Introduction

Transparent conductive oxide (TCO) is an indispensable material in optoelectronic devices, such as flat panel displays and photovoltaics. Currently, indium tin oxide (ITO) is widely used as TCO because of its excellent conductivity and transparency in the visible region. However, the shortage of indium, a major component of ITO, requires the development of TCO alternatives composed of more abundant elements. Recently, it has been reported that Nb-doped anatase TiO2 (Ti1-xNbxO2; TNO) thin films in both epitaxial [1] and polycrystalline [2] form exhibit low resistivities (ρ) of the order of 10-4 Ωcm and high transmittances of 60-90% in the visible region, demonstrating the potential of TNO as a next-generation TCO.

There are two main methods for fabricating polycrystalline TNO films: direct deposition of crystallized films on heated substrates [3] and crystallization of amorphous films deposited on unheated substrates by post-deposition annealing [4]. Since it is much easier to obtain TNO films with ρ < 1 × 10-3 Ωcm by the latter method, the crystallization process of amorphous TNO films, including nucleation and crystal growth from the nuclei, has attracted great interest. However, even the basic parameters of crystallization kinetics of amorphous TiO2 have scarcely been examined thus far. In addition, experimental factors that determine grain sizes after crystallization are still unclear.

In this study, the crystallization kinetics of sputter-deposited amorphous TNO films during isothermal annealing was studied by in situ X-ray diffraction (XRD). The crystallization of TNO films was found to have two-dimensional saturation growth. This implies that the density of nuclei, which limits the grain size after crystallization, is constant irrespective of annealing condition. To further explore the factors determining the density of nuclei, the microstructures of the amorphous films was investigated. The results show that the microstructures of amorphous films followed Thornton model, in which working pressure has a great influence on microstructure through shadowing effect. The grain size was strongly affected by the working pressure of sputtering, suggesting that voids formed by shadowing effect behave as the nucleation sites.

Experimental

Amorphous TNO films were deposited on alkaline-free glass substrates and thermally oxidized SiO2-coated Si substrates without intentional substrate heating by RF magnetron sputtering technique. A 2-inch-diameter ceramic disk with Ti0.963Nb0.037O2-δ composition was used as a target. For in situ XRD measurements, TNO films were deposited at a working pressure P =1 Pa in a mixture of Ar and O2. For structural studies, TNO films were deposited in a mixture of Ar and H2 at the P range of 0.2 to 2 Pa. The as-deposited amorphous films were crystallized by H2 annealing at 600 oC for 1 h. The crystallization kinetics of the amorphous films during isothermal annealing was monitored by an in situ X-ray diffractometer equipped with a heating unit. The in situ XRD measurements were continued until the diffraction intensity of anatase (101) no longer changed. Ex situ polarized-light optical microscopic (POM) observations were carried out at room temperature to confirm the completion of crystallization and to characterize grain sizes. The surface and cross sectional topography was examined by scanning electron microscope (SEM).

Results and discussions

Crystallization kinetics of amorphous thin films

Fig. 1 shows the evolution of the anatase (101) diffraction peak for the TNO film deposited at f(O2) = [O2/ (Ar+O2)] = 1% (Ta = 300 °C). The amorphous TNO film appears to gradually crystallize into a polycrystalline anatase phase. The volume fraction of the crystallized anatase phase at each annealing time t was evaluated as x(t) = I(t)/I(∞), where I(t) and I(∞) are the integrated intensities of the (101) diffraction peak at time t and at the end of the measurement, respectively. The obtained x vs t curve is shown in Fig. 2(a).

According to the JMA formula [5], the crystallization kinetics is given by

x(t) = 1 - exp(-ktn), (1)

where k is a time constant depending on both the nucleation rate and growth rate, n is the Avrami exponent, which reflects the dimensionality of crystal growth and the nucleation rate relative to the growth rate. For isothermal crystallization, the Avrami plot of ln(-ln(1-x)) vs ln(t) yields a straight line with a slope of n. Fig. 2(b) shows ln(-ln(1-x)) vs ln(t) curves for the three TNO films. From Fig. 2(b), The value of n equal to 2.7, 2.0, and 2.2 were obtained for the TNO films deposited at f(O2) = 0, 1, and 5 %, respectively. In a conventional three-dimensional crystal growth process, n is known to be in the range of 3-4, while a two-dimensional growth process is characterized by n in the range of 2-3. Thus, it can be concluded that the crystallization of the present sputter-deposited TNO films proceeds mostly in a two-dimensional manner. The range of n values also gives information about the nucleation process. When nucleation occurs continuously with a constant rate, n should be 3. In a process with saturated nucleation sites, where the total number of nucleation sites is limited, n takes a value of 2. The n values of the TNO films deposited at f(O2) = 1 and 5 % are very close to 2, indicating that the nucleation sites are consumed in the early stage of growth and that crystal growth propagates in lateral directions from the nuclei. On the other hand, the n value of the TNO film deposited at f(O2) = 0 % is rather close to 3, implying continuous nucleation during the crystal growth.

The two-dimensional growth of fully crystallized TNO films was also confirmed by POM observations. As shown in Fig. 3(a)-3(c), the grains have lateral sizes exceeding several μm. Therefore the crystal growth of TNO films is two-dimensional as the grain size in the lateral direction was much larger than the film thickness (300 nm). This result is consistent with the in situ XRD measurements described in the paragraph above.

Impact of microstructure due to shadowing effect

As described above, amorphous TNO films crystallize with a site saturation mode in a two-dimensional manner. The site saturation growth means that the density of nuclei, which reflects grain size after crystallization, is constant irrespective of annealing condition. Therefore, the difference in the density of nuclei should be attributed by the structure of amorphous films.

According to Thornton model, films deposited at higher P tend to show porous structures and rough surfaces due to the shadowing effect [6]. Because of collisions with Ar atoms, the mean free path of sputtered particles decreases with increasing P, which enhances the shadow effect and thus affects the microstructures of the films. Figs. 3(a), 3(b), 3(c) show cross sectional SEM micrographs of amorphous films deposited at P = 0.2, 1, and 2 Pa, respectively. The microstructure becomes more porous with increasing P, being consistent with Thornton model. Figs. 4(a), 4(b) show POM images of crystallized TNO films fabricated under P = 0.2 and 1, respectively. The microstructure of the TNO films fabricated at P = 2 Pa was observed by SEM, instead of POM, due to small grain, as shown in Fig. 4(c). It is evident that the TNO films deposited at lower P, where the influence of shadow effect is weak, exhibit larger grain sizes. It is suggested that the shadowing effect becomes less significant at lower P, resulting in the decrease of the density of nuclei and thus the enhancement of grain sizes. That is reasonable because the surface of void, a kind of interface, could work as nucleation centers. The existence of shadowing effect was further confirmed by the following experiment shown in Fig. 5. Three holes with diameters of 1, 1.5, and 2 mm were made on a 1-mm-thick metal mask. As the diameter decreases, slant particles, which are responsible for the shadowing effect, are expected to be filtered. Fig. 5 shows POM images of the films crystallized from amorphous ones prepared by this mask at P = 1 Pa. Notably, the grain size increases with decreasing the diameter of the holes, supporting the conclusion that shadowing effect determines the density of nucleation centers and thus the size of TNO grains.

Summary

The crystallization kinetics of amorphous sputtered TNO thin film during isothermal annealing was studied using in situ XRD measurements. The obtained Avrami exponents were in the range of 2.0 to 2.7, suggesting that the crystallization of amorphous TNO films is two-dimensional with dominant site saturation. This result is consistent with the results of ex situ POM . The microstructures of amorphous films were well explained by Thornton model. That is, the microstructure became more porous with increasing P. On the other hand, the grain size of crystallized films increased with decreasing P. These results suggest that voids act as nucleation sites. By using a mask to prevent the arrival of slant particles on substrates, the grain size was drastically improved even at high P.

Fig. 1 In situ XR profile of anatase (101) peak of TNO film deposited at f(O2) = 1%

Fig. 2 (a) Crystalline fraction x as a function of time t for TNO films deposited at f(O2) = 0%, 1% (vertically shifted by 1 for clarity), and 5% ( vertically shifted by 2 for clarity), (b) Avrami plots for TNO films derived from (a).

Fig. 3 POM images of the isothermally crystallized TNO films for f(O2) = (a) 0, (b) 1, and (c) 5% at P = 1 Pa

Fig. 4 Cross sectional SEM micrographs of amorphous TNO films deposited under different conditions: (a) P = 0. Pa, (b) P = 1 Pa, and (c) P = 2 Pa.

Fig. 5 (a)-(b) POM images of TNO films crystallized from amorphous films deposited under conditions: (a) P = 0.2 Pa, (b) P = 1 Pa, and (c) Surface SEM micrograph of film crystallized from amorphous film deposited at P = 2 Pa

Fig. 6 (a) Schematic illustration of metal mask. (b)-(d) POM images of TNO films crystallized from amorphous films deposited at P = 1 Pa with metal mask.

透明導電体は、高い電気導電性と高い可視光透過率を併せ持つ物質であり、液晶パネルなど広範囲で実用に供されている。透明導電体の主流はITO(SnドープIn2O3)であるが、近年、Inの枯渇が問題視されており、ITO代替物質への要求が高まっている。アナターゼ型のNbドープTiO2(TNO)は有望なITO代替材料の一つであり、実用的な合成プロセスの確立が急務とされている。本論文では、ガラス上に低抵抗TNO薄膜を作製する方法として、アモルファス前駆体を結晶化させる手法に着目し、アモルファスTNOの結晶化過程の解析、膜内組織構造と合成条件との関係、抵抗率を制限する要因について調べ報告している。

本研究は以下の6章より構成されている。

第一章は序論であり、本論文の背景および目的が述べられている。この章では、まず透明導電膜の電導機構、歴史と応用に関する研究を概観し、次世代透明導電膜の開発動向についても述べている。さらに、TNOに関する既往の研究をまとめ、現在の課題について議論している。TNOは、従来型の透明導電膜には見られない特徴を有しており、有望な代替材料であるが、導電性の面では実用レベルに達していないと指摘している。また、大面積化に対応できるスパッタ法は有望な合成手法であるが、結晶粒径の改善が課題であると述べている。

第2章は実験手法とその原理の説明である。TNO多結晶薄膜の合成には、RFスパッタ法により作製したアモルファス薄膜を還元雰囲気下でアニール処理し、結晶化させる手法を用いているが、本章では、スパッタ法の機構と特徴を解説するとともに、評価手法であるホール効果、偏光顕微鏡(POM)、走査型電子顕微鏡(SEM)、X線解析(XRD)、X線反射率法(XRR)などの原理と、そこから得られる情報について詳説している。

第3章はアモルファスTNOの結晶化過程について述べている。上記の薄膜作製方法では、アニール中の結晶化過程は非常に重要であるが、メカニズムの詳細な検討はなされていない。本章では、高温XRDを用いて結晶化速度を求め、これをJohn-Mehl-Avrami(JMA)式を用いて解析して、アモルファスTNOの結晶化機構について調べている。JMA式から見積もった反応次数nは2.0から2.7の間にあることから、結晶化過程は二次元の結晶核飽和過程であると結論している。またこの結果より、結晶化後の粒径は、アモルファス中の結晶核の密度に依存すると指摘している。

第4章はアモルファスTNOの微細構造について述べている。アモルファスの微細構造はThorntonモデルで説明できることを指摘し、成膜圧力Pの増加に従ってアモルファス薄は多孔質になると述べている。一方、Pの減少につれ結晶化後の粒径は大きくなることから、斜影効果により形成された空隙などの欠陥が結晶核となり、その密度が結晶粒径を規定していると推論している。さらに、同様の傾向はドープしていないTiO2にも見られ、ThorntonモデルはNb濃度に依存しないと述べている。

第5章は抵抗率の改善法について述べている。低抵抗を実現するには、結晶粒の拡大が必要であるが、低圧下で作製した粒径の大きい薄膜でも抵抗率の低下はみられないと報告している。低圧下で作製した薄膜からはAr不純物が検出されたことから、高エネルギー粒子によるプラズマダメージに加え、Ar不純物による電子散乱が抵抗を上昇させる原因となっていると推論している。そこで、高エネルギー粒子の割合が小さい高圧下で斜影効果を抑制するため、斜め方向飛来粒子をフィルターで遮蔽する方法を提案している。また、実際にフィルターを用いて成膜を行い、高圧下で作製した薄膜でも劇的に結晶粒径が増大することを見出している。抵抗率も5×10-4 Ωcmを実現し、この値は実用化レベルに達していると評価している。

第6章は結論と総括である。

以上のように、本論文は、アモルファスTNOが結晶化する過程の速度論的解析ならびにアモルファス内部の構造観察を基に、低抵抗TNO多結晶薄膜を作製する方法を提案するものである。これらの研究は理学の展開に大きく寄与する成果であり、博士(理学)に値する。なお本論文は複数の研究者との共同研究であるが、論文提出者が主体となって行ったものであり、論文提出者の寄与は十分であると判断する。

したがって、博士(理学)の学位を授与できると認める。