SIMaP-EPM laboratory of Grenoble and INES institute of Chambery have both financed this thesis which investigates the effect of thermo-mechanical stresses on the crystal quality during production of silicon ingots for photovoltaic applications. This work begins by showing how photovoltaic industry makes solar panels and the influence of dislocations (defects induced by stresses) on the conversion efficiency. Bibliographic review is also performed in order to describe physical and numerical models of dislocation motion and their multiplication in silicon. Several characterization methods of the dislocation density at the surface of a sample are also presented in the first part of this work.In the second part of this manuscript, comparative study of different quick characterization methods is done in order to show their strength and weaknesses. Therefore, a sample, which is wide, not containing grain boundaries, and having areas of high and low dislocation density, is used as reference sample for the comparison. The first characterization technique studied in this work is the “accurate method” consisting in manually counting the dislocations at the surface of the sample in order to have a precise characterization of dislocation density. The “INES method” uses numerical treatment of SEM pictures to count dislocations. The “Ganapati method” links the grey scale of a sample picture taken with a scanner and the dislocation density. Finally, the “PVScan method”, using the eponymous device, uses diffusion of a laser beam on the surface of the sample for characterization. This comparative study underlines the best applications for each method and which questions should be thought about before performing dislocation characterization.The third part of this work is intended to build two numerical simulations using Comsol commercial software in order to predict dislocation density in silicon ingot at the end of its production. Therefore, Alexander and Haasen model, describing dislocation density and plastic relaxation rate, is implemented into the software and coupled with the thermo-mechanical stress calculation. In the first model, named “continuous evolution”, the entire ingot is taken into account (liquid and solid parts) and, during solving of this numerical simulation, temperature changes continuously. In the second model, named “step by step” only the solid part of the ingot is taken into account with new geometry and new temperature at each step. Both of these models are compared to numerical simulations performed by Japanese and Norwegian teams. Results of the first one are also compared to the experimental characterization of a sample. Thus, this part shows the pertinence of using commercial software for the prediction of dislocation density in a silicon ingot at the end of its production. Its use is simple and shows good adaptability to different furnace geometries and thermal fields.In the last part, ingot/crucible attachment is studied because it creates high stresses and then dislocations in the crystal. This problem is also solved by numerical simulation using Comsol software. Therefore, a physical model is created: the J-integral is used to estimate elastic energy at the attachment area and then this value is compared to ingot/crucible adhesion energy. This model is implemented into the software and the results are compared to an experiment realized during a previous thesis. This numerical simulation is also applied to two attachment configurations of a silicon ingot in order to study the attachment duration, the localization and the size of crystal area impacted by plasticity.
展开▼
机译:格勒诺布尔的SIMaP-EPM实验室和尚贝里的INES研究所都资助了该论文,该论文研究了在光伏应用硅锭生产过程中热机械应力对晶体质量的影响。这项工作从展示光伏产业如何制造太阳能电池板以及位错(应力引起的缺陷)对转换效率的影响开始。还进行了书目审查,以描述位错运动及其在硅中的相乘的物理和数值模型。本文的第一部分还介绍了几种样品表面位错密度的表征方法。第二部分,对不同快速表征方法的优缺点进行了比较研究。因此,将宽的,不包含晶界并且具有高和低位错密度的区域的样品用作比较的参考样品。在这项工作中研究的第一个表征技术是“精确方法”,该方法包括手动计算样品表面的位错,以便对位错密度进行精确表征。 “ INES方法”使用SEM图片的数值处理来计算位错。 “ Ganapati方法”将用扫描仪拍摄的样本图片的灰度与位错密度联系起来。最后,使用同名设备的“ PVScan方法”使用激光束在样品表面上的扩散进行表征。这项比较研究强调了每种方法的最佳应用,以及在进行位错表征之前应考虑哪些问题。本工作的第三部分旨在使用Comsol商业软件建立两个数值模拟,以预测硅锭在位错处的密度。生产结束。因此,将描述位错密度和塑性弛豫速率的Alexander and Haasen模型实现到软件中,并与热机械应力计算相结合。在名为“连续演化”的第一个模型中,考虑了整个铸锭(液态和固态部分),在求解此数值模拟期间,温度连续变化。在名为“逐步”的第二个模型中,仅考虑铸锭的固体部分,并在每个步骤中采用新的几何形状和新的温度。将这两个模型与日本和挪威团队执行的数值模拟进行了比较。第一个结果也与样品的实验特性进行了比较。因此,该部分说明了在生产结束时使用商业软件预测硅锭中位错密度的相关性。它的使用简单,并且对不同的炉子几何形状和温度场具有良好的适应性。最后,研究了晶锭/坩埚的附着,因为它会产生高应力,然后在晶体中产生位错。使用Comsol软件进行数值模拟也可以解决此问题。因此,创建了一个物理模型:使用J积分估算附着区域的弹性能,然后将该值与铸锭/坩埚粘附能进行比较。该模型在软件中实现,并将结果与上一篇论文中实现的实验进行比较。该数值模拟还应用于硅锭的两种附着构造,以研究附着持续时间,受塑性影响的晶体区域的大小和位置。
展开▼