Predictive multi-stage modelling for complex process control

Patent metadata

Field	Value
Publication number	US-11568101-B2
Application number	US-201916539382-A
Country	US
Kind code	B2
Filing date	Aug 13, 2019
Priority date	Aug 13, 2019
Publication date	Jan 31, 2023
Grant date	Jan 31, 2023

Abstract

Predictive multi-stage modelling for complex semiconductor device manufacturing process control is provided. In one aspect, a method of predictive multi-stage modelling for controlling a complex semiconductor device manufacturing process includes: collecting geometrical data from metrology measurements made at select stages of the manufacturing process; and making an outcome probability prediction at each of the select stages using a multiplicative kernel Gaussian process, wherein the outcome probability prediction is a function of a current stage and all prior stages. Machine-learning models can be trained for each of the select stages of the manufacturing process using the multiplicative kernel Gaussian process. The machine-learning models can be used to provide probabilistic predictions for a final outcome in real-time for production wafers. The probabilistic predictions can then be used to select production wafers for rework, sort, scrap or disposition.

First claim

What is claimed is:

1. A method of predictive multi-stage modelling for controlling a complex semiconductor device manufacturing process, the method comprising the steps of:
collecting geometrical data from metrology measurements made of production wafers at select stages of the manufacturing process;
making an outcome probability prediction as to whether an output of the manufacturing process is within specifications at each of the select stages of the manufacturing process by machine-learning models trained for each of the select stages of the manufacturing process using a multiplicative kernel Gaussian process, wherein the outcome probability prediction is a function of a current stage and all prior stages; and
using the outcome probability prediction from the machine-learning models to identify those of the production wafers needing to be reworked or scrapped which are then removed from production, wherein the method further comprises the steps of:
using the machine-learning models for each of the select stages of the manufacturing process to provide probabilistic predictions for a final outcome in real-time for the production wafers; and
using the probabilistic predictions to select the production wafers for rework, sort, scrap or disposition.

2. The method of claim 1 , wherein the manufacturing process comprises a self-aligned quadruple patterning (SAQP) process performed during fin field-effect transistor (finFET) fabrication.

3. The method of claim 2 , wherein the geometrical data is selected from the group consisting of: mandrel height, mandrel top critical dimension (CD), mandrel bottom CD, mandrel sidewall angle, nitride thickness, spacer thickness, spacer bottom thickness, fin space widths, and combinations thereof.

4. The method of claim 1 , wherein the step of making the outcome probability prediction, further comprises the step of:
computing a predictive distribution p(y*|X* 1:j ,X 1:j ,Y)=N(y*|μ j ,λ j ), for j=1, . . . , M for each of the select stages of the manufacturing process, wherein expectation μ and precision λ are computed by a multiplicative kernel s j (·,·),
wherein μ j =S j (X* 1:j ,X 1:j )[S j (X 1:j )+σ j I] −1 Y,
wherein λ j =s j (X* 1:j ,X* 1:j )−S j (X* 1:j ,X 1:j )[S j (X 1:j ,X 1:j )+σ j I] −1 S j T (X* 1:j ,X 1:j ), and
wherein S j denotes a matrix of the multiplicative kernel s j (·,·).

5. The method of claim 4 , further comprising the step of:
computing the multiplicative kernel s j (·,·) as s j (X* 1:j ,X* 1:j )=s j−1 (X* 1:j−1 ,X* 1:j−1 )k j (x* j ,x* j ),
wherein k j (x* j ,x* j )=v j exp[−x* j T L j x* j ].

6. The method of claim 1 , further comprising the steps of:
defining geometric parameters specific to each stage of the manufacturing process;
obtaining data values for the geometric parameters from measurements made of sample wafers at each stage of the manufacturing process; and
removing data values having an incomplete set of measurements across the select stages.

7. The method of claim 6 , wherein the measurements are made of the sample wafers using a semiconductor fabrication metrology tool selected from the group consisting of: scanning electron micrograph (SEM) imaging, thin film measurements, overlay measurements, optical critical dimension measurements, scatterometry measurements, and combinations thereof.

8. A method of predictive multi-stage modelling for controlling a complex semiconductor device manufacturing process, the method comprising the steps of:
collecting geometrical data from metrology measurements made of production wafers at select stages of the manufacturing process;
training machine-learning models for each of the select stages of the manufacturing process using a multiplicative kernel Gaussian process;
using the machine-learning models for each of the select stages of the manufacturing process to provide probabilistic predictions for a final outcome in real-time for the production wafers, wherein the probabilistic predictions are a function of a current stage and all prior stages; and
using the probabilistic predictions to identify those of the production wafers needing to be reworked or scrapped which are then removed from production.

9. The method of claim 8 , wherein the manufacturing process comprises a SAQP process performed during finFET fabrication.

10. The method of claim 9 , wherein the geometrical data is selected from the group consisting of: mandrel height, mandrel top CD, mandrel bottom CD, mandrel sidewall angle, nitride thickness, spacer thickness, spacer bottom thickness, fin space widths, and combinations thereof.

11. The method of claim 8 , further comprising the step of:
computing a predictive distribution p(y*|X* 1:j ,X 1:j ,Y)=N(y*|μ j ,λ j ) for j=1, . . . , M for each of the select stages of the manufacturing process, wherein expectation μ and precision λ are computed by a multiplicative kernel s j (·,·),
wherein μ j =S j (X* 1:j ,X 1:j )[S j (X 1:j )+σ j I] −1 Y,
wherein λ j =s j (X* 1:j ,X* 1:j )−S j (X* 1:j ,X 1:j )[S j (X 1:j ,X 1:j )+σ j I] −1 S j T (X* 1:j ,X 1:j ), and
wherein S j denotes a matrix of the multiplicative kernel s j (·,·).

12. The method of claim 11 , further comprising the step of:
computing the multiplicative kernel s j (·,·) as s j (X* 1:j ,X* 1:j )=s j−1 (X* 1:j−1 ,X* 1:j−1 )k j (x* j ,x* j ),
wherein k j (x* j ,x* j )=v j exp[−x* j T L j x* j ].

13. The method of claim 8 , further comprising the steps of:
defining geometric parameters specific to each stage of the manufacturing process;
obtaining data values for the geometric parameters from measurements made of sample wafers at each stage of the manufacturing process; and
removing data values having an incomplete set of measurements across the select stages.

14. The method of claim 13 , wherein the measurements are made of the sample wafers using a semiconductor fabrication metrology tool selected from the group consisting of: SEM imaging, thin film measurements, overlay measurements, optical critical dimension measurements, scatterometry measurements, and combinations thereof.

15. A computer program product of predictive multi-stage modelling for controlling a complex semiconductor device manufacturing process, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform the steps of:
collecting geometrical data from metrology measurements of production wafers made at select stages of the manufacturing process;
making an outcome probability prediction as to whether an output of the manufacturing process is within specifications at each of the select stages of the manufacturing process by machine-learning models trained for each of the select stages of the manufacturing process using a multiplicative kernel Gaussian process, wherein the outcome probability prediction is a function of a current stage and all prior stages; and
using the outcome probability prediction from the machine-learning models to identify those of the production wafers needing to be reworked or scrapped which are then removed from production, and wherein the program instructions further cause the computer to perform the steps of:
using the machine-learning models for each of the select stages of the manufacturing process to provide probabilistic predictions for a final outcome in real-time for the production wafers; and
using the probabilistic predictions to select the production wafers for rework, sort, scrap or disposition.

16. The computer program product of

Assignees

• IBM

Inventors

• Halle Scott
• Yeo Kyong Min
• Chao Robin Hsin Kuo
• Dunn Derren

Classifications

• G06F17/18

  for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)} · CPC title

• G06F30/20 Primary

  Design optimisation, verification or simulation (optimisation, verification or simulation of circuit designs G06F30/30) · CPC title

• G06F30/392

  Floor-planning or layout, e.g. partitioning or placement · CPC title

• G06F17/16

  Matrix or vector computation {, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization (matrix transposition G06F7/78)} · CPC title

• G06F2111/08

  Probabilistic or stochastic CAD · CPC title