Abnormality cause specifying method, abnormality cause specifying system, and semiconductor device fabrication method

• US 20060281199A1
• Filed: 06/14/2006
• Published: 12/14/2006
• Est. Priority Date: 06/14/2005
• Status: Active Grant

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Matsushita, Hiroshi
• Granted Patent: US 7,599,817 B2
• Time in Patent Office: Days
• US Class Current: 438/14
• CPC Class Codes: G05B 23/024 Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
• CPC Class Codes: H01L 22/20 Sequence of activities consisting of a plurality of measurements, corrections, marking or sorting steps

Abstract

A feature amount is generated by standardizing inspection data related to a fabrication unit for each type, a similar set including fabrication units corresponding to similar feature amounts is formed by comparing the feature amounts, and apparatus difference analysis is performed between a plurality of fabrication units forming the similar set. A two-level orthogonal table is used to determine whether to adopt a feature amount of each type, and some feature amounts are not used in the apparatus difference analysis and the like by optimizing the smaller-the-better characteristic or the larger-the-better characteristic of a test value of the apparatus difference analysis, thereby reducing the calculation amount and accurately and efficiently specifying an abnormality cause.

Claims

1. An abnormality cause specifying method comprising:

acquiring, for each fabrication unit, a plurality of types of inspection data in a plurality of fabrication steps of a product fabrication process;

generating a feature amount by standardizing the inspection data for each type;

generating, for each fabrication unit, a trial data set by selecting the feature amount corresponding to the inspection data of the type on an adoption level for each trial experiment in a two-level orthogonal table, by using two levels of the two-level orthogonal table as the adoption level and a non-adoption level of the type, and using a factor of the two-level orthogonal table as the type;

calculating, for the trial data set, a trial similarity representing a degree of similarity between the fabrication units;

generating, for each fabrication unit, a trial set by extracting another fabrication unit whose trial similarity is not less than a threshold value;

calculating, for each trial experiment and for each fabrication step, a step test value representing a degree of a causal relation of the fabrication unit forming the trial set to a difference between a plurality of fabrication apparatuses used in the fabrication step;

setting, for each trial experiment, the step test value whose degree of the nominal-the-best characteristic is largest of the trial set as a trial experiment test value;

generating, for each fabrication unit, an optimum data set including the feature amount corresponding to the type for which the adoption level is selected, on the basis of a factorial effect diagram for optimizing the nominal-the-best characteristic of the trial experiment test value by using the adoption level and the non-adoption level of each type;

calculating, for the optimum data set, an optimum similarity representing a degree of similarity between the fabrication units;

generating, for each fabrication unit, an optimum set by extracting another fabrication unit whose optimum similarity is not less than a threshold value;

calculating, for each fabrication step, an optimum test value representing a degree of a causal relation of the fabrication unit forming the optimum set to a difference between a plurality of fabrication apparatuses used in the fabrication step;

determining, for each optimum set, whether the causal relation of the optimum set to the difference between the fabrication apparatuses is significant, on the basis of the optimum test value; and

extracting, for each optimum set, the fabrication apparatus as an object of the causal relation found to be significant, as an abnormality cause.

2. The method according to claim 1, wherein

the inspection data includes at least one of equipment engineering system (EES) data, defect data, QC data, characteristic data, and wafer test data, and

the EES data is a physical amount representing an operating condition of a fabrication apparatus used to fabricate a product, the defect data indicates a result of product abnormality inspection performed before and after the fabrication step, the QC data is one of a film thickness of a thin film added by processing in the fabrication step and a dimension of a shape processed in the fabrication step, the characteristic data is an electrical characteristic obtained by measuring a test element group for electrical characteristic evaluation formed together with the product in the fabrication step, and the wafer test data is obtained by an electrical characteristic test conducted on the product after a wafer process is completed.

3. The method according to claim 1, wherein the feature amount is obtained by calculating an average μ

and a standard deviation σ

for the inspection data, and standardizing inspection data x into a generalized feature amount x′

by

x′

=((x−

μ

)/2σ

)+1  

 

(A)

4. The method according to claim 1, wherein a correlation coefficient between all fabrication units as objects concerning the feature amount of the trial data set is used as the trial similarity.

5. The method according to claim 1, wherein

in the calculation of the step test value, a bias of said plurality of fabrication apparatuses used in the fabrication step of each of the fabrication units forming the trial set is statistically tested, and

in the calculation of the optimum test value, a bias of said plurality of fabrication apparatuses used in the fabrication step of each of the fabrication units forming the optimum set is statistically tested.

6. The method according to claim 1, wherein in the determination of the significance of the causal relation, it is determined that a causal relation of the optimum set to a fabrication apparatus whose optimum test value satisfies a predetermined threshold value is significant, and that a causal relation of the optimum set to a fabrication apparatus whose optimum test value does not satisfy the predetermined threshold value is not significant.

7. An abnormality cause specifying system comprising:

a data acquisition unit which acquires, for each fabrication unit, a plurality of types of inspection data in a plurality of fabrication steps of a product fabrication process;

a feature amount generator which generates a feature amount by standardizing the inspection data for each type;

a trial data set generator which generates, for each fabrication unit, a trial data set by selecting the feature amount corresponding to the inspection data of the type on an adoption level for each trial experiment in a two-level orthogonal table, by using two levels of the two-level orthogonal table as the adoption level and a non-adoption level of the type, and using a factor of the two-level orthogonal table as the type;

a trial similarity calculator which calculates, for the trial data set, a trial similarity representing a degree of similarity between the fabrication units;

a trial set generator which generates, for each fabrication unit, a trial set by extracting another fabrication unit whose trial similarity is not less than a threshold value;

a step test value calculator which calculates, for each trial experiment and for each fabrication step, a step test value representing a degree of a causal relation of the fabrication unit forming the trial set to a difference between a plurality of fabrication apparatuses used in the fabrication step;

a maximum test value setting unit which sets, for each trial experiment, the step test value whose degree of the nominal-the-best characteristic is largest of the trial set as a maximum test value;

an optimum data set generator which generates, for each fabrication unit, an optimum data set including the feature amount corresponding to the type for which the adoption level is selected, on the basis of a factorial effect diagram for optimizing the nominal-the-best characteristic of the maximum test value by using the adoption level and the non-adoption level of each type;

an optimum similarity calculator which calculates, for the optimum data set, an optimum similarity representing a degree of similarity between the fabrication units;

an optimum set generator which generates, for each fabrication unit, an optimum set by extracting another fabrication unit whose optimum similarity is not less than a threshold value;

an optimum test value calculator which calculates, for each fabrication step, an optimum test value representing a degree of a causal relation of the fabrication unit forming the optimum set to a difference between a plurality of fabrication apparatuses used in the fabrication step;

a determination unit which determines, for each optimum set, whether the causal relation of the optimum set to the difference between the fabrication apparatuses is significant, on the basis of the optimum test value; and

an extractor which extracts, for each optimum set, the fabrication apparatus as an object of the causal relation found to be significant, as an abnormality cause.

8. A system according to claim 7, wherein

the inspection data includes at least one of equipment engineering system (EES) data, defect data, QC data, characteristic data, and wafer test data, and

the EES data is a physical amount representing an operating condition of a fabrication apparatus used to fabricate a product, the defect data indicates a result of product abnormality inspection performed before and after the fabrication step, the QC data is one of a film thickness of a thin film added by processing in the fabrication step and a dimension of a shape processed in the fabrication step, the characteristic data is an electrical characteristic obtained by measuring a test element group for electrical characteristic evaluation formed together with the product in the fabrication step, and the wafer test data is obtained by an electrical characteristic test conducted on the product after a wafer process is completed.

9. A system according to claim 7, wherein the feature amount is obtained by calculating an average μ

and a standard deviation σ

for the inspection data, and standardizing inspection data x into a generalized feature amount x′

by

x′

=((x−

μ

)/2σ

)+1  

 

(A)

10. A system according to claim 7, wherein a correlation coefficient between all fabrication units as objects concerning the feature amount of the trial data set is used as the trial similarity.

11. A system according to claim 7, wherein

in the calculation of the step test value, a bias of said plurality of fabrication apparatuses used in the fabrication step of each of the fabrication units forming the trial set is statistically tested, and

in the calculation of the optimum test value, a bias of said plurality of fabrication apparatuses used in the fabrication step of each of the fabrication units forming the optimum set is statistically tested.

12. A system according to claim 7, wherein in the determination of the significance of the causal relation, it is determined that a causal relation of the optimum set to a fabrication apparatus whose optimum test value satisfies a predetermined threshold value is significant, and that a causal relation of the optimum set to a fabrication apparatus whose optimum test value does not satisfy the predetermined threshold value is not significant.

13. A semiconductor device fabrication method comprising:

fabricating a semiconductor device as a product;

performing an abnormality cause specifying method for the semiconductor device, the abnormality cause specifying method comprising

acquiring, for each fabrication unit, a plurality of types of inspection data in a plurality of fabrication steps of a product fabrication process,

generating a feature amount by standardizing the inspection data for each type,

generating, for each fabrication unit, a trial data set by selecting the feature amount corresponding to the inspection data of the type on an adoption level for each trial experiment in a two-level orthogonal table, by using two levels of the two-level orthogonal table as the adoption level and a non-adoption level of the type, and using a factor of the two-level orthogonal table as the type,

calculating, for the trial data set, a trial similarity representing a degree of similarity between the fabrication units,

generating, for each fabrication unit, a trial set by extracting another fabrication unit whose trial similarity is not less than a threshold value,

calculating, for each trial experiment and for each fabrication step, a step test value representing a degree of a causal relation of the fabrication unit forming the trial set to a difference between a plurality of fabrication apparatuses used in the fabrication step,

setting, for each trial experiment, the step test value whose degree of the nominal-the-best characteristic is largest of the trial set as a trial experiment test value,

generating, for each fabrication unit, an optimum data set including the feature amount corresponding to the type for which the adoption level is selected, on the basis of a factorial effect diagram for optimizing the nominal-the-best characteristic of the trial experiment test value by using the adoption level and the non-adoption level of each type,

calculating, for the optimum data set, an optimum similarity representing a degree of similarity between the fabrication units,

generating, for each fabrication unit, an optimum set by extracting another fabrication unit whose optimum similarity is not less than a threshold value,

calculating, for each fabrication step, an optimum test value representing a degree of a causal relation of the fabrication unit forming the optimum set to a difference between a plurality of fabrication apparatuses used in the fabrication step,

determining, for each optimum set, whether the causal relation of the optimum set to the difference between the fabrication apparatuses is significant, on the basis of the optimum test value, and

extracting, for each optimum set, the fabrication apparatus as an object of the causal relation found to be significant, as an abnormality cause;

extracting, for each optimum set, the inspection data related to the fabrication step in which the fabrication apparatus as the abnormality cause is used, as abnormality data; and

adjusting the fabrication apparatus extracted as the abnormality cause such that the abnormality data is corrected.

14. The method according to claim 13, wherein

the inspection data includes at least one of equipment engineering system (EES) data, defect data, QC data, characteristic data, and wafer test data, and

the EES data is a physical amount representing an operating condition of a fabrication apparatus used to fabricate a product, the defect data indicates a result of product abnormality inspection performed before and after the fabrication step, the QC data is one of a film thickness of a thin film added by processing in the fabrication step and a dimension of a shape processed in the fabrication step, the characteristic data is an electrical characteristic obtained by measuring a test element group for electrical characteristic evaluation formed together with the product in the fabrication step, and the wafer test data is obtained by an electrical characteristic test conducted on the product after a wafer process is completed.

15. The method according to claim 13, wherein the feature amount is obtained by calculating an average μ

and a standard deviation σ

for the inspection data, and standardizing inspection data x into a generalized feature amount x′

by

x′

=((x−

μ

)/2σ

)+1  

 

(A)

16. The method according to claim 13, wherein a correlation coefficient between all fabrication units as objects concerning the feature amount of the trial data set is used as the trial similarity.

17. The method according to claim 13, wherein

in the calculation of the step test value, a bias of said plurality of fabrication apparatuses used in the fabrication step of each of the fabrication units forming the trial set is statistically tested, and

in the calculation of the optimum test value, a bias of said plurality of fabrication apparatuses used in the fabrication step of each of the fabrication units forming the optimum set is statistically tested.

18. The method according to claim 13, wherein in the determination of the significance of the causal relation, it is determined that a causal relation of the optimum set to a fabrication apparatus whose optimum test value satisfies a predetermined threshold value is significant, and that a causal relation of the optimum set to a fabrication apparatus whose optimum test value does not satisfy the predetermined threshold value is not significant.