Dislocation nucleation with a critical role on plastic deformation of nano-scale structure has been studied for years, while still remains several unknown features. By means of the atomistic calculation based on reaction pathway analysis, the activation parameters of dislocation nucleation transition have been quantified and some issues have been clarified. However, there is a lack of calculation on dislocation nucleation in binary compounds providing the related minimum energy pathway and saddle-point configurations.

At the more basal point, we start from the calculation of dislocation nucleation in a simple substance Cu, which is a classic fcc crystal. The nucleations of 90° and 30° partial dislocations from a sharp corner in Cu have been investigated. The anisotropy aspects of dislocation nucleation revealed by the results have shown that the stress-dependent activation energy of 30° partial dislocation is approximately twice over the counterpart of 90° partial dislocation, and that the maximum inelastic displacement for the former is also higher. Moreover, the shape of the saddle-point configuration of 30° partial dislocation is similar to a half-ellipse whereas in the case of 90° partial dislocation it is more like a semi-circle, which could be principally ascribed to the Peierls barrier differences between edge and screw components and the inhomogeneous stress field.

To further investigate the dislocation nucleation in binary compounds, 3C-SiC is taken as an example, as a classic zinc blende crystal, moreover as one of the promising semiconductors drawing much attention. By focusing on the core element effect, studies on glide-set 90° partial dislocations with Si-core and C-core nucleated from a sharp corner in 3C-SiC have been carried out. The results present that both of the stress-dependent activation energy and athermal strain of C-core dislocation nucleation are higher than that of Si-core. Whereas the gradient of the energy curve for C-core nucleation is lower than the counterpart of Si-core, manifesting the greater thermal sensitivity of C-core dislocation. Such results show that the nucleation of Si-core dislocation is more energy-favorable while the C-core dislocation would nucleate with thermal assistance under high temperature, which explicitly explain the previous experiment observation of core nature effects on dislocation performance in SiC. Besides, estimation of dislocation nucleation behavior under high temperature has also been conduced. The free energy is obviously reduced at deposition temperature, implicating the much higher occurrence rate during crystal growth.

In addition, the role of surface step on dislocation nucleation in 3C-SiC has also been revealed. The activation energy required by dislocation nucleation from a surface step is slightly reduced comparing with that from a sharp corner, which demonstrates the facilitative effect of surface step on dislocation nucleation. Discussion on nucleation rate under deposition temperature has been attempted as well, indicating that the temperature could be greatly influential on nucleation criterion.

After the ideal calculation under theoretical level, more realistic technology problems are confronted. The deposition of 3C-SiC on undulant-Si substrate published by Nagasawa and the co-workers is found to efficiently reduce the planar defects and improved crystal quality. However, the saddle-shape warpage of 3C-SiC wafer becomes a problem ([1,2]). And the mechanisms of stacking faults formation remain unclear as well. The wafer warpage and poor crystalline quality which are the two main issues hindering the developments and applications of 3C-SiC/Si, are devoted by the following work to provide a reasonable analysis and explanation in some extent.

With the purpose of clarify the causes of wafer warpage, from ex-situ curvature measurements and stress calculation, it is found that large compressive intrinsic stress is generated during high temperature growth (1623 K) in both directions parallel and perpendicular to the ridge of undulation on Si substrate. To investigate the intrinsic stress distribution along the 3C-SiC film thickness, reactive ion etching (RIE) has been performed on 3C-SiC film to determine the dependence of SiC/Si system curvature on the remained 3C-SiC thickness. Then, by analyzing the experiment data, it is reflected that intrinsic stress component perpendicular to the ridge of undulation shows nonuniform distribution along the film thickness. Below the 50 μm thickness region, the distribution presents large variation. To explain the warpage in a quantitative manner, the calculation by finite element method taking the intrinsic stress distribution as input has been performed and the obtained curvature of 3C-SiC wafer is in agreement with experiment data. Besides, the obtained intrinsic stress distribution shows very similar distribution of stacking faults, which displays high density near SiC/Si interface and rapidly decreases within 50 μm thickness. It is speculated that the microstructure changes induced by the stacking fault reduction process (stacking fault collision) would be the cause of the intrinsic stress variation.

Moreover, from the viewpoint on glide behavior of partial dislocations as the boundary of stacking faults, the distinct performances involving expanding or shrinking of stacking faults with Si-face (SF(Si)) and C-face (SFC) exposing on (001) surface observed by Nagasawa have been deliberated. By the perspective of dislocation mobility, it is thought that the SFSi bounded by two Si-core 30° partials or one Si-core 90° partial and one C-core 30° partial should be expanding during crystal growth. Whereas, the SFC sandwiched by two C-core 30° partials should be shrinking, which gives rise to the reduction of SFC having no relation with undulation on Si substrate. While the other kind of SFC contains one C-core 90° partial and one Si-core 30° partial as two sides tends to extend itself. Such one could be reduced with the help of enhanced stacking faults collisions due to the undulant-Si substrate. Consequently, it is supposed that the reduction of stacking faults is attained not only thanks to undulation on Si substrate but also because of the different characters of the bounding partial dislocations, which could be instructive for the controlling of stacking faults and improvement of 3C-SiC crystal quality.

The calculations and proposals have mainly promoted the understanding on mechanisms of dislocation formation in 3C-SiC at both theoretical and technological level, which may supply possibility to the further progress of binary semiconductors.

本論文では、原子レベルの解析手法である反応経路解析を使って、銅並びに3C-SiCにおける転位生成の活性化エネルギー及び活性化体積の応力依存性を調べ、さらに、3C-SiCウェーハ製造プロセスにおけるSiC基板の反り発生及び積層欠陥発生・消滅メカニズムの解明への応用を試みたものである。

第1章では、序論として、本論文で扱う銅と3C-SiCの転位生成に関する学術的背景として、それぞれの転位の特性、単元系と二元系における転位の特性の違いについて述べられている。また、転位生成のような分子動力学では解析不能である非常に遅い現象への、本論文で適用する反応経路解析の有効性について述べられている。

第2章では、本研究で用いている分子動力学の手法及び反応経路解析の詳細な手法について、説明がなされている。

第3章では、角部からの銅の転位生成の反応経路解析により、30°部分転位と90°部分転位の生成の活性化エネルギーの応力依存性および鞍点における転位核の原子形態について求めている。結果、90°部分転位が30°部分転位より活性化エネルギーが低いことを示した。これは、90°転位がパイエルスエネルギが低いことに起因すると考えられる。また、90°部分転位はおおよそ等方的な転位ループが生成するのに対して、30°部分転位はパイエルスエネルギーが異なるらせん成分と刃状成分を転位ループの両端に含むため、生成した転位核の形状に異方性が見られることがわかった。また、この異方性が活性化エネルギを高くしていることがわかった。このような結果は、連続体力学ベースの転位論では見出せず、原子レベルのシミュレーションによって初めてわかったことである。

第4章では、角部からの銅の転位生成において確立した反応経路解析の手法を3C-SiCに応用し、二元系特有の二つの転位芯形状(Si-core, C-core)の特徴の違いを明らかにした。活性化エネルギーの応力依存性を調べた結果、Si-coreのほうがC-coreと比較して、活性化エネルギーが低く、より低い応力で転位が生成することがわかった。これはC-coreの転位芯におけるC-C結合の強さに起因するものと考えられる。

Meyer-Neldel則に基づく、活性化自由エネルギーの考察により、低温域においては、活性化自由エネルギーが低いSi-coreの転位が生成するが、高温域においては、高応力でC-coreの転位が生成する可能性があることが示され、実験結果を定性的に説明することができた。また、表面ステップから発生する3C-SiCの転位の反応経路解析を行い、応力場の違いによる発生条件の違いについて論じた。結果、ステップ起因の転位の生成の活性化エネルギーが角部からの生成のエネルギーより小さくなることがわかった。

第5章では、3C-SiC基板製造プロセスに貢献するため研究が述べられている。最初に、発生している応力を知るために、基板の反り解析による真性応力の算出を行った。成長初期には炭化プロセスに起因する圧縮応力が発生し、その後の真性応力は膜厚に依存し、基板となるSiとの界面近くが小さく、界面から離れると大きくなることがわかった。これは、積層欠陥密度の変化に起因するものであり、積層欠陥がより多く消滅している結晶方位でのみ強く生じることがわかった。

次に、第4章の結果により、SiCウェーハ製造プロセスの部分転位起因の積層欠陥生成・消滅メカニズムの解明を試みた。結果、発生しにくく移動度が低いC-coreの部分転位に起因して発生する積層欠陥(SFc)が、自己消滅及び積層欠陥同士の衝突による停止を起こし、密度が低くなると推測される。一方、Si-coreの部分転位に起因する積層欠陥(SF(si))は消滅や停止が起こりにくく、密度が高くなることが推測される。このメカニズムは、Si-core起因のSFsiが最終的に支配的に残留するという実験結果を定性的に説明することができ、3C-SiC基板の積層欠陥密度制御技術への指針を提供することが出来た。

第6章では、結論と研究の展望を述べた。加速分子動力学法の適用による活性化自由エネルギーの算出が課題である。

これらの成果によって、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。