Wafer failure pattern recognition can be used for root cause analysis, which is very important for yield learning. Recently, TestDNA was proposed to improve diagnosis resolution with data collected from wafer test. Previous studies on wafer failure pattern recognition using machine learning achieve good classification results. In this letter, we propose to enhance the classification accuracy with the help of spatial information and ensemble learning algorithms. Experimental results indicate that the proposed method can further improve the accuracy by 8.9%.

In semiconductor manufacturing processes, yield analysis aims to increase the yield by determining and managing the causes of low yield. The variable data collected from semiconductor manufacturing processes, in which hundreds of unit processes are implemented according to specific conditions and sequences, are interdependent, and the variables related to previous processes influence the variables in subsequent processes. Therefore, the order of processes should be considered when building a model that searches for the causes of low yield. However, there have been few studies in this area. This paper proposes a low-yield root cause search method considering the order of processes using a long short-term memory with attention mechanism (LSTM-AM) model. Specifically, the LSTM-AM model is applied to data classified according to the process structure of semiconductor products, and the causes of low yield are determined considering the order of processes by extracting attention weights. Experiments are conducted to verify the suitability of the proposed method using real yield data from a semiconductor company. The experimental results confirm that the proposed method outperforms the existing low yield root cause search methods in terms of low yield prediction.

As the critical dimension of transistors has become lower and the stacked layer of semiconductors has become higher, virtual diagnostics to monitor the status of plasma in an etching process has been important because of the reliability of process. In this study, we proposed the model to predict a plasma density of the etch equipment with high accuracy using OES data despite a small number of process conditions. The proposed model could improve the prediction performance of multilayer perceptron by using pre-trained variational auto-encoder as an initializer and had the best performance of several regression methods. At application point of view, it is expected that the model can be used to monitor a plasma density when a wafer is absent from the chamber. Moreover, using the proposed model can more easily understand the status of plasma rather than monitoring thousands of sensor data even without the knowledge about plasma.

Semiconductor manufacturing is on the cusp of a revolution: the Internet of Things (IoT). With IoT we can connect all the equipment and feed information back to the factory so that quality issues can be detected. In this situation, more and more edge devices are used in wafer inspection equipment. This edge device must have the ability to quickly detect defects. Therefore, how to develop a high-efficiency architecture for automatic defect classification to be suitable for edge devices is the primary task. In this paper, we present a novel architecture that can perform defect classification in a more efficient way. The first function is self-proliferation, using a series of linear transformations to generate more feature maps at a cheaper cost. The second function is self-attention, capturing the long-range dependencies of feature map by the channel-wise and spatial-wise attention mechanism. We named this method as self-proliferation-and-attention neural network. This method has been successfully applied to various defect pattern classification tasks. Compared with other latest methods, SP&A-Net has higher accuracy and lower computation cost in many defect inspection tasks.

A special container is made to detect the liquid contents used in the semiconductor manufacturing process by electrical measurements in both static and dynamic conditions. For static measurements, the capacitance measurement shows the dispersive capacitance resulting from ionic response at low frequency (<2 kHz). However, it has negligible dispersion and is immune from the air exposure at high frequency (≥2 kHz). In addition, the capacitance and the corresponding dielectric constant show good linearity with respect to the IPA concentration at 20 kHz, indicating that it is a sweet spot to determine the liquid contents. The resistivity is independent of the measurement frequency (20 Hz to 200 kHz). However, it is sensitive to the exposure time to the air due to the atmospheric moisture and carbon dioxide dissolved in the liquid. For dynamic measurement, the longer transition time is observed when high-density liquid is flushed with low-density liquid, as compared to the other way around.

The number of variables involved in the formation of Ohmic contacts to SiC is large, and their relationships to the final contact resistance are often unclear. As such, trial-and-error methods are typically employed to develop or improve SiC contacts. In pursuit of a better alternative, we developed and tested several regression models to predict the specific contact resistance of Ni, Ti, and Al based contacts on both n- and p-type SiC. Literature data was used to train linear regression, Gaussian process regression, and neural network (NN) ensemble models; of these, the NN ensemble was the most effective at predicting contact resistances. We then applied the model to optimize the annealing schedule for Ni contacts to n-type 4H-SiC, and Ti/Al contacts to p-type 4H-SiC. Finally, we use the model to generate optimal simultaneous contact recipes.