Abstract

The initial stage of non-epitaxial, heterogeneous growth of TiN thin films onto Si (111) was studied by using transition electronic microscopy. Based on the experimental findings on TiN initial growth and the previous experimental studies on metals growth. A model was proposed to understand the non-epitaxial, heterogeneous growth.

This dissertation includes nine chapters. In chapter 1, the classical theories were reviewed firstly, followed by description of the necessity of understanding non-epitaxial growth, introduction of the previous investigation on non-epitaxial growth of metals and the research background of TiN thin films, purposes and experimental approaches of this research. In chapter 2, the experimental setup and analysis was described in details. Deposition condition optimization was described in chapter 3, and the plasma properties during reactive sputtering under various conditions were given firstly, followed by listing deposition conditions for studying the initial growth mechanism. In chapter 4, the evidential results of formation of continuous, amorphous layers and thickness-dependent crystal nucleation were given. In chapter 5, effects of substrate heating and biasing on nanostructural evolution of non-epitaxially grown TiN nano-films were investigated. In chapter 6, texture formation during reactive magnetron sputtering of TiN on Si (111) was described in details. The interface properties during sputter-deposition of TiNx onto Si (111) were investigated in chapter 7. Based on the experimental results, distinguishing features of non-epitaxial, heterogeneous growth were summarized

Chapter 1

Thin films are of interest from many different points of views. Classical initial growth models for thin film growth have not distinguished between epitaxial growth and non-epitaxial growth. For non-epitaxial growth, besides the nucleus density, size distribution etc, which were considered principally in classical theories, the nucleus structure and the nucleus kinetic behavior should be considered in detail due to special features of non-epitaxial growth, such as weaker film-substrate interaction compared with epitaxial growth. Previous researches on non-epitaxial growth have concentrated on metals on amorphous SiO2. In this research we investigate the initial growth of TiN on Si (111) in details at the scales of nanometers, to understand more comprehensively the non-epitaxial growth.

Chapter 2

The Si (111) substrates were H-terminated in HF solution. TiN films were deposited by using magnetron reactive sputtering with a titanium target in an N2/Ar atmosphere, equipped with dc and rf power sources. The total flow, F, was maintained at F = 20 sccm through independently controlling the Ar and N2 flows. The total pressure, P, in the sputter chamber was controlled at P = 0.93 Pa. PN2 was calculated from N2 flow and P, F. The microstructure of the films was investigated with XRD, and with plan-view and cross-sectional HRTEM. The thickness of the film (40nm ? 200 nm) deposited under defined time was measured with a stylus profilometer to determine the deposition rate. The thickness of initial films was calculated from the deposition rate and time. The film composition was determined with in situ Auger Electron Spectroscopy (AES) and ex situ X-ray photoelectron spectroscopy (XPS).

Chapter 3

The cathode discharge voltage, VT, depends on PN2 when the discharge power was kept constant at 69 W dc. VT increased about 40 V when PN2 increased from 0 to 0.47 Pa. This might be due to the less secondary electron emission on the “poisoned Ti” (TiN) surface. The floating substrate potentials, VF, which induced from the different impinging fluxes of positive ions and electrons, increased from - 23 V to ? 29 V when PN2 changed from 0 to 0.47 Pa. The ion flux (Ar+, N2+…) impinging onto the substrate or film surface, calculated from the saturated ion current, was in the order of 1018 cm-2s-1.

The deposition conditions needed to obtain stoichiometric TiN films with good crystallinity were determined by using various N2 partial pressures of the sputtering gases for both dc and rf sputtering. Both films, deposited under dc and rf power, show the same trend: for an increase in N2 partial pressure from 0.015 Pa to 0.47 Pa, the preferred orientation of the film changed from TiN (111) to TiN (200). The deposition conditions needed to clarify the initial growth mechanism of TiN were an N2 partial pressure of 0.047 Pa and 0.47 Pa for dc sputtering and 0.015 Pa for rf sputtering. The other parameters used to clarify the initial growth mechanism were substrate temperature, Ts, and substrate bias voltage, Vb. The deposition conditions are shown in detail in Table 1.

Chapter 4

The initial growth stage of titanium nitride (TiN) deposited by reactive magnetron dc sputtering onto (111)-oriented Si substrates was investigated by using high-resolution transmission electron microscopy (HRTEM). During the initial growth stage, a continuous, amorphous layer was observed when the deposited film was less than 1 nm thick, as shown in Fig. 1(a) and (d). Crystal nucleation occurred from the amorphous layer when the film grew to about 2 nm thick [Fig. 1(b) and (e)]. No preferred orientation was found for the initial crystal nuclei.

The growth of the crystal grains depended on the N2 partial pressure, PN2. When the films were deposited under Condition I, as the deposition time increased from 10 s (lc = 2 nm) to 30 s (lc = 6 nm), the average lateral grain size increased from 3.0 to 4.2 nm. On the other hand, however, when the films were deposited under Condition II, as the deposition time increased from 32 s (lc = 2 nm) to 60 s (lc = 4 nm), the average grain size increased from 3.7 to 4.8 nm. Moreover, for PN2 = 0.47 Pa, all the planar grains with large lateral dimension were found to be (200) oriented. This large lateral size under Condition II was presumably caused by grain lateral growth and coalescence. The grains with anisotropic planar shapes have a larger top area than side area, and tend to restructure into (200)-oriented grains to minimize the surface energy under energetic bombardment of the incident atoms/ions. In Fig. 2, several grains [(e)-(h)] with a coalesced shape and large lateral size are showed, and lattice imaging shows that these grains were (200) oriented, four grains [(a)-(d)] formed during 30-s deposition under Condition I were showed as comparison.

Chapter 5

The effects of substrate heating and substrate biasing on the initial stage of non-epitaxial heterogeneous growth of TiN on Si (111) was studied. Although TiN films deposited at room temperature (RT) undergo a transition from continuous, amorphous films to polycrystalline films with three-dimensional grains when the film thickness increased from ~ 1 to 2 nm, crystallization occurred at a substrate temperature, Ts = 570 K, even for film thickness less than 1 nm [shown in Fig. 3(a)]. During the successive grain growth process, when lc increased from lc = 0.8 to 1.6 nm the lateral grain size increased from 2.2 to 3.0 nm, and the lateral grain size was almost the same as the grain height. (shown in Fig. 4) This indicates three-dimensional grain growth and is identical to the XHRTEM results shown in Fig. 3. At lc = 4.8 nm, the lateral grain growth rate decreased due to the formation of a continuous film, which caused grains to mainly grow vertically. Compared with the grains deposited at Ts = 570 K, the grains grown at Ts = RT were smaller and had lower lateral growth rate, which indicates the formation of a continuous film earlier in the film-growth process. The evolutionary selection growth, therefore occurs later at Ts = 570 K than that at Ts = RT, due to that the grains become continuous later than that at Ts = RT.

At a substrate bias voltage, Vb = -70 V, the grains were laterally larger and were planar as shown clearly in Fig. 5. At the film thickness 50 nm, the film deposited at Vb = -70 V showed thermodynamically favored (200) preferred orientation, whereas the film deposited at Ts = 570 K showed (111) preferred orientation with a weak (200) peak (shown in Fig. 6).

Chapter 6

The initial texture formation mechanism of TiN films were investigated by using transmission electron microscopy (TEM) and X-ray diffraction (XRD). Two power sources for the sputtering, dc and rf, were compared. As has been shown in Chapter 4 that at the initial growth stage, randomly oriented nucleation occurred from continuous amorphous layers when the film thickness was about 3 nm. The nuclei grew and formed a polycrystalline layer when the film thickness was about 6 nm as shown in Fig. 7(a). As the film grew further, its orientation changed depending on the deposition conditions. For dc sputtering, the appearance of (111) or (200)-preferred orientations depended on the N2 partial pressure, and the intensity of the preferred orientation increased with increasing film thickness. This can be seen in Fig. 7(b) and Fig. 8(a). In Fig. 7(b), the grains on the bottom of the film were still randomly oriented, whereas those on the top of the film were dominantly oriented in the (111) direction. For rf sputtering, however, when the film thickness was small, the film showed (200) orientation, independent of the N2 partial pressure, and further growth caused the film to orient to the (111) orientation when the N2 partial pressure was low (about 0.015 Pa) [as shown in Fig. 8(b)]. The results indicated that preferred orientation of TiN films is controlled by a competition between kinetic and thermodynamic effects.

Chapter 7

Fig. 9 shows the XHRTEM images of TiNx films deposited at various PN2. Fig. 9 (a) and (b) show the films deposited at PN2 = 0 and 0.015 Pa. No interlayer can be found between the deposits and the Si substrate for these two conditions. A small intermixing, with thickness less than 0.5 nm, can be found for film deposited at PN2 = 0.047 Pa (Condition I). For film deposited at PN2 0.47 Pa (Condition II), however, a distinct interlayer with thickness 1.5-1.8 nm was formed. The contrast of the interlayer was much brighter than that of the TiNx layer, and closer to that of Si. This indicates that the interlayer was composed of little Ti, because the relatively heavy Ti atoms efficiently scatter electrons, and would therefore darken the contrast of the layer if it contained high concentrations of Ti.

In order to investigate in details the compositions of the interlayer, ex situ XPS depth profiling was made. Fig. 10 shows the depth profiles for 20-nm thick TiNx layers deposited under Conditions I and II. Ti and N compositions maintained a stoichiometric ratio in the films for both conditions I and II. However, for Condition II, N became rich at the TiN/Si interface. Considering that TiN film had a stoichiometric composition, the excess N at the TiN/Si interface could have been caused by the N from the amorphous interlayer. The compositions of the interlayer were mainly SiNx.

Chapter 8

Based on the above experimental findings for initial growth of TiN films onto Si and the previous studies for initial growth of metals, distinguishing features of non-epitaxial heterogeneous growth has been revealed, and the difference from the traditional models for epitaxial thin film growth was clarified. For non-epitaxial growth, the deposit structure is affected by the interface and surface properties, and may lead to an amorphous phase in the initial stage. Two stages (condensation and crystallization) should be considered for the initial amorphous phase of non-epitaxial growth. If the critical size of condensation nucleation is smaller than that of crystallization nucleation, amorphous to crystal transition may be observed, such as in the case of TiN/Si. If the critical size of condensation nucleation is near to that of crystallization nucleation, amorphous to crystal transition may not be detected, such as Au/SiO2. The orientation of the initial grains is mainly determined by the grain surface energy. Restructuring may occur when there is a grain shape-change to meet the free energy minimization. An overall growth model that accounts for the mechanisms above is proposed.

Chapter 9 Conclusions

Finding of evidence of crystal nucleation from amorphous continuous initial layer

Larger lateral grain size at high PN2 or substrate-biasing, and inducing restructuring into (200) orientation

Increasing Ts promotes initial aggregation and delays evolution selection growth

Films grown with rf power have different orientation evolution pass-way due to different plasma effect on initial films

Growth under the conditions with high flux ratio of N2+ causes distinct SiNx interlayer

An overall non-epitaxial initial growth model was proposed.

Deposition conditions used to clarify the initial growth mechanism

Cross-sectional TEM images of TiN deposited onto (111) Si substrates under Condition I for deposition times of (a) 4 s, (b) 10 s, and (c) 30 s, and under Condition II for deposition times of (d) 10.5 s, (e) 32 s, and (f) 60 s.

Enlarged plan-view HRTEM images of grains formed under Cond. I (a-d), and Cond. II (e-h).

XHRTEM images of TiN deposited at Ts = 570 K (Condition III) for deposition times of (a) 4 s, (b) 8.5 s, and (c) 25 s.

Lateral grain size vs. film thickness.

(a) Plan-view TEM image and corresponding SAED patterns of a 3.4-nm film and (b) XHRTEM image of a 2.5-nm film deposited under Condition IV.

Effect of substrate bias voltage on the texture of 50-nm films.

XHRTEM micrographs of TiN films deposited under Condition I, (a) for 30 s, (b) for 120 s, and under Condition V, (c) for 180 s.

Effect of film thickness on preferred orientation of films deposited by using (a) dc and (b) rf sputtering.

Interlayer thickness vs PN2: (a) PN2 = 0, (b) PN2 = 0.015 Pa, (c) PN2 = 0.047 Pa, (d) PN2 = 0.47 Pa

XPS depth profiles of 20-nm thick TiN layers deposited at PN2 = 0.047 Pa (a) and PN2 = 0.47 Pa (b).

薄膜材料は、様々な電気、電子デバイスに利用されており、デバイスの小型化、高密度化に伴って薄膜材料の品質（形状、構造、配向、機械的特性、電気的特性）に対する要求も、ますます厳しいものになっている。この要求に応えるためには、薄膜の初期成長過程と最終的な薄膜の形状、構造、配向、さらには機械的、電気的特性との関係を把握することによって、成膜プロセスを制御することが重要であると考えられる。本論文は「TiN Thin Film Growth at the Nanometer-Scale: Toward an understanding of the initial stage of non-epitaxial growth (TiN薄膜のナノスケールでの成長機構：非エピタキシャル成長の初期過程の解明を目指して)」 と題し、TiNの成膜条件とナノメータースケールにおける初期成長機構及びその後の成長過程並びに最終的な薄膜の形状、構造、配向との関係を解明することを目的としたものであり、全体で9章より構成されている。

第1章は序章であり、物質の気相から基板表面への凝縮、核形成、粒成長、粒融合についての既往の研究がまとめられ、薄膜の初期成長過程を研究することの意義及び本研究の方針並びに目的が述べられている。

第2章は、実験装置及び実験方法が述べられ、また、高分解能透過型電子顕微鏡（HRTEM）による薄膜分析方法が記載されている。

第3章は、TiNの成膜条件と薄膜の組成、結晶性、配向性との関係についての概要が紹介され、詳細解析を行うために選定した5つの成膜条件の根拠と狙いが述べられている。

第4章は、窒素分圧を変えて成膜を行った結果が述べられている。どの窒素分圧下においても、厚さが1 nm未満の初期成長過程にあるTiN薄膜は、連続したアモルファス相を形成しており、2 nm程度の厚さになってから結晶核が発生することが見出された。この初期成長過程の結晶はランダム配向であるが、低窒素分圧下で薄膜の成長が進むと、漸進的選択成長（Evolutionary Selection Growth）によって（111）配向となる。一方、高窒素分圧下では、成長中のTiN薄膜に対して単位時間あたりに入射するイオン量が低窒素分圧下に比べて多くなるため、粒子の融合が進んで横方向の粒子径が大きくなるとともに、熱力学的に安定な（200）に配向することが報告されている。

第5章は、基板加熱と基板バイアスの影響について述べられている。基板を570Kに加熱して成膜すると、室温成膜時に比べて基板表面での物質移動、拡散が盛んになるため、1 nm未満の膜厚においても結晶核の発生が確認された。核発生時のTiNは、球状の結晶粒子になるため三次元的な（凸凹の）構造をとるが、その後の成長によって（111）に配向した連続膜になる。一方、-70Vの基板バイアスをかけた場合には、薄膜に入射するイオンのエネルギーが高いため粒子の融合が進んで横方向の粒子径が大きい平べったい結晶となり、熱力学的に安定な（200）に配向していくと報告されている。

第6章は、DCおよびRFでスパッタした場合の比較が述べられている。どちらの場合でも、初期成長過程の結晶核はランダム配向であるが、RFスパッタではDCスパッタに比べて薄膜へ入射するイオン量が多いため、成長に伴って熱力学的に安定な（200）配向になりやすいと報告されている。

第7章は、反応性スパッタによってTiN薄膜とSi基板との間に形成される中間層（Interlayer）について述べられている。XPS及びHRTEM分析の結果から、形成された中間層はSiNxと推定されている。また、窒素分圧、基板温度、基板バイアスを変えた成膜結果を比較した結果、中間層の厚さは主としてN2+/Tiというパラメーター（イオンフラックス比）によって決まることが見出されている。

第8章は、4章から7章までに記載した実験事実をもとにして、様々な成膜条件下での薄膜の初期成長過程を考察し、モデル化した内容が提示されている。まず、熱力学的なモデルから初期核発生が起こる臨界膜厚を推定した結果が示されている。次に、粒成長及び粒融合の過程におけるイオン照射の影響をN2+/Tiで整理し、TiN薄膜が最終的に（111）配向となったり（200）配向になったりする原因が説明され、配向を制御するための方法が示唆されている。

第9章は、本論文のまとめと展望である。

以上の研究によって、非エピタキシャル薄膜の初期成長機構及びその後の成長過程並びに最終的な薄膜の形状、構造、配向との関係が明らかにされ、成膜プロセスを制御するための基礎情報が得られた。さらに、反応性スパッタによって生成する中間層についても貴重な情報を与えた。これらの結果は、化学システム工学の発展に大きく寄与するものである。よって本論文は博士（工学）の学位請求論文として合格と認められる。