Method and device for evaluating a statistically distributed measured value in the examination of an element of a photolithography process

Patent metadata

Field	Value
Publication number	US-12307334-B2
Application number	US-202117140250-A
Country	US
Kind code	B2
Filing date	Jan 4, 2021
Priority date	Jul 5, 2018
Publication date	May 20, 2025
Grant date	May 20, 2025

Abstract

The present invention relates to a method for evaluating a statistically distributed measured value in the examination of an element for a photolithography process, comprising the following steps: (a) using a plurality of parameters in a trained machine learning model, wherein the parameters characterize a state of a measurement environment in a time period assigned to a measurement of the measured value; and (b) executing the trained machine learning model in order to evaluate the measured value.

First claim

What is claimed is:

1. A method of evaluating a statistically distributed value to be measured in an examination of an element for a photolithography process prior to its measurement for improving a measurement accuracy of the statistically distributed value to be measured, the method comprising the following steps:
a. using a plurality of characterizing parameters in a trained machine learning model, wherein the plurality of characterizing parameters characterize a state of a measurement environment in a time period assigned to a measurement;
b. executing the trained machine learning model in order to evaluate the time period assigned to the measurement of a value to be measured, wherein the measurement is to be carried out in the assigned time period, the trained machine learning model predicting a quality criterion for the assigned time period assigned to the measurement of the value to be measured, wherein the trained machine learning model has been trained with a training data set comprising N data pairs, wherein each data pair of the N data pairs comprises two or more of the plurality of characterizing parameters which characterize the state of the measurement environment at an i-th time period and the quality criterion associated with the two or more of the plurality of characterizing parameters at the i-th time period; and
c. deferring the measurement of the value to be measured if the quality criterion predicted by the trained machine learning model indicates that a threshold is not met.

2. The method of claim 1 , comprising performing the planned measurement if the quality criterion predicted by the trained machine learning model is satisfied in the time period associated with the planned measurement.

3. The method of claim 2 , wherein the quality criterion comprises at least one element from the following group: a threshold value with regard to an expected value of a statistical distribution assigned to the value to be measured, an assignment to one range of a plurality of ranges predefined for the statistical distribution of the value to be measured, and a deviation of the value to be measured from the expected value of the statistical distribution.

4. The method of claim 2 , wherein the value to be measured is evaluated before a planned measurement, and furthermore comprising the following step: not performing the planned measurement of the value to be measured if the quality criterion is not satisfied in the time period assigned to the planned measurement of the value to be measured.

5. The method of claim 4 , furthermore comprising the following step: deferring the planned measurement until the quality criterion is satisfied in the time period assigned to the planned measurement of the value to be measured.

6. The method of claim 2 , further comprising performing a measurement to obtain a measured value, evaluating the measured value after it has been measured, and furthermore comprising the following step: rejecting the measured value if the quality criterion was not satisfied in the time period assigned to the planned measurement of the measured value.

7. The method of claim 2 , wherein a training data set for training the machine learning model comprises data pairs: characterizing parameters of an i-th measured value at a j-th position of the element of the photolithography process and the quality criterion of the i-th measured value at the j-th position of the element of the photolithography process.

8. The method of claim 1 , furthermore comprising the following step: producing a confidence statement with respect to the measured value.

9. The method of claim 1 , wherein the plurality of characterizing parameters comprise two or more elements from the following group: temperature of the measurement environment, pressure of the measurement environment, air humidity of the measurement environment, refractive index of the measurement environment, focus position of a device for measuring the statistically distributed value to be measured, wavelength of an optical system of the device, exposure intensity of the optical system of the device; exposure setting of the optical system of the device, degree of coherence of the optical system of the device, detector settings of the device, settings of one or more interferometers of the device, settings of one or more damping systems of the device, and settings of one or more drives of the device.

10. The method of claim 1 , wherein the plurality of characterizing parameters comprise a temporal development of their numerical values.

11. The method of claim 1 , wherein the value to be measured comprises a plurality of measurement recordings.

12. The method of claim 11 , wherein the plurality of measurement recordings comprise at least one changed characterizing parameter.

13. The method of claim 12 , wherein at least one of the plurality of characterizing parameters comprises at least one characteristic variable of its static distribution during the plurality of measurement recordings for the value to be measured.

14. The method of claim 1 , wherein the machine learning model comprises at least one element from the following group: a kernel density estimator, a statistical model, a decision tree, a linear model, a time-invariant model, a nearest neighbor classification, and a k-nearest neighbor algorithm, and their nonlinear extensions with nonlinear feature transformations.

15. The method of claim 14 , wherein the machine learning model comprises two or more different types of model of machine learning model from the group.

16. A computer program comprising instructions stored in a storage medium which, when they are executed by a computer system comprising one or more processors and the storage medium, cause the computer system to perform the method steps:
a. using a plurality of characterizing parameters in a trained machine learning model, wherein the plurality of characterizing parameters characterize a state of a measurement environment in a time period assigned to a measurement;
b. executing the trained machine learning model in order to evaluate the time period assigned to the measurement of a value to be measured, wherein the measurement is to be carried out in the assigned time period, the trained machine learning model predicting a quality criterion for the assigned time period assigned to the measurement of the value to be measured, wherein the trained machine learning model has been trained with a training data set comprising N data pairs, wherein each data pair of the N data pairs comprises two or more of the plurality of characterizing parameters which characterize the state of the measurement environment at an i-th time period and the quality criterion associated with the two or more of the plurality of characterizing parameters at the i-th time period; and
c. deferring the measurement of the value to be measured if the quality criterion predicted by the trained machine learning model indicates that a threshold is not met.

17. A device for evaluating a statistically distributed value to be measured in an examination of an element for a photolithography process prior to its measurement for improving a measurement accuracy of the statistically distributed value to be measured, comprising:
a. means for using a plurality of characterizing parameters in a trained machine learning model, wherein the plurality of characterizing parameters characterize a state of a measurement environment in a time period assigned to a measurement of the measured value;
b. means for executing the trained machine learning model in order to evalua

Assignees

• Zeiss Carl Smt Gmbh

Inventors

• Seidel Dirk
• Freytag Alexander
• Wojek Christian
• Töpfer Susanne
• Schmidt Carsten
• Husemann Christoph

Classifications

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/0985

  Hyperparameter optimisation; Meta-learning; Learning-to-learn · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06N20/20 Primary

  Ensemble learning · CPC title