Determining process condition in substrate processing module

• US 8,046,193 B2
• Filed: 04/21/2008
• Issued: 10/25/2011
• Est. Priority Date: 07/07/2004
• Status: Active Grant

• Current Assignee: KLA-Tencor Corporation
• Original Assignee: KLA-Tencor Corporation
• Inventors: Parker, Jeffrey M., Yetter, Forrest Gilbert Jr., Pieper, John B., Renken, Wayne G.
• Primary Examiner(s): WACHSMAN, HAL D
• Application Number:
  US12/107,007
• Publication Number:
  US 20080228306A1
• Time in Patent Office: 1,282 Days
• Field of Search: 702/183, 702/182, 702/84
• US Class Current: 702/183
• CPC Class Codes: G05B 23/0254 based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks

Abstract

A sensor network collects time-series data from a process tool and supplies the data to an analysis system where pattern analysis techniques are used to identify structures and to monitor subsequent data based on analysis instructions or a composite model. Time-series data from multiple process runs are used to form a composite model of a data structure including variation. Comparison with the composite model gives an indication of tool health. A sensor network may have distributed memory for a more simplified configuration.

Claims

1. A method for determining a condition of a process in a substrate processing module, comprising:

obtaining a first set of time-series data from measurement of a process parameter during a first occurrence of the process in the substrate processing module;

obtaining at least a second set of time-series data from measurement of the process parameter during at least a second occurrence of the process in the substrate processing module;

identifying a structure corresponding to a first portion of the first set of time-series data, the structure selected from a finite set of structures;

identifying a first portion in each set of the at least a second set of time series data that corresponds to the structure;

deriving a composite model from the first portion of the first set of time-series data and the first portion of each set of the at least a second set of time series data;

obtaining a third set of time-series data from measurement of the process parameter during a third occurrence of the process in the substrate processing module;

determining a degree of correlation between a first portion of the third set of timeseries data and the composite model; and

determining a condition of the process based on the degree of correlation.

2. The method of claim 1 wherein the first occurrence of the process is the processing of a semiconductor wafer that is subsequently processed to form semiconductor devices.

3. The method of claim 1 wherein the first occurrence of the process is the processing of a glass substrate that is subsequently processed to form a flat panel display.

4. The method of claim 1 wherein the degree of correlation is determined from comparison of the first portion of the third set of time-series data and the model, the model including a variation between the first portion of the first set of time-series data and additional sets of time-series data.

5. The method of claim 1 wherein the first set of time-series data and the at least a second set of time-series data are collected using one or more sensors mounted to a process chamber, the one or more sensors in communication with a controller over a network.

6. The method of claim 1 further comprising determining the locations of the boundaries of the portion of the third set of time-series data, the locations determined by modifying the locations to obtain good correlation with structures on both sides of each boundary.

7. The method of claim 6 wherein the composite model for a process of variable length has a range of acceptable time between boundaries.

8. The method of claim 1 wherein identifying the structure includes selecting a polynomial having an order from a plurality of polynomials of different orders, selecting the polynomial based on goodness of fit and the order so that the polynomial selected is not the polynomial that gives the best fit.

9. The method of claim 8 wherein individual polynomials of the plurality of polynomials are shifted Legendre polynomials.

10. The method of claim 1 further comprising deriving additional composite models from additional portions of the first set of time-series data and additional portions of each set of the at least a second set of time series data and determining additional degrees of correlation between additional portions of the third set of time-series data and the additional composite models.

11. The method of claim 1 wherein the degree of correlation between the first portion of the third set of time-series data and the composite model is determined by calculating a scalar Mahalanobis distance between a first polynomial that represents the third set of time-series data and a second polynomial that represents the composite model and dividing the scalar Mahalanobis distance by the number of terms in the first polynomial, or by a value derived from the first polynomial, to obtain a scaled Mahalanobis distance that is compared to a threshold distance.