用于亚0.1μmCMOS器件技术中的镍硅化物研究

【作者】 蒋玉龙；

【导师】 李炳宗；

【作者基本信息】 复旦大学 ， 微电子学与固体电子学， 2005， 博士

【摘要】 随着CMOS集成电路制造技术持续向亚0.1μm工艺推进,与之相应的自对准硅化物（SALICIDE）材料和工艺也在不断发展创新。由于窄线条效应限制,在0.25/0.18μm技术代SALICIDE工艺使用CoSi₂取代TiSi₂。与NiSi相比,最近的研究表明,当线条物理宽度小于40nm时,CoSi₂在多晶硅上的薄层电阻陡然升高。因此,NiSi被广泛看作是研究发展新一代SALICIDE工艺的优选材料。 由于NiSi形成温度低,硅化反应过程中Ni始终是主导扩散粒子,因此早期研究都认为只要使用单步快速热处理工艺（RTP）即可。但近期实验表明,使用单步RTP的NiSi SALICIDE工艺导致在器件有源区边缘出现过度硅化,严重破坏器件电学特性。已有研究表明,使用两步RTP形成NiSi可以抑制过度硅化反应,但这种NiSi SALICIDE工艺尚有许多材料和工艺问题有待研究。本论文系统研究在较低温度范围内Ni/Si固相反应规律和两步RTP NiSi SALICIDE工艺的有关技术。论文主要内容及结果归纳为以下四个部分: 1.论文系统研究了超薄Ni膜在重掺杂n+/p和p+/n浅结型硅衬底上、低温范围内Ni/Si固相反应规律,分析了硅化镍薄膜的组分、结构和特性的变化及其原因。 - 实验从实际生产的角度证实在两种掺杂衬底上都存在第一步退火和第二步退火的工艺窗口。 - 实验发现,经过第一步低温退火,同样厚度的Ni膜硅化反应后在p+/n-Si衬底上的薄层电阻值比n+/p-Si衬底上大很多。测试分析表明,两种衬底上形成的硅化物晶粒尺寸大小不同是造成这种差别的主要原因。 - 实验发现,在重掺杂衬底上Ni/Si反应导致杂质（As、B）显著再分布:在硅化物/Si的界面处形成一个杂质富集峰;在硅化物薄膜表面以下数纳米的地方存在另一个杂质富集峰。研究认为,Ni/Si固相反应过程中的Kirkendall空洞效应（kirkendall voiding effect）是造成近硅化物表面杂质富集峰的原因。 - 根据Kirkendall空洞层在薄膜中的位置,研究发现Ni/Si反应中尽管Ni原子是主导扩散粒子,但Si原子也会向Ni膜中扩散参与反应。 2.为深入了解和更好控制第一步低温退火下的Ni/Si反应,论文系统研究了Ni₂Si薄膜形成动力学。 - 论文建立了Ni/Si固相反应形成Ni₂Si薄膜的反应动力学模型,即线性—抛物线型薄膜生长模型。 - 实验发现,Ni₂Si薄膜在n+/p-Si衬底上比p+/n-Si衬底上更加稳定。更多还原

【Abstract】 As CMOS IC fabrication technology continuously progresses toward sub-0.1μm nodes, the self-aligned silicide （SALICIDE） material and technology also require corresponding innovations. Due to the limitation of narrow line effect, TiSi₂ is replaced by CoSi₂ in SALICIDE technology at 0.25/0.18μm node. Recent studies show that the sheet resistance （Rs） of CoSi₂ formed on poly-Si line with physical line width less than 40nm will dramatically increase. The silicon consumption for CoSi₂ formation is also much larger than NiSi, which conflicts with the shallow junction application. Besides, it is difficult for Co film to form low resistivity compound with SiGe material. With the development of nanometer scale CMOS technologies, NiSi has been considered as the most promising silicide for next generation SALICIDE technology.Because the formation temperature of NiSi is low enough and Ni atoms are always the dominating diffusion species during Ni/Si reaction, a single-step rapid thermal process （RTP） is accepted as one of its advantages in early studies. Recent investigation, however, has shown that NiSi formation by a single-step RTP may result in excess silicidation at the edges of patterned Si regions, which significantly degrades the device characteristics. A novel process technology, i.e., a two-step RTP is introduced to form NiSi, in which only partial as-deposited Ni film will be consumed and the high resistivity Ni₂Si phase will generally form during the first step RTP （RTP1） with an anneal temperature as low as 300℃, the unreacted Ni will then be removed by a wet etch, a second step RTP （RTP2） with a higher temperature is followed to form the low resistivity NiSi phase. So it is interesting to investigate the nickel silicidation in low temperature range and the two-step RTP NiSi SALICIDE technology. The main content and results of this thesis are summarized as follows.1. The solid state reaction of ultra-thin nickel film on both n+/p and p+/n junction substrates has been investigated in low temperature range. The film composition, structure and properties of the formed nickel silicide are characterized.- The experimental results demonstrate that the optimized process windows for RTP1 and RTP2 have been found for silicidation on both n+/p and p+/n junction substrates for CMOS fabrication.- The experimental results show that the Rs value of Ni-silicide formed on p+/n junction substrate is much higher than that on n+/p junction substrate after low temperature RTP1 anneal. With various material characterizations and analysis, it is revealed that the observed difference in Rs between Ni-silicide films should be attributed to the difference between the formed grain sizes. It is revealed that the dopant （As and B） will significantly redistribute after silicidation in the low temperature range, such as 300°C or 450°C. SIMS analysis shows that there is a dopant accumulation peak near the silicide/Si interface, which is caused by segregation and helpful to lower the contact resistance. In the meanwhile, the study reveals another dopant accumulation peak just several nm beneath the silicide surface. Further analysis shows that it is related to the void layer formed by the Kirkendall voiding effect usually observed in thin film reaction. According to the location of the void layer, the study demonstrates that although Ni atoms are the dominating diffusion species during Ni/Si reaction, Si atoms are not immobile, on the contrary they also diffuse into the Ni film and react with Ni there.2. To better control Ni/Si solid state reaction during RTP1, the formation kinetics of Ni2Si thin film is investigated. The kinetics model for Ni2Si formation is proposed, in which the linear-parabolic thin film growth kinetics is discussed. The study demonstrates that NiiSi thin film is more stable on n+/p junction substrate than on p+/n junction substrate. In the anneal temperature range between 262²77°C, the experimental results reveal that the growth of N12S1 thin film linearly depends on reaction time. It means that the film growth is dominated by Ni/Si interface reaction. The study shows that the activation energy for Ni2Si thin film growth in low temperature range is 1.35eV. This activation energy is successfully applied to explain different Rs-T curves.3. The NiSi SALICIDE technology with different RTP schemes has been studied by means of NiSi/Si Schottky contacts."" The experimental results demonstrate that the area-contact characteristics of NiSi/Si Schottky diodes by a two-step RTP are significantly improved comparing with a single-step RTP, i.e., the diode leakage is suppressed when a two-step RTP is applied. The study successfully investigates the dependence of the Schottky barrier height （SBH） distribution inhomogeneity of the NiSi/Si contacts on different RTP schemes by means of temperature dependent I-V technique （I-V-T）, which demonstrates that the two-step RTP NiSi SALICIDE technology is superior to the single-step RTP case in the view of the SBH inhomogeneity.更多还原