Focusing on the monolithic integration of OEICs, selective area growth (SAG) achieved by metal organic vapor phase epitaxy (MOVPE) is one of the most important monolithic integration methods. With photolithography and etching techniques, dielectric mask can be formed on the substrate. In SAG-MOVPE, because there will be no-consumption of precursors on dielectric mask, concentration gradient of precursor in gas phase would occur between mask and intact region on the substrate. Then the precursor gradient would induce extra gas phase diffusion, which results in growth rate enhancement in selective area (SA). One of the features of SAG is that by deliberately designing the mask pattern and key parameters, growth rate and elemental composition can be controlled in each SA regions. In fact, SAG combined with MOVPE technique has been well applied in fabricating OEICs and it gains tremendous advantages over other alternative ways.

For SAG-MOVPE technique, the challenge is how to precisely construct more complex device structures on the even narrow area. MOVPE is a quite complicated process, it consists of many important steps such as gas phase decomposition of precursor, their diffusion to the surface, adsorption, desorption, diffusion on surface, and surface reactions to form epi-layer. Though various mechanisms have been suggested to explain the MOVPE process, due to complicated transfer of heat and mass and chemical reactions both in the gas phase and on the surface, many details still remain unclear, especially the growing process taking place on the surface of epi-layer. Thus, till now the design of SAG process is only based on empirical optimization procedures. Such trial and error procedure could be effective for rather simple circuits, but not a suitable solution of densely integrated circuits for enormous time and resources expense. Moreover, it fails to predict the growth rate and composition when interference phenomenon occurs between neighbor masks for highly compact OEICs.

Consequently, we try to set up a systematic and reliable method to predict the growth behavior in SAG-MOVPE and to assist the design of integration process. For this target, it becomes quite necessary to acquire surface kinetics by understanding the mechanism of epitaxial process. In my study, SAG is not only used as an integrating method, but an important tool for kinetic analysis of MOVPE as well. With this analysis process, we are going to build up a kinetic database which includes surface reaction rate constants (ks) of elemental binary compound such as InP and GaAs to much complicated quaternary compound, InGaAsP. The purpose of this research is to optimize the mask designing and process designing for OEICs integration on basis of some key kinetic parameters in SAG process and to control macro scale growth behavior such as growth rate and composition distribution in device structure.

Due to the complicity of InGaAsP material system, this research started from the binary material, and same kinetic analysis was applied for ternary material. In this research, relation of surface kinetics between simple material and complicate material system has been clarified and mechanism was proposed on base of kinetic data from these investigations and some previous work. With kinetic model and perspective on InGaAsP growth, emission wavelength in SAG was predicted based on elemental composition and thickness distribution control. In the last part of this work, some issues in the design of SAG process has been discussed with my established kinetic database, which cover poly-crystal free growth, mask interference in compact pattern and abrupt growth in edge of mask.

From this research, some important surface kinetics has been revealed and useful suggestions can be concluded for SAG based integration of OEICs in MOVPE. The important features are summarized as below.

1.Investigation exhibits that gas phase diffusion model can well explain surface phenomena in MOVPE, and it can be applied in collecting kinetic information in MOVPE process. With SAG technique, surface kinetic which is hindered in mass transfer limited regime can be successfully extracted by numerical simulation and experimental data fitting. Results show that in MOVPE process, surface reaction rate constant present a dominant role in determine growth behavior in the selective area and the relative investigation of this parameter would be valuable to process design of fabrication for OEICs.

2.Kinetic analysis had been successfully made for InGaAsP material system, kinetic database of surface reaction rate constant (ks) has been built up for better understanding on MOVPE and for achieving accurate process design of SAG. My investigations indicate that ks have temperature and concentration dependency. Generally, ks would increase with increasing temperature. Under practical growth conditions, it indicates indium species in MOVPE has rather moderate temperature dependence and is not as sensitive as ks of GaAs whose activation energy is about 84.1 kJ between 475 ℃ to 610 ℃. Additionally, gallium and indium species shows quite different surface kinetics. Compared with InAs and InP, the data indicate values of ks for both InP and InAs are significantly larger than that of ks for GaAs. Their value of ks is two or three times of the value of gallium species in MOVPE. As for indium related binary compounds, ks value of InP growth is always larger than that of InAs. All these kinetic information suggest that group-V elements have significant influence on the value of ks for III-V binary compound. These results show group- III species have quite different reactivity on phosphorus and arsenic site which could be a fundament for kinetic analysis of ternary and quaternary compounds, such as InAsP and InGaAsP.

3.In this study, non-linear phenomena which depend on mask width and partial pressure of precursors were observed in kinetic analysis for both indium and gallium related materials. And all evidences proved that it is essential to consider concentration dependency due to over saturated adsorption of species on surface. Then non-linear simulation based on Langmuir-Hinshelwood model was introduced into kinetic analysis procedure to explain mask dependency and partial pressure dependency of precursor on surface reaction rate constant. Results showed non-linear kinetic analysis had been well applied in understanding these non-linear phenomena. It revealed that adsorption-desorption process should not be neglected in MOVPE at low temperature or high concentration. Combining nonlinear analysis data and practical operating condition in MOVPE, it suggests that under low precursor concentration, SAG could avoid non-linear phenomena and linear kinetic data would be appropriate to precise process design. Such kinetic information could be very precious and useful to gain perspective of surface process or design device fabrication with MOVPE technique.

4.On base of binary kinetic database of ks, investigation revealed overall ks obtained in experiments for ternary compounds can be linearly calculated from binary ks and elemental ratio of group-V species. It clarified the relation of surface kinetics between binary material and ternary material system, and suggested that incorporation of indium and gallium species in MOVPE should be independent to each other.

5.With kinetic model and established database for InGaAsP material system, emission wavelength in SAG had been simulated for a multi-width mask pattern based on elemental composition and thickness distribution control in SAG. This demonstration was designed to test reliability of both kinetic model and our database, and the simulated results showed very good consistency with experimental data. Therefore, it clearly indicated the reliability of my model and database. The most significant meaning of this demonstration is that it confirms the feasibility of process design via numerical simulation on base of kinetic model, and it could be expanded to compact OEICs design with the help of computer.

6.In the last part of this work, some issues in the design of SAG process has been discussed with my established kinetic database, which cover free-polycrystal growth, mask interference in compact pattern and abrupt growth in edge of mask. These applications exhibit great advantage to use my process design methods in SAG, it can save time and resources. Instead of repeating growth and analysis in trial and error way, all these work can be done on computer with sufficient kinetic data for MOVPE. The discussion on issues, such as poly-crystal free growth, mask interference in compact pattern and abrupt growth in edge of mask, indicate that poly-crystal free growth and abrupt growth in edge of mask can be easily achieved and mask interference even be used to reduce the mask coverage with help of such method. A reliable kinetic database shows most significance in these applications, it reveals that the different surface kinetics between indium and gallium should be concerned in all the mask design.

光ファイバー通信の普及に伴い,発光デバイス,受光デバイス,光スイッチ,光アンプ,導波路などを集積した,光電子集積回路(Opto-Electronic Integrated Circuit,OEIC)への期待が高まっている。OEICを作製するには,個々の部品を個別に作製して張り合わせるハイブリット方式と,1枚の半導体基板に集積作製するモノリシック集積がある。後者の方がアライメントなどの問題がなく,高密度集積も望めるが,従来の薄膜形成とエッチングを繰り返す作製手段では,製造歩留まりが低く,現実的でない。そこで,少ない結晶成長によりOEICを作製する技術の開発が要求される。有機金属ガスを原料とするMOVPE(有機金属気相エピタキシー)による化合物半導体の結晶成長は,良質な結晶を大面積基板上に一括して成長できるだけでなく,シリコン酸化膜などをマスクとして形成して部分的に被覆すると,マスク上には結晶成長せず,開口部のみに結晶が選択的に成長する。この選択成長では,マスク面積が大きいほど,開口部へ集まる原料濃度が高くなるため,マスク形状や大きさの制御により,成長する結晶の膜厚や組成を制御することが可能となる。このことを利用すれば,1回の成長により基板上に局所的にアンプや発光素子などを作りこむことができ,OEIC作製に応用することが可能となる。

本論文は,"Process development of selective area MOVPE for Opto-Electronic Integrated Circuits based on InGaAsP"(InGaAsP系光集積回路用MOVPE選択成長プロセスの開発)と題し,上記選択MOVPEプロセスの開発とInGaAsP系OEIC作製への展開を検討した結果をまとめたものであり,全部で7章からなる。

第1章では,研究背景と既往の研究をとりまとめ,選択成長技術を利用してOEICを作製する際の課題などを取りまとめ,第2章では,実験に用いた装置や実験結果の解析方法について述べている。

第3章では,選択成長により形成した膜厚分布の解析から,表面反応に関する情報が得られることを示している。すなわち,選択成長領域の膜厚分布は,気相中での拡散と表面での反応のバランスによって決まる。そのため,選択成長のプロファイルをシミュレーション結果と比較することにより,D/ks(Dは拡散係数,ksは表面反応速度定数)の値が求まり,Dは予測可能な物性値であることから,表面反応速度に関する情報を得ることが可能であることを示している。その際にマスク端から20μmくらいは表面拡散の影響があるため,解析対象から除外することにより解析の精度が高まることを,表面拡散を考慮したシミュレーションから明らかにしている。

第4章では,InP,InAs,GaP,GaAsの2元系化合物での選択成長を利用した表面反応機構の解析結果についてまとめている。まず,InP,InAs系では,表面反応速度定数が大きく,比較的小さな活性化エネルギーを持っていること,低温・高濃度の条件では,表面反応速度定数に濃度依存性が表れることなどを報告している。特に,表面反応速度定数に濃度依存性が表れることについては,Langmuir-Hinshelwood型の速度式を用いて整理することにより,実験結果をうまく再現できることを示しており,これによりマスクが大きい場合(実効的な濃度が高い場合)の選択成長形状の再現性を高めている。GaPについては,選択成長を行うと異常成長が起こり,選択成長の解析から表面反応に関する情報を得ることができなかった。このため,GaAsP成長での成長プロファイルから,GaP成長時の表面反応に関する情報を推測している。なお,GaAsについては既往の研究結果を利用できるため,本論文では実験的解析を行っていない。

第5章では,3元系,4元系の結晶成長における選択成長の解析と予測について検討を行っている。まず,InAsP系では,InAsとInP成長でのIn種の表面反応性が異なり,P上へのIn種の反応性の方が,As上への反応よりも高いことを示している。このため,InAsP成長では,In種の反応速度は,P上への反応速度とAs上への反応速度の線形的な和として与えられると考え,解析を行っている。解析の結果からは,3元系での反応速度は2元系での成長速度の線形結合として評価することでほぼ問題がないことが示されている。このため,第4章で構築した2元系でのIn種,Ga種の反応速度を基に,3元系,4元系での選択成長プロファイルを予測可能であることを示している。さらに,実際にInGaAsP4元化合物半導体をマスク幅が異なるパターンを用いて成長させたところ,PL(Photo Luminescence)発光波長のマスク幅依存性が予測と良く一致することを示し,今回の研究で得た速度パラメータが正しく,選択成長を利用した化合物半導体の成長をうまく設計・制御可能であることを証明した。

第6章では,選択MOVPEプロセスを利用したOEIC作製において,問題となるいくつかの事項についての検討結果を示している。まず,マスク上に多結晶薄膜が成長する場合があり,この場合にはマスク上へも析出が起こるために選択成長プロファイルが予測どおりに形成されない。このマスク上の多結晶薄膜の成長は臨界濃度が存在するため,これを実験的に求め,実際のマスク設計においては,マスク上の濃度がこの臨界濃度を超えないように設計する必要があることを示している。また,マスク同士が隣接すると選択成長に他のマスクからの干渉が現れる。この効果について定量的に検討を行い,マスク間の干渉を考慮したマスク設計が必要となることを示している。さらに,急峻な膜厚変化,膜組成変化を実現するためのマスク形状に関する考察を行い,通常のマスクの外側と内側に三角形のテーパー状マスクを追加することにより上記要求を満たせることを示している。

第7章では上記の成果を取りまとめ,本論文の結論を述べている。

このように,本論文は,OEICのモノリシックインテグレーションを実現する選択成長プロセスの基礎となる表面反応速度定数のデータベースを構築し,これをもとにOEIC作製プロセスの設計を可能としたものである。マテリアルプロセスの高度化を具現化した本研究は,マテリアル工学の発展に多大な貢献をするものであり,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。