Method and system for predicting process performance using material processing tool and sensor data

1. A method for constructing a process performance prediction model for a material processing system, the method comprising the steps of:

recording tool data for a plurality of observations during a process in a process tool, said tool data comprises a plurality of tool data parameters;

recording process performance data for said plurality of observations during said process in said process tool, said process performance data comprises one or more process performance parameters;

performing a partial least squares analysis using said tool data and said process performance data; and

computing correlation data from said partial least squares analysis,

wherein said process performance data comprises at least one of mean etch depth and etch depth range.

2. The method for constructing a process performance prediction model as recited in claim 1 , wherein said step of recording tool data for a plurality of observations includes the step of forming a tool data matrix.

3. The method for constructing a process performance prediction model as recited in claim 1 , wherein said step of recording process performance data for said plurality of observations includes the step of forming a process performance data matrix.

4. The method for constructing a process performance prediction model as recited in claim 1 , wherein said correlation data comprises a correlation matrix.

5. A method for constructing a process performance prediction model for a material processing system, the method comprising the steps of:

recording tool data for a plurality of observations during a process in a process tool, said tool data comprises a plurality of tool data parameters;

recording process performance data for said plurality of observations during said process in said process tool, said process performance data comprises one or more process performance parameters;

performing a partial least squares analysis using said tool data and said process performance data; and

computing correlation data from said partial least squares analysis,

wherein said tool data comprises at least one of a capacitor position, a forward radio frequency (RF) power, a reflected RF power, a voltage, a current, a phase, an impedance, a RF peak-to-peak voltage, a RF self-induced direct current bias, a chamber pressure, a gas flow rate, a temperature, a backside gas pressure, a backside gas flow rate, an electrostatic clamp voltage, an electrostatic clamp current, a focus ring thickness, RF hours, and focus ring RF hours.

6. A method for constructing a process performance prediction model for a material processing system, the method comprising the steps of:

recording tool data for a plurality of observations during a process in a process tool, said tool data comprises a plurality of tool data parameters;

recording process performance data for said plurality of observations during said process in said process tool, said process performance data comprises one or more process performance parameters;

performing a partial least squares analysis using said tool data and said process performance data; and

computing correlation data from said partial least squares analysis,

wherein said step of performing a partial least squares analysis includes,

calculating variable importance in projection data defined as an influence on said process performance data of said tool data, and

refining said tool data based upon an analysis of the variable importance in projection data.

7. The method for constructing a process performance prediction model as recited in claim 6 , wherein said step of performing a partial least squares analysis includes the step of determining a minimum number of partial least squares components that substantially explain a correlation between said tool data and said process performance data.

8. The method for constructing a process performance prediction model as recited in claim 6 , wherein said step of performing a partial least squares analysis includes the step of verifying the fitting power of said process performance prediction model by inspecting at least one of a fraction of a sum of squares of all of said tool data and said process performance data explained by the partial least squares component, a fraction of a variance of all of said tool data and said process performance data explained by the partial least squares component, a cumulative sum of squares of all of said tool data and said process performance data explained by the partial least squares component, and a cumulative variance of all of said tool data and said process performance data explained by the partial least squares component.

9. The method for constructing a process performance prediction model as recited in claim 6 , wherein said step of performing a partial least squares analysis includes the step of verifying the predictive power of said process performance prediction model by performing cross validation.

10. A method for detecting a fault in a material processing system using a process performance prediction model, the method comprising the steps of:

initiating a process in a process tool of said material processing system;

recording tool data for at least one observation during said process in said process tool to form a tool data matrix, said tool data comprises a plurality of tool data parameters;

performing a matrix multiplication of said tool data matrix and a correlation matrix to form predicted process performance data, said correlation matrix comprising said process performance prediction model;

comparing said predicted process performance data with target process performance data; and

determining a fault condition of said material processing system from said comparing step,

wherein said step of comparing said predicted process performance data with target process performance data comprises determining difference data between said predicted process performance data and said target process performance data, and

said step of determining a fault condition of said material processing system comprises comparing said difference data with threshold difference data.

11. The method for detecting a fault in a material processing system as recited in claim 10 , wherein a fault in said material processing system is detected when said difference data exceeds said threshold difference data.

12. A method for detecting a fault in a material processing system, the method comprising the steps of:

recording first tool data for a plurality of observations during a first process in a process tool to form a first tool data matrix, said first tool data comprises a plurality of tool data parameters;

recording first process performance data for said plurality of observations during said first process in said process tool to form a first process performance data matrix, said first process performance data comprises one or more process performance parameters;

performing a partial least squares analysis using said first tool data matrix and said first process performance data matrix;

computing a correlation matrix from said partial least squares analysis, said correlation matrix comprises a process performance prediction model;

performing a matrix multiplication of said first tool data matrix and said correlation matrix to form target process performance data;

initiating a second process in said process tool of said material processing system;

recording second tool data for at least one observation during said second process in said process tool to form a second tool data matrix, said second tool data matrix comprises said plurality of tool data parameters;

performing a matrix multiplication of said second tool data matrix and said correlation matrix to form predicted process performance data;

comparing said predicted process performance data with target process performance data; and

determining a fault condition of said material processing system from said comparing step,

wherein said step of comparing said predicted process performance data with a target process performance data comprises determining difference data between said predicted process performance data and said target process performance data, and

said step of determining a fault condition of said material processing system comprises comparing said difference data with threshold difference data.

13. The method for detecting a fault in a material processing system as recited in claim 12 , wherein said first process performance data and said predicted process performance data comprises at least one of mean etch depth and etch depth range.

14. The method for detecting a fault in a material processing system as recited in claim 12 , wherein said first tool data and said second tool data comprises at least one of a capacitor position, a forward RF power, a reflected RF power, a voltage, a current, a phase, an impedance, a RF peak-to-peak voltage, a RF self-induced DC bias, a chamber pressure, a gas flow rate, a temperature, a backside gas pressure, a backside gas flow rate, an electrostatic clamp voltage, an electrostatic clamp current, a focus ring thickness, RF hours, and focus ring RF hours.

15. The method for detecting a fault in a material processing system as recited in claim 12 , wherein said step of performing a partial least squares analysis includes the step of determining a minimum number of partial least squares components that substantially explains a correlation between said first tool data and said first process performance data.

16. The method for detecting a fault in a material processing system as recited in claim 12 , wherein said step of performing a partial least squares analysis includes the steps of:

calculating variable importance in projection data defined as an influence on said first process performance data matrix of said first tool data matrix; and

refining said first tool data matrix based upon an analysis of the variable importance in projection data.

17. The method for detecting a fault in a material processing system as recited in claim 12 , wherein said step of performing a partial least squares analysis includes the step of verifying the fitting power of said process performance prediction model by inspecting at least one of a fraction of a sum of squares of all of said first tool data and said first process performance data explained by a partial least squares component, a fraction of a variance of all of said first tool data and said first process performance data explained by a partial least squares component, a cumulative sum of squares of all of said first tool data and said first process performance data explained by a partial least squares component, and a cumulative variance of all of said first tool data and said first process performance data explained by a partial least squares component.

18. The method for detecting a fault in a material processing system as recited in claim 12 , wherein said step of performing a partial least squares analysis includes the step of verifying the predictive power of said process performance prediction model by performing cross validation.

19. The method for detecting a fault in a material processing system as recited in claim 12 , wherein a fault in said material processing system is detected when said difference data exceeds said threshold difference data.