Fault detection and classification (FDC) using a run-to-run controller

A computer-implemented method of operating a semiconductor processing system comprising:

• positioning a first wafer of a plurality of wafers in a lot in a process module;
• receiving process context information for the first wafer at a processor;
• executing a control strategy on the processor using the process context information;
• executing an analysis strategy on the processor using process context information for the first wafer;
• calculating a first trim amount for the first wafer using a run-to-run (R2R) control model, wherein a first calculated trim amount comprises a vertical trim value, or a horizontal trim value, or a combination thereof;
• computing a first set of process parameters to achieve the first calculated trim amount for the first wafer using an R2R controller;
• executing a first process recipe using the first set of process parameters in the process module to process the first wafer;
• calculating a predicted trim amount for the first wafer using an FDC model, the first process recipe and chamber and wafer state data;
• determining an actual trim amount for the first wafer;
• comparing the actual trim amount for only the first wafer to the predicted trim amount for only the first wafer;
• creating an FDC alarm when the actual trim amount for the first wafer is substantially greater than the predicted trim amount for the first wafer;
• executing an intervention plan on the processor when an a first alarm for the first wafer has been established by at least one executed strategy;
• and
• removing the first wafer from the process module when an alarm condition has not been established by at least one executed plan, wherein the FDC model is distinct from the R2R control model.

A computer-directed system for implementing Fault Detection and Classification (FDC) in an. Advanced Process Control (APC) system including a run-to-run (R2R) controller comprising:

• a processing tool configured to process a wafer;
• and
• a processor configured to,
• receive process data from a first executed process run,
• receive a process recipe comprising a set of recipe parameters from the R2R controller based on an R2R control model, the R2R control model distinct from an FDC model,
• receive chamber and wafer state data,
• receive the FDC model from memory, the FDC model based on normal tool level parameters,
• compute a vector of predicted dependent process parameters using the FDC model, the chamber and wafer state data, and the process recipe,
• execute the process recipe for only a first wafer of a plurality of wafers in a lot to produce a first vector of measured dependent process parameters,
• calculate a first difference for the first wafer using the vector of predicted dependent process parameters and the first vector of measured dependent process parameters for the first wafer,
• compare the first difference for the first wafer to a threshold value,
• declare an alarm condition when the first difference for the first wafer is greater than the threshold value, and
• display alarm condition information on a GUI screen.

A computer storage medium containing a software routine that, when executed, causes a general purpose computer to control an apparatus using a processing method, comprising:

• receiving an FDC model from memory, the FDC model based on normal tool level parameters;
• providing the FDC model to a process model calculation engine;
• receiving a process recipe comprising a set of recipe parameters from the R2R controller based on an R2R control model, the R2R control model distinct from the FDC model;
• receiving chamber and wafer state data;
• computing a vector of predicted dependent process parameters using the process model calculation engine based on the provided FDC model the chamber state and wafer state data and the process recipe;
• providing the process recipe to a process module;
• executing the process recipe using only a first wafer of a plurality of wafers in a lot to produce a first vector of measured dependent process parameters;
• calculating a first difference for the first wafer using the vector of predicted dependent process parameters and the first vector of measured dependent process parameters for the first wafer;
• comparing the first difference for the first wafer to a threshold value;
• declaring an alarm condition when the first difference for the first wafer is greater than the threshold value;
• and
• displaying alarm condition information on a GUI screen.