半导体晶圆匀胶系统在线缺陷检测方法研究

发布时间：2018-10-17 12:22

【摘要】：近年来,随着我国半导体行业的迅速发展,光刻工艺复杂度不断提高,品圆的缺陷来源越来越多元化,晶圆检测的传统方式主要为人工目检,该检测方法效率低,且易受主观因素影响,产生漏检、误检的情况,已不能满足工业需要。为了提高晶圆缺陷检测速度与可靠性,实现晶圆缺陷在线视觉检测,降低缺陷对半导体制造业的影响,提高产品的良品率,本文在分析现有缺陷检测方法的基础上,结合计算机技术与图像处理技术,对该课题展开研究。首先,为了获取高分辨率的晶圆表面图像,选用合适的图像采集设备,搭建了图像采集系统。在光源选择中,系统选用红色同轴光源,不仅克服了晶圆表面高反射度的缺点,而且红光较长的波长也最大程度的保护了晶圆表面的光刻胶。微米级的检测精度限制了相机的视场,为了获取完整的晶圆图像,且要避免相机的抖动造成干扰,搭建了三轴运动平台,负责吸附晶圆,按照规定的路径移动,捕获到整个晶圆的完整图像。其次,针对图像在采集、传输过程中,受到光源性能、通道带宽、噪声等的影响而产生噪声的问题,根据图像噪声的特点与算法实验对比,确定采用改进的中值滤波法—多重中值滤波,滤除图像噪声；为了提高图像的对比度,突出我们感兴趣的区域,滤波完成后,采用直方图规定化的方法对图像进行增强,增强图像的判读和识别效果。最后,为了搜索匹配到晶圆图像中所有的Die单元,对模版匹配算法进行了深入的研究,由Die单元分布特点,结合算法实验效果对比,确定采用归一化互相关的模板匹配算法,该方法对图像灰度值的线性变化具有“免疫性”,对光照改变具有较好的鲁棒性。通过图像差影法得到差分图像,将得到的差值与预设的阈值进行比较,大于阈值,则标记为缺陷点。通过大量的实验,验证了该检测系统的可行性,有效的检测出品圆表面的缺陷,精度到达15μm。该系统可在实际应用中取代人工,实现快速准确的缺陷检测。
[Abstract]:In recent years, with the rapid development of semiconductor industry in China, the complexity of lithography technology is increasing, and the source of defects of wafer is more and more diverse. The traditional method of wafer detection is mainly artificial inspection, and the efficiency of this method is low. And it is easy to be affected by subjective factors, which can not meet the needs of industry. In order to improve the speed and reliability of wafer defect detection, realize the on-line vision detection of wafer defects, reduce the influence of defects on semiconductor manufacturing industry, and improve the rate of good products, this paper analyzes the existing defect detection methods. Combined with computer technology and image processing technology, the research on this subject is carried out. Firstly, in order to obtain high resolution wafer surface image, an image acquisition system is built by selecting appropriate image acquisition equipment. In the selection of light source, the system not only overcomes the shortcoming of high reflectivity of wafer surface, but also protects the photoresist of wafer surface with the longest wavelength of red light. In order to obtain the complete wafer image and avoid the interference caused by the camera's jitter, a three-axis motion platform is built, which is responsible for adsorbing wafer and moving according to the prescribed path. Capture the complete image of the entire wafer. Secondly, in the process of image acquisition and transmission, the noise is caused by the influence of the performance of light source, channel bandwidth, noise and so on, according to the characteristics of image noise and algorithm experimental comparison, In order to improve the contrast of the image and highlight the region of interest, the image is enhanced by the method of histogram specification after the completion of the filtering, the method of multi-multiple median filter is adopted to filter the noise of the image, so as to improve the contrast of the image and highlight the region of interest to us. Enhance the image interpretation and recognition effect. Finally, in order to search all the Die elements matching to the wafer image, the template matching algorithm is deeply studied. According to the distribution characteristics of the Die elements and the comparison of the experimental results of the algorithm, the template matching algorithm of normalized cross-correlation is adopted. This method is immune to the linear change of image gray value and robust to illumination change. The difference value is compared with the preset threshold value by the method of image difference. If the difference value is greater than the threshold value, the difference will be marked as a defect point. Through a large number of experiments, the feasibility of the detection system is verified, and the defect of the circular surface of the product is effectively detected, and the precision is up to 15 渭 m. The system can replace manual in practical application and realize fast and accurate defect detection.
【学位授予单位】：沈阳建筑大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN305;TP391.41

【相似文献】