sqrt{\text{n}}$ and the main contribution to SS comes from the Cit. The mechanism that connects OH^- content to SS behavior is explained by 'subthreshold conduction' as elaborated in the previous description of (2), (3), (4) for the correlation between the reduced MP and the improvement in the improvement in Ioff.

2.4 IGZO composition ratio by TOF-SIMS

Compared to m-SIMS that could be measured only by the bulk samples, Time of flight (TOF)-SIMS was useful for the real patterned TFT devices. Moreover, it was possible to make 3D imaging of the TFT structure.

Related with IGZO composition ratio as In, Ga, and Zn elements, no difference is found between each split condition by m-SIMS and TOF-SIMS.

Fig. 9. Top view by TOF-SIMS at (a) No_N2_anneal at Al2O3, (b) N2_anneal at Al2O3, (c) estimated areas of each nodes.

Fig. 10. 3D imaging by TOF-SIMS at (a) XY view, and (b) each element such as Si, Ti, Al, In, and Mo.

2.5 IGZO roughness by AFM

Finally, we checked the IGZO roughness by AFM (Atomic Force Microscopy). The measured value is 0.468 nm (rms) at the N₂_anneal process and 0.489 nm (rms) at the No_N₂_anneal. According to these data, we concluded that the IGZO surface roughness was not changed and no influence for the electrical results.

Fig. 11. AFM data at (a) No_N2_anneal, and (b) N2_anneal .