【作者】 王金星；

【导师】 刘天模；

【作者基本信息】 重庆大学 ， 材料科学与工程， 2009， 博士

【摘要】 随着晶体管的进一步小型化,由于存在漏电流,传统的SiO₂已经无法满足下一代金属氧化物半导体场效应管（MOSFET）的栅介质要求。为了继续维持摩尔定律预测的发展速度,人们迫切需要找到一种更加合适的高介电材料（即High-k材料）,以取代SiO₂作为晶体管的栅介质。可以说,将来理想的High-k介质的成功研究与应用必将极大地推动半导体技术的快速发展。正因为如此,这些年来有关High-k介质的研究已经成为微电子领域里最关键的热门课题。本课题为德国教育科学研究部（BMBF）的MEGAEPOS科研项目的分支课题,主要目的为寻找合适的、用于下一代晶体管的栅介质材料。在本论文的工作中,我们利用先进的超高真空分子束外延技术（UHV-MBE）成功地在Si基底上制备了三种氧化物纳米薄膜材料:Gd₂O₃、Nd₂O₃以及二者的复合体（GdxNd1-x）2O3。通过原位RHEED、同步辐射光源衍射（GIXD掠角衍射倒易图扫描和线扫描、θ-2θ扫描等）、HRXRD（θ-2θ扫描、Φ扫描、摇摆曲线扫描等）、HRTEM （HRTEM观察、EDX分析、电子衍射分析等）、XRR等手段深入地研究了这些薄膜的生长情况,研究结果表明:（1）Gd₂O₃可以以高质量的晶体结构外延生长在4o斜切的Si（100）表面上。通过对比分析发现,Gd₂O₃薄膜在Si（100）基底上的生长与基底表面的台阶结构有很大关系。干净的、未处理的Si（100）基底表面上存在单原子层台阶结构,而经过1150K/15min热处理后的Si（100）基底表面则转变为单一的双原子层台阶结构。在这两种台阶结构上,生长的Gd₂O₃薄膜均为立方相的方铁锰矿晶体结构,空间群为Ia-3,且均以[110]为面外方向。但是不同的是,在未处理的Si（100）表面上,Gd₂O₃以互相垂直的双晶畴结构生长,而在热处理后的Si（100）表面上,Gd₂O₃以单晶畴结构生长。前者与基底的匹配关系为:面外[110]Gd₂O₃//[100]Si,面内[001]Gd₂O₃//[011]Si和[₁10]Gd₂O₃//[011]Si;后者与基底的匹配关系为:面外[110]Gd₂O₃//[100]Si,面内[001]Gd₂O₃//[011]Si。（2）Gd₂O₃可以以高质量的单晶结构外延生长在S（i111）表面上。在S（i111）-（7×7）再构表面上,生长的异质结构的各界面和表面的粗糙度均小于0.6 nm,生长的Gd₂O₃薄膜为立方相的方铁锰矿结构,空间群为Ia-3,且是以[111]为面外方向生长的。非常重要的是,生长的薄膜中未出现多个晶畴,而是显示出高质量的（111）单晶性能。薄膜立方晶格与Si立方晶格在面内方向存在180°旋转,为A/B匹配结构,匹配关系为:面外[111]Gd₂O₃//[111]Si,面内[1₁0]Gd₂O₃//[₁10]Si。在本课题中还利用同步辐射掠角衍射（GIXD）绘制了Gd₂O₃（111）在面内方向的360o倒易空间图,详细直观地解释了其立方单晶体结构。薄膜和基底匹配非常好,在面内[₁10]Si和倾斜[1₁3]Si方向的失配率分别为-0.1%和-0.2%（相对于2aSi）,这证明了利用MBE外延生长的Gd₂O₃晶格比Si晶格略小的结论。10.89 nm厚的Gd₂O₃薄膜的晶格在面内方向产生了拉伸应变,在面外方向产生了压缩应变,薄膜晶格中发生了部分应变弛豫现象。（3）Nd₂O₃可以以高质量的单晶结构外延生长在Si（111）表面上。在同样的Si（111）-（7×7）再构表面上,生长的异质结构的各界面和表面粗糙度均小于0.7 nm。与Gd₂O₃相同,生长的Nd₂O₃薄膜也为立方相的方铁锰矿结构,空间群为Ia-3,以[111]为面外方向,Nd₂O₃薄膜也具有高质量的（111）单晶性能,与基底的匹配关系也与Gd₂O₃/Si（111）完全相同。但是与Gd₂O₃不同的是,在面内和面外方向,外延生长的Nd₂O₃的晶格都明显比Si晶格大(2a_Si)。8.13 nm厚的薄膜在面内和面外方向的失配率分别为0.66%和3.25%,Nd₂O₃晶格在面内方向产生了压缩应变,应变大小为-1.32%,在面外方向产生了拉伸应变,应变大小为1.22%,薄膜中发生了部分应变弛豫,弛豫度为33%。另外,Nd₂O₃薄膜在面内和面外的晶格常数分别为10.9339?和11.2153A。（4）（GdxNd1-x）2O3（简记为GNO）可以以高质量的单晶结构外延生长在Si（111）表面上。由于与Si相比,Gd₂O₃和Nd₂O₃的失配率一负一正,因此利用MBE外延生长GNO复合薄膜,以希望达到晶格互补、减小失配的目的。分析表明,在Si（111）-（7×7）再构表面上,复合生长获得成功,GNO薄膜晶格中约14%的Gd原子成功地被Nd原子所替代。生成的GNO薄膜的晶体结构与Gd₂O₃和Nd₂O₃薄膜完全相同,仍以[111]为面外方向。更重要的是,GNO表现出比Gd₂O₃和Nd₂O₃还要完美的单晶性能。相比Gd₂O₃和Nd₂O₃的负失配和正失配,约14%Nd原子替换Gd原子的GNO薄膜,在面外和面内方向的失配率甚至连强大的同步辐射光源都难以区分,因此认为该薄膜的晶格与Si晶格大小相等（2aSi）,失配率为零,晶格中也不存在失配应变!所有这些都证明,晶格互补的思路是可行的!另外,初步RTA快速退火研究发现,a-Si/GNO/Si结构即使经过1000℃/30s的退火,在XRR和面外HRXRDθ-2θ扫描手段下仍呈现出良好的热稳定性。本课题研究的三种纳米薄膜,生长在Si基底上均具有较好的晶体结构特性,因此均有希望成为下一代High-k栅介质的候选材料。GNO晶格互补的思路被证明是可行的,互补得到的完美晶体结构和零失配使得GNO极有可能成为最终的栅介质替代材料。更多还原

【Abstract】 With the further downscaling of the transistors, conventional SiO₂ can not meet the gate dielectric demands of the metal-oxide-semicoductor-field-effect transistor （MOSFET） of the next generation because of the leakage current problem. To continue to remain the development speed predicted by the Moore’s law, a suitable material with high dielectric constants have to be urgently introduced to replace SiO₂ as the gate dielectric material of the transistors. Doubtlessly, successful development and application of the future high-k materials will remarkably drive the development of the semiconductor technology. Therefore, the works on high-k materials in these years have become the most crucial research focuses in the microelectronics field. This work was supported by the German Federal Ministry of Education and Research （BMBF） under the MEGAEPOS project. The main aim of this work is to search suitable high-k oxides materials as the gate dielectric for the next generation transistors.In this work, with the employment of an advanced ultra-high-vacuum molecular beam epitaxy system （UHV-MBE）, we successfully grew three types of nano-thick oxide films: Gd₂O₃,Nd₂O₃,and （GdxNd1-x）2O3 which is a mixed material of these two oxides. During and after the layers growth, the growth information was widely investigated, by means of in-situ RHEED, synchrotron radiation diffraction, HRXRD, HRTEM, XRR, etc. Synchrotron radiation diffraction includes the reciprocal space map scans and line scans by grazing incidence x-ray diffraction （GIXD） and theθ-2θscans. Apart from TEM observations, EDX and electron diffraction means mounted in the HRTEM system were also employed to investigate the films in details. Apart fromθ-2θscans, rocking curve scans andΦscans were included in HRXRD. The main results in this work are as follows.Gd₂O₃ was found to have high-quality structure after grown on 4o-offcut Si（100） surface by MBE. Through comparing, we find that the Gd₂O₃ growth largely depends on the steps structure of the Si（100） surface. Single-atomic-layer steps exist in clean Si（100） substrate surfaces without thermal annealing, while double-atomic-layer steps exist in Si（100） substrate surfaces with the thermal annealing of 1150K/15min. On the surface of both types, Gd₂O₃ can be grown with the same cubic bixbyite structure with the space group of Ia-3 and with [110] orientation as the surface normal. What is different is that Gd₂O₃ films grown on the substrates without thermal annealing have two crystal domains perpendicular to each other, while Gd₂O₃ films on the substrates with thermal annealing have only single crystal domain. The former has the epitaxial relationship: out-of-plane [110]Gd₂O₃//[100]Si, in-plane [001]Gd₂O₃//[011]Si and [ 1₁0]Gd₂O₃//[011]Si. The latter has the epitaxial relationship: out-of-plane [110]Gd₂O₃//[100]Si, in-plane [001]Gd₂O₃//[011]Si.Gd₂O₃ was found to have high-quality single-crystalline structure after grown on Si（111） surface by MBE. On Si（111）-（7×7） reconstructed surfaces, all the surface and interface roughnesses of the heterostructure were found to be below 0.6 nm. The Gd₂O₃ films grown on Si（111） have the cubic bixbyite structure with the space group of Ia-3 and with [111] orientation as the surface normal. More important is that there is no domain information in the Gd₂O₃ layer, but a high-quality single crystalline structure. Cubic crystal lattice of the Gd₂O₃ layer was rotated by 180o in the in-plane azimuth away from the Si lattice, revealing an A/B matching structure. The epitaxial relationship is: out-of-plane [111]Gd₂O₃//[111]Si, in-plane [1 1 0]Gd₂O₃//[ ₁ 10]Si. By synchrotron radiation GIXD, 360o in-plane reciprocal space map was successfully made which can clearly interprets the in-plane single-crystalline structure of the cubic Gd₂O₃（111） film in details. Gd₂O₃ and Si were found to have good lattice matching. In the in-plane [₁10]Si direction, the Gd₂O₃ layer is only -0.1% mismatched with the Si substrate and in asymmetric [1₁3]Si direction -0.2% mismatched （relative to 2aSi）. This indicates the lattice parameter of Gd₂O₃ is slightly smaller than the one of the Si substrate. The 10.89 nm layer exhibits compressive strain in the out-of-plane direction and tensile strain in-plane. The small mismatches indicate that the layer is only partially strained.Nd₂O₃ was found to have high-quality single-crystalline structure after grown on Si（111） surface by MBE. On the Si（111）-（7×7） reconstructed surfaces, all the surface and interface roughnesses of the heterostructure were found to be below 0.7 nm. The Nd₂O₃ films grown on Si（111） have the cubic bixbyite structure with the space group of Ia-3 and a high-quality single crystalline structure with [111] orientation as the surface normal, which are the same to those of the Gd₂O₃ films. Different from Gd₂O₃ film, the lattice of the epitaxial Nd₂O₃ is obviously larger than the one of the Si substrate （2aSi）. In the in-plane [₁10]Si direction, the 8.13 nm Nd₂O₃ layer is 0.66% mismatched and out-of-plane 3.25% mismatched with the Si substrate （relative to 2aSi）. The layer exhibits compressive strain of -1.32% in the in-plane direction and tensile strain of 1.22% out-of-plane. The degree of strain relaxation of Nd₂O₃ was estimated to be 33%. In addition, the lattice constants of this Nd₂O₃ film are calculated to be 10.9339? in in-plane direction and 11.2153? out-of-plane.（GdxNd1-x）2O3 （GNO） was found to have high-quality single-crystalline structure after grown on Si（111） surface by MBE. With respect to Si, cubic Gd₂O₃ and Nd₂O₃ have the negative and positive mismatches, respectively. Therefore, GNO films were grown to expect the lattice complementation and smaller mismatch. All the analysis by different means indicates that on the same type Si（111）-（7×7） reconstructed surfaces, the lattice complementation is successful. 14% Gd atoms in the Gd₂O₃ lattice were replaced with Nd atoms. The as-grown GNO exhibits completely the same crystal structure to those of the Gd₂O₃ and the Nd₂O₃ with the [111] orientation as the surface normal. More important is that GNO film exhibits better structure. Compared to the negative and positive mismatches of Gd₂O₃ and Nd₂O₃, G_0.86N_0.14O with 13¹4% Gd atoms replaced with Nd atoms shows little mismatches both in in-plane and out-of-plane. Even with the synchrotron radiation, it is difficult to determine the mismatches. Therefore, the lattice sizes of G_0.86N_0.14O and Si could be considered equivalent! The mismatches and mismatch strains in the film could be considered zero! All the analysis indicate that the method of lattice complementation is feasible. In addition, preliminary investigation reveals that a-Si/GNO/Si structures exhibit good thermal stability in XRR and out-of-plane HRXRDθ-2θanalysis even after the rapid thermal annealing of 1000℃/30s.All these three nano-thick films studied in this work grown on the Si substrates exhibit good crystal structures. These make them to be the promising candidates for the high-k gate dielectric of the next generation. The method of GNO lattice complementation proves feasible. Perfect lattice structure and zero mismatch makes GNO more promising as the replacement.更多还原