The spatial failure patterns in wafer defect maps can be related to problems in the manufacturing and test process. Therefore, failure pattern recognition can be used for root cause analysis, which is very important for defect diagnosis resolution improvement and yield learning. However, previous studies show that wafers with recognizable failure patterns only account for a small part of all defective wafers. In order to further improve diagnosis resolution with data collected from wafer test, we propose a novel way to present results of various test items in order to facilitate feature extraction. The test results are exhibited in a DNA-like signature called TestDNA such that sources of wafer defects can be effectively and efficiently identified. Since more defect information can be incorporated into a TestDNA, more defect types of wafer defects can be identified and predicted with higher accuracy. Experimental results show that the proposed TestDNA method can be used to identify and predict more defect types such that both higher diagnosis resolution and prediction accuracy are achieved.

中文翻译：

---

TestDNA：用于诊断和模式识别的新型晶圆缺陷特征


晶圆缺陷图中的空间失效模式可能与制造和测试过程中的问题有关。因此，故障模式识别可用于根本原因分析，这对于缺陷诊断分辨率的提高和良率学习非常重要。然而，之前的研究表明，具有可识别失效模式的晶圆仅占所有缺陷晶圆的一小部分。为了进一步提高从晶圆测试收集的数据的诊断分辨率，我们提出了一种新的方法来呈现各种测试项目的结果，以便于特征提取。测试结果显示在名为 TestDNA 的类似 DNA 的签名中，以便可以有效且高效地识别晶圆缺陷的来源。由于TestDNA可以包含更多的缺陷信息，因此可以以更高的准确度识别和预测更多的晶圆缺陷类型。实验结果表明，所提出的TestDNA方法可以用于识别和预测更多的缺陷类型，从而实现更高的诊断分辨率和预测精度。