[Introduction]

Transition metal oxides have attracted great interest as a stage in which a wide variety of unique physical properties, such as metal-insulator transition, high-Tc superconductivity and colossal magnetoresistance and, appear, particularly at low temperatures. For applying these materials to electronics, it is of essential importance to make them in high-quality thin film form, which requires comprehensive knowledge of their initial growth processes on lattice matched substrates. The initial growth processes have been studied using atomic force microscopy (AFM) in ambient atmosphere at ten-nm scale or diffraction methods, such as reflection highenergy electron diffraction (RHEED) in vacuum. Although real-space observations using scanning tunneling microscopy (STM) are thought to provide more direct information about the initial growth processes, such attempts have been scarcely made so far, mainly due to dull-contrast STM images caused by ionic nature of metal-oxygen bonds and poor surface preparation techniques for epitaxial growth on an atomic scale, compared to Si or III-V compound semiconductors. To overcome these difficulties, an ultrastable STM instrument with enhanced signal-to-noise ratio and well-established oxide substrate preparation methods are highly desired.

The purpose of the present study is to observe homoepitaxial growth process of perovskite SrTiO3 (STO) on the truly atomically-ordered STO substrate at the atomic level, using ultrastable STM. What I have done in this study covers three topics. The first issue is development of vibration-free STM to overcome thermal drift for long-time measurements and to enhance the STM image contrast. The second issue is establishment of surface preparation methods for oxide substrates from the viewpoint of thin film growth, as well as of surface science. The final subject is atomic-scale observation of homoepitaxial growth process of sub-ML SrTiO3 using pulsed laser deposition (PLD) technique.

[Vibration isolation for ultrastable STM]

Figure 1 shows a schematic illustration of the newly-developed STM system combined with a PLD chamber. Based on the general principles, the STM has the following outstanding features for vibration isolation. First, the STM head and the damping system are located on a foundation completely separated from a building floor to eliminate vibration from the building. Second, the STM system is equipped with a two-stage damping system composed of a typical passive table and actively-controlled damping legs. In general, double passive suspension springs are considered to be most effective for vibration isolation [1]. However, the two passive dampers were found to enhance original vibrations below 20 Hz, rather than to reduce them, because their resonant frequencies are around a few Hz. To avoid this problem, I replaced the lower passive damper with actively-controlled damping legs, which can suppress vibration transfer even around the resonant frequency and also work similarly to usual passive one above a few ten Hz. Moreover, suspension springs supporting the STM head were also replaced with firm poles for obtaining higher rigidity, resulting in higher resonant frequency near the head.

Figure 2 shows a typical fast-Fourier-transformed dark current noise spectrum with a preamplifier-gain of 109 acquiring on the STM. As can be seen, there are no significant peaks compared with the reference spectrum [2], and the noise level is lower than 10 fA/Hz in a whole frequency range, which is one of the lowest values in the world.

[Preparation of SrTiO3(001)-(√13×√13) substrates]

Strontium titanate (SrTiO3) has been widely used as a single crystalline substrate for epitaxial growth of perovskite oxides due to its well-defined surface with wide terraces separated by equidistant steps, hereafter called as step/terrace structure. However, I found that the step/terrace structure does not assure atomic-scale structural ordering on the topmost surface and that the structural ordering strongly depends on the amount of oxygen deficiencies, manifesting the importance of controlling oxygen vacancies in surface preparation. Here, to establish the preparation process for obtaining atomically-ordered STO substrate surfaces in a PLD chamber, I have examined the thermal and chemical stability of STO substrates by means of ultrastable STM, low- nergy electron diffraction (LEED) and RHEED.

Single crystalline substrates of SrTi(1-x)NbxO3 (x= 0.001) (Nb:STO) (001) were chemically-etched with buffered-HF solution for TiO2-single termination, and then loaded into a PLD chamber with a background pressure of 5×10(-9) Torr. Figs. 3(b) and 3(c) are a RHEED pattern and an STM image, respectively, taken on the STO surfaces prepared by a typical heat-treatment as shown in Fig. 3(a). Although a clear step/terrace structure is visible from the wide-scan STM image (inset), the close-up picture shows structural disordering on an atomic scale. This is consistent with the corresponding RHEED pattern which shows only bulk-like (1×1) sharp streaks. Such structural disorder is possibly due to oxygen deficiencies, which are easily introduced into substrate surfaces during heating in a vacuum condition. This suggests that a heat-treatment process under a more oxidative condition is required to reduce the density of oxygen deficiencies.

As a consequence, I developed a new annealing process, as shown in Fig. 3(d). The annealing in oxygen partial pressure (PO2) of 10(-6) Torr at 500℃ is a degassing process for sample and sample holder without the removal of surface carbon-contamination. Heat-treatment at 850℃ in PO2=10(-5) Torr is for preparing a single wider domain, and short pulse-like heating at 1000℃in PO2=10(-5) Torr makes steps straight. The contaminated carbon can also be removed at 850℃.The STO surface prepared by the new procedure shows a clearer RHEED pattern and STM image, characterized by well-ordered surface reconstruction (Fig. 3(e), (f)). This reconstructed surface can be identified as R33.7°-(√13×√13) structure, being consistent with LEED observations (data not shown) [3,4]. It should be noted that the (13×13)-reconstructed surface can be prepared in an oxide-growth chamber in a reproducible manner. Thus, the STO (13×13) surface is suited as a template on which oxide thin films epitaxially grow.

The (√13×√13) surface shows strong bias-dependent STM images as shown in Fig. 4. It is of great surprise that contrast reversal occurs with increasing sample bias voltages (Vs), where the darker areas at Vs < +1:5 V get brighter with Vs increasing.Recent DFT calculation with transmission electron microscopy measurements proposed a (√13×13√) reconstruction model composed of additional TiO2 topmost layer (TiO2 double-layer model) [4]. Based on the model, Hamada et al. simulated STM images with theoretical calculation [5]. However, the STM simulation cannot reproduce the experimental bias-dependent STM images in the empty states (Fig. 5). Further structural studies on the (13×13) surfaces are needed.

[Atomic-scale investigation of initial stage of STO homoepitaxial growth]

It is of peculiar interest to elucidate how the thin film is grown on the truly atomically-controlled substrate surfaces in the initial stage of oxide growth. I have observed the initial stage of homoepitaxial growth of STO on the (√13×√13) substrate surface by using atomic-resolution STM. Two STO thin films with 0.3- and 1.6-monolayer (ML) thickness were homoepitaxially grown on the Nb:STO(001)-(√13×√13) substrates by PLD at a substrate temperature of 700℃ under PO2=10(-6) Torr. The film thickness was monitored in-situ by RHEED intensity oscillation during deposition. STM measurements were carried out at 77 K on the deposited thin film surfaces without exposing them to air.

Figure 6 shows an STM image taken on the STO film with 0.3 ML coverage at 77 K. The (√13×√13)-based mesh structure is clearly recognizable not only on the substrate surface but also on the first layer of the STO film. The mesh structure can also be seen on the second layer of the 1.6-ML thick film (not shown here). Thus, the STO(001)-(√13×√13) substrate can be used as an atomic-scale template to grow single-crystalline perovskitetype oxides even with 1-ML thickness, leading to further studies on the mechanisms of initial growth processes.

[Summary]

In this study, I have developed a new vibration-free STM with a two-stage damping system consisting of a typical passive table and actively-controlled damping legs. I have also established a preparation process for obtaining truly atomically-controlled SrTiO3(001) step/terrace substrates. Furthermore, the initial growth processes of STO homoepitaxial ultrathin films grown on the (√13×√13) substrate surfaces were observed on an atomic scale using the ultrastable STM. The same (13×13) reconstruction structure was observed on the mono-layer STO film surface as well as on the original substrate. This growth technique is widely applicable to heteroepitaxial growth of perovskite-type functional materials, which would help to fabricate higher-quality epitaxial thin films with well-defined interfaces.

Figure 1: Schematic of newly-developed STM system.

Figure 2: A fast-Fourier-transformed dark current noise spectrum with a reference one [2]. The preamplifier is the same as that of the reference.

Figure 3: Schematic diagram of (a) typical and (d) new annealing processes. (b), (e): RHEED patterns (c), (f): STM images (20×20nm2, Vs=+1.5 V, It=30 pA at 77 K) on the surfaces annealed by process (a), (b), respectively. Insets show wide-view STM images (400×400nm2, Vs=+2.0 V, It=0:2 nA at RT).

Figure 4: Bias-dependent STM images of the STO(001)-(√13×√13) surface (2.9×2.9nm2,It=30 pA at 77 K).

Figure 5: Simulated STM images at (a): Vs=+1.5 V and (b): Vs=+2.5 V based on the proposed model [5].

Figure 6: STM image (15×15nm2, Vs =+1.5 V, It=30 pA at 77 K) taken on the ultrathin STO film with thickness of 0.3 ML. Red broken lines are guides of the (√13×√13) mesh structure.

チタン酸ストロンチウム(SrTiO3)は、ペロブスカイト系酸化物薄膜を作製する際に汎用的に用いられる単結晶基板であり、その上でのホモエピタキシャル成長は、酸化物薄膜の成長様式を理解する上での最も基礎的な系と位置付けられている。本研究では、酸化物薄膜のさらなる高品質化のためのボトムアップアプローチとして、真に原子レベルの秩序を有するSrTiO3(001)基板表面(原子制御基板表面)を得るための手法の確立と、その基板上における薄膜成長初期過程の原子スケール観察について議論している。

第1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。SrTiO3(001)単結晶表面は、薄膜成長、表面科学の両分野において盛んに研究が繰り広げられているが、互いの表面準備条件には大きな隔たりがあり、薄膜成長時に原子制御表面を保障する基板処理技術は確立されていない。とりわけ、酸化物極薄膜における金属-絶縁体転移などの興味ある現象を理解するには、原子制御基板の準備と、初期成長過程の原子スケール観察を通じて、薄膜の結晶性と物性との関係を明らかにすることが重要であると指摘している。

第2章では、実験手法と装置の概略について説明している。まず、本研究における表面観察の主な手法である走査型トンネル顕微鏡(STM)について詳説し、続いて、他の表面観察手法として、低速電子線回折(LEED)、反射高速電子線回折(RHEED)、Auger電子分光(AES)、原子間力顕微鏡(AFM)について紹介している。また、薄膜結晶構造評価手法であるX線回折(XRD)、及び薄膜作製手法であるパルスレーザー堆積法(PLD)についても解説し、最後に本研究で使用したPLD-STM装置の仕様について概説している。

第3章では、STM装置の超安定化のために本研究で導入した除振機構について議論している。除振における理論的背景をまとめた後、その原理に基づいて導入したアクティブ制御除振台を含む除振機構の振動伝達について、系統的かつ定量的な解析を行っている。また、実際のトンネル電流測定時における影響について議論し、その除振機構の有用性を示している。

第4章では、原子制御酸化物基板の準備手法の確立とその表面構造について議論している。まず、ルチル型TiO2(110)基板をモデルケースとして、薄膜成長に向けた酸化物基板表面作製の基礎について議論し、表面不純物及びアニール時の雰囲気、とりわけ酸素分圧の影響について考察している。それらの知見を踏まえた上でSrTiO3(001)基板表面の処理を行い、通常の薄膜成長条件下で(√13×√13)再構成表面を再現性良く得る手法を確立している。この(√13×√13)再構成表面の構造は、過去に提案されたモデルでは説明できないことを示した後、新たな構造提案に向け、走査トンネル分光法(STS)による電子状態観察のデータを提示している。

第5章では、第4章で作製したSrTiO3(001)-(√13×√13)基板表面上にホモエピタキシャル成長を行い、その成長過程初期における原子スケール観察について議論している。特に、原子制御基板を用いることで、薄膜第1層目からコヒーレントなエピタキシーが実現していることを明らかにしている。

第6章では、本研究の結論と今後の展望が述べられている。

以上のように、本研究では、真に原子レベルの秩序を有するSrTiO3(001)基板表面の準備手法を確立するとともに、その上にホモエピタキシャル成長を行うことで、薄膜第1層目からのコヒーレントエピタキシーが実現していることを明らかにした。この基板準備・薄膜作製・表面観察という一連の手法は、他のヘテロエピタキシャル薄膜成長にも容易に応用できることから、これらの研究は機能性薄膜のさらなる高品質化、低次元酸化物における物性物理の解明に大きく寄与する成果である。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。