Semiconductor yield management system and method

Abstract

A system and method for yield management are disclosed wherein a data set containing one or more prediction variable values and one or more response variable values is input into the system. The system can process the input data set to remove prediction variables with missing values and data sets with missing values based on a tiered splitting method to maximize usage of all valid data points. The processed data can then be used to generate a model that may be a decision tree. The system can accept user input to modify the generated model. Once the model is complete, one or more statistical analysis tools can be used to analyze the data and generate a list of the key yield factors for the particular data set.

Description

BACKGROUND OF THE INVENTION [0001] 1. Field of the Invention [0002] The present invention relates generally to a system and method for managing a semiconductor manufacturing process and, more particularly, to a system and method for managing yield in a semiconductor fabrication process. [0003] 2. Description of the Prior Art [0004] The semiconductor manufacturing industry is continually evolving its fabrication processes and developing new processes to produce smaller and smaller geometries of the semiconductor devices being manufactured, because smaller devices typically generate less heat and operate at higher speeds than larger devices. Currently, a single integrated circuit chip may contain over one billion patterns. Consequently, semiconductor fabrication processes are extremely complex, and hundreds of processing steps may be involved. The occurrence of a mistake or small error at any of the process steps or tool specifications may cause lower yield in the final semiconductor ...

AI Technical Summary

Benefits of technology

[0011] One embodiment of the yield management system and method in accordance with the present invention provides many advantages over conventional yield management systems and techniques, which make the yield management system and method in accordance with the present invention more useful to semiconductor manufacturers. The system may be fully automated and is easy to use, so that no extra training is necessary to make use of the yield management system. In addition, the yield management system handles both continuous and categorical variables. The system also automatically handles missing data during a processing step that is optimized to consider data for all significant yield parameters. The system can rapidly search through hundreds of yield parameters and generate an output indicating the one or more key yield factors / parameters. The system generates an output preferably in the form of a decision tree that is easy to interpret and understand. The system may employ advanced splitting rules to parse the data and is also very flexible in that it permits prior yield parameter knowledge from one or more users to be easily incorporated into the building of the model. Unlike conventional yield management systems, if there is more than one yield factor / parameter affecting the yield of the process, the system can identify all of the parameters / factors simultaneously, so that the multiple factors / parameters are identified during a single pass through the yield data.

[0012] In accordance with various embodiments of the present invention, the yield management system and method may receive a yield data set. When an input data set is received, one embodiment of the yield management system and method in accordance with the present invention first performs a data processing step in which the validity of the data in the data set is checked, and cases or parameters with missing data are identified. One embodiment of the semiconductor yield management system and method in accordance with the present invention provides a tiered splitting method to maximize usage of all valid data points. Another embodiment of the yield management system and method in accordance with the present invention provides an outlier filtering method. Also, in accordance with various other embodiments of the yield management system and method of the present invention, a user can select from among 1) add tool usage parameters, 2) treat an integer as categorical, and 3) auto-categorize methods for better data manipulation capability and flexibility.

[0013] The semiconductor yield management system and method in accordance with one embodiment of the present invention also preferably provide a linear type split and a range type split for use in constructing the model when the response variable and the prediction variable have a linear relationship, in order to overcome the shortcoming of a binary decision tree that has to split on the prediction variable several times on different levels and does not necessarily show that the relationship is linear. The semiconductor yield management system and method in accordance with various embodiments of the present invention also provide user control in formulating the rules for splitting nodes, so that the user may assure that more appropriate and accurate models are generated. Preferably, the user selectable split methods include: 1) consider tool and date parameters jointly; 2) consider tool and event parameters jointly; 3) maximize class distinction; 4) prefer simple splits; 5) minimum purity; 6) parameter weighting; 7) minimum group size; 8) maximum number of descendants; and 9) raw data mapping.

[0014] Additionally, if the prediction variable is categorical, one embodiment of the yield management system and method in accordance with the present invention enables the user to select any combination of classes of the variable and include them in one sub-node of the decision tree. The remainder of the data is included in the other sub-node. On the other hand, if the prediction variable is continuous, there are preferably three types of split formats from which the user may select. The available split formats are 1) a default type (a≦X), 2) a range type (a1≦X<a2), and 3) a linear type (X<a1, X in [a1, a2], X in [a2, a3], X>a3). These different split formats facilitate the user being able to produce an accurate model.

[0017] Another embodiment of the yield management system and method in accordance with the present invention additionally enables the user to invoke a method to redisplay the setup window and quickly modify his or her previous selections, so that the model may be adjusted. Finally, the yield management system and method in accordance with another embodiment of the present invention enable the user to invoke methods to collapse / expand a node to collapse the node when the user decides that the split of the node is unnecessary or, alternatively, to expand the node when the user wants to examine the aggregate statistics of the entire subset. The method to expand a node may also be invoked by the user to expand a previously collapsed node, so that the node returns to its original length.

[0018] After the model has been modified, the data set may be processed using various statistical analysis tools to help the user better understand the relationship between the prediction and response variables. The yield management system and method in accordance with the present invention provide a yield management tool that is much more powerful and flexible than conventional tools.

Problems solved by technology

Consequently, semiconductor fabrication processes are extremely complex, and hundreds of processing steps may be involved.

The occurrence of a mistake or small error at any of the process steps or tool specifications may cause lower yield in the final semiconductor product, where yield may be defined as the number of functional devices produced by the process as compared to the theoretical number of devices that could be produced assuming no bad devices.

In particular, the database technology has far outpaced the yield management analysis capability when using conventional statistical techniques to interpret and relate yield to major yield factors.

Many conventional yield management systems have a number of limitations and disadvantages which make them undesirable to the semiconductor manufacturing industry.

For example, conventional systems may require some manual processing which slows the analysis and makes the system susceptible to human error.

In addition, these conventional systems may not handle both continuous (e.g., temperature) and categorical (e.g., Lot 1, Lot 2, etc.) yield management variables.

Some conventional systems cannot handle missing data elements and do not permit rapid searching through hundreds of yield parameters to identify key yield factors.

Some conventional systems output data that is difficult to understand or interpret even by knowledgeable semiconductor yield management personnel.

In addition, conventional systems typically process each yield parameter separately, which is time consuming and cumbersome and cannot identify more than one parameter at a time.

While the yield management system and technique disclosed in aforementioned U.S. Pat. No. 6,470,229 B1 provide a powerful yield management tool, one limitation is that the criteria employed for processing data sets may remove data sets with missing values, even though the data sets may contain usable data respecting a significant prediction variable that may be useful in generating the model.

Also, while the disclosed system and technique provide fundamental splitting rules for generating a decision-tree based model, there are instances in which the system is limited in the variety of splitting rules and also limited in accommodating modification of the model based on the knowledge of the user.

Images