US 7,003,366 B1Granted

Diagnostic system and operating method for the same

AssigneeProMOS Technologies Inc.

InventorHung-Wen Chiou

Life of this patent

Estimating the term…

Loading patent…

Abstract

An operating method for a fault detection of a semiconductor process and a diagnostic system for fault detection in a semiconductor process are described. By using the method and the diagnostic system, the real-time process parameters collected during the process is performed by the tool become meaningful and are correlated with the historic process performance data obtained by the post process metrology process. Moreover, the method and the diagnostic system further provide an alarm index for the process performed on the tool to actually reflect the process environment during the process is performed after correlating the real-time process parameters and the historic process performance data. With referring to the alarm index, the current process performance under the real-time process parameters in the tool can be accurately diagnosed.

Claims18

Dependent claims

A diagnostic system for a semiconductor process, comprising:

means for receiving a set of variables including a tool information of a first tool, a process information with respect to a first process performed by the first tool and at least a set of real-time process parameters of the first tool obtained during the first process is performed by the first tool;
means for storing a knowledge database comprising a plurality of sets of historic process performance data, a plurality of sets of historic process parameters, a plurality of sets of historic process parameter membership functions and a plurality of sets of process performance membership functions;
means for retrieving a portion of the historic process parameter sets from the knowledge database with reference to the received tool information and the received process information;
means for approximately matching the received set of real-time process parameters with the retrieved sets of the historic process parameters to trace a candidate set of the historic process performance data corresponding to the matched set of the historic process parameters;
means for fuzzifying each of the received real-time process parameters by using the corresponding sets of the process parameter membership functions to correspondingly generate a set of process parameter fuzzy values;
means for applying a set of fuzzy rules corresponding to the first process to the set of process parameter fuzzy values with the candidate set of the historic process performance data and the corresponding sets of the process performance membership functions to generate a plurality of sets of metrology fuzzy sets;
and
means for defuzzifying the sets of metrology fuzzy sets to generate an alarm index.

5 more elements

A diagnostic method for a first process performed by a first tool with at least a set of real-time process parameters of the first tool obtained during the first process is performed on a first wafer by the first tool, the diagnostic method comprising:

providing a set of variables including a tool information of the first tool, a process information with respect to the first process and the set of real-time process parameters to a diagnostic system having a knowledge database, wherein the knowledge database comprises a plurality of sets of historic process performance data, a plurality of sets of historic process parameters, a plurality of sets of process parameter membership functions and a plurality of sets of process performance membership functions;
approximately matching the provided set of real-time process parameters with the sets of the historic process parameters in the knowledge database to trace a candidate set of the historic process performance data corresponding to the matched set of the historic process parameters;
fuzzifying each of the provided real-time process parameters by using the corresponding sets of the process parameter membership functions to correspondingly generate a set of process parameter fuzzy values;
applying a set of fuzzy rules corresponding to the first process to the set of the process parameter fuzzy values with the candidate set of the historic process performance data and the corresponding sets of the process performance membership functions to generate a plurality of sets of metrology fuzzy sets;
and
defuzzifying the sets of metrology fuzzy sets to generate an alarm index.

3 more elements

Dependencies are inferred from the claim text.