Hidetaka Tsuda

Yield improvement using data mining system

It is ideal to prevent all failures. However, when a failure occurs, it is important to quickly specify the cause stage and take countermeasures. There are various types of failures, ranging from the failures due to simple mis-operation to the failures whose cause analysis takes many highly skilled engineers a long time. If the failure cause in the latter case can be specified simply by anyone, the yield enhancement will be accelerated. We are developing a method that enables us to specify a failure cause, without depending on the experience and skills of engineers. Data mining is a method for extracting buried information and rules from data of enormous quantity, by using a statistical method. Some examples have been reported in various fields but only a few in the semiconductor field. This time, we have applied a regression tree analysis system, which is one of data mining tools codeveloped by Fujitsu Laboratories Ltd. and FLT, to failure analysis in LSI manufacturing. As a result, a failure cause which has been difficult to be detected even in the in-process monitoring was specified automatically only in six hours. Then, through the verification process, we ascertained that the failure cause was correct. We could specify the cause and take countermeasures at a speed six times faster than by the conventional method.

Yield analysis and improvement by reducing manufacturing fluctuation noise

There are various types of failures and their causes intertwined one another complicatedly. Therefore, we need to recognize timely what any kind of failure causes affect to total yield quantitatively to decide the countermeasures for yield improvement. Collected data for data analysis involve the noise with various influences. For example, analyzing all collected data always does not bring us precious correlation between the collected data. We prefer to adapt a new method to get more precious information. We have got many success stories for yield improvement by data mining which is one of the statistical methods, now we have succeeded in developing a method to clarify the correlation between yield and various wafer parametrical data value, because the data can be extracted to reduce the influence of the manufacturing fluctuation automatically.

Improvement of Photolithography Process by Second Generation Data Mining

The advanced process control (APC) system has been developed. The APC system has already been introduced regarding critical dimension (CD) and overlay controls in a photolithography process. It has improved the productivity and device performance. However, the current APC is based on the inspection data where process deviation is mingled with machine fluctuation and which has a very small quantity to be analyzed, then it has the limit in the effect. We have collected and stored the CD and overlay inspection data as well as the log data of the exposure tool in a relational database. So, we have investigated the method to compensate and solve the above-mentioned problem. First, we have extracted relationships between inspection data and many equipment parameters, especially correlation coefficients, in huge tool log data. Next, we have investigated the issues with significant relationships and have consequently extracted useful information not extracted by the conventional method. The purpose of this paper is to show that we have developed a second generation data mining system in cooperation with APC to prove the effect of stabilizing machine fluctuation.

Log data extraction and correlation miner for lithography management system : LMS-LEC

To attain quick turn-around time (TAT) and high yield, it is very important to remove all the problems affecting the semiconductor volume production line. For this purpose, we have used a lithography management system (LMS) as an advanced process control system. The LMS stores the critical dimension and overlay inspection results as well as the log data of the exposure tool in a relational database. This enables a quick and efficient grasp of the productivity under the present conditions and helps to identify the causes of errors. Furthermore, we developed a mining tool, called a log data extraction and correlation miner (LMS-LEC), for factor analysis on the LMS. Despite low correlation between all data, a high correlation may exist between parameters in a certain data domain. The LMS-LEC can mine such correlations easily. With this tool, we can discover previously unknown error sources that have been buried in the vast quantity of data handled by the LMS and thereby increase of the effectiveness of the exposure and inspection tool. The LMS-LEC is an extremely useful software mining tool for “equipment health” monitoring, advanced fault detection, and sophisticated data analysis.