DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 7-14 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claims 7-14 depend from “the method of Claim 1.” However, Claim 1 does not recite a method but rather recites a measurement system. This leaves the scope of the claims unclear. The Examiner is interpreting Claims 7-14 as depending from the method of Claim 6 for the purposes of this Office Action.
Claim 9 recites the term “the one or more reference measurement systems.” This term lacks antecedent basis even if Claim 9 was intended to depend from Claim 6. The Examiner is interpreting Claim 9 as depending from the method of Claim 7 for the purposes of this Office Action.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-23 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim(s) recite(s) the abstract idea of a mathematical and/or mental activity algorithm for generating and/or using a model for determining parameters of interest from semiconductor wafer inspection data.
This judicial exception is not integrated into a practical application because no improvement to the semiconductor wafers or their manufacturing process is realized through performance of the algorithm.
The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the recited computing system and machine learning techniques amount to the recitation of general-purpose computer elements for performing the abstract idea through use of a general-purpose computer and do not serve to amount to significantly more than the recitation of the abstract idea itself (see Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014)). The recited illumination system and detector amount to well-understood, routine, and conventional semiconductor wafer data gathering elements [See Fig. 1 of US 20230124431 A1 and Paragraph [0055] of US 20170109646 A1] and do not serve to amount to significantly more than the recitation of the abstract idea itself in gathering the data needed to perform the algorithm.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 16-18 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Bringoltz et al. (US 20230124431 A1)[hereinafter “Bringoltz”].
Regarding Claim 16, Bringoltz discloses a measurement system comprising:
an illumination subsystem configured to illuminate a structure with a first amount of radiation at a first measurement site disposed on a first semiconductor wafer wafers [See Fig. 1 and Paragraph [0028] – “The system 10 includes a light source 20, which generates a beam of light 22 of a predetermined wavelength range. The beam of light 22 is reflected from wafer patterns at a wafer site 14 (indicated as reflected, or “scattered,” light 24) towards a spectrophotometric detector 26.”] in a pre-process state and illuminate the structure with a second amount of radiation at the first measurement site disposed on the first semiconductor wafer in a post-process state, wherein the pre-process state and the post-process state are separated by one or more intervening semiconductor manufacturing process steps [Abstract – “A system and methods for Advance Process Control (APC) in semiconductor manufacturing include: for each of a plurality of waiter sites, receiving a pre-process set of scatterometric training data, measured before implementation of a processing step”];
a detector configured to detect a first amount of raw measurement data associated with a measurement of the structure at the first measurement site disposed on the first semiconductor wafer in response to the first amount of radiation [See Fig. 1 and Paragraph [0028] – “The system 10 includes a light source 20, which generates a beam of light 22 of a predetermined wavelength range. The beam of light 22 is reflected from wafer patterns at a wafer site 14 (indicated as reflected, or “scattered,” light 24) towards a spectrophotometric detector 26.”Paragraph [0029] – “The scatterometric data 32 generated by the metrology system 30 typically includes various types of plotted data 34, which may be represented in vector form (e.g., a spectrogram, whose data points are measures of reflected light intensity “I” at different light wavelengths, or a mapping of reflected irradiance vs. incident angle). As described above, variations between sets of scatterometric data are indicative of variations in pattern parameters at the respective wafer sites.”] and detect a second amount of raw measurement data associated with a measurement of the structure at the first measurement site disposed on the first semiconductor wafer in response to the second amount of radiation [Abstract – “A system and methods for Advance Process Control (APC) in semiconductor manufacturing include: for each of a plurality of waiter sites, receiving a pre-process set of scatterometric training data, measured before implementation of a processing step”]; and
a computing system [See Fig. 1 and Paragraph [0031] – “In embodiments of the present invention, a computer system including machine learning (ML) tools known in the art, referred to herein as an ML modeling system 40, may be configured for training an ML model for OCD metrology.”] configured to:
receive the first and second amounts of raw measurement data [Abstract – “receiving a pre-process set of scatterometric training data, measured before implementation of a processing step, receiving a corresponding post-process set of scatterometric training data measured after implementation of the process step”]; and
estimate a value of a parameter of interest characterizing the structure at the first measurement site disposed on the first semiconductor wafer based on a trained, combined machine learning based measurement model and the first and second amounts of raw measurement data [Abstract – “generating a machine learning model correlating variations in the pre-process sets of scatterometric training data and the corresponding process control knob training data with the corresponding post-process sets of scatterometric training data, to train the machine learning model to recommend changes to process control knob settings to compensate for variations in the pre-process scatterometric data.”].
Regarding Claim 17, Bringoltz discloses that the trained, combined machine learning based measurement model includes a trained, pre-process measurement model [Paragraph [0049] – “FIG. 5 depicts a process of training a machine learning model 500 that directly transforms a representation of wafer patterns, before and after a process step, and then calibrates the post-process signal (i.e., scatterometric data, {right arrow over (S)}.sub.Post) to the target (controlled) parameter, {right arrow over (P)}.sub.Post. The model 500 includes an encoder-decoder neural network 510, which has encoder layers 520 that compress the dimensionality of the pre-process signal (i.e., the pre-process scatterometric data), a bottleneck layer 540, and decoder layers 550.”], a trained, post-process measurement model [Paragraph [0050] – “The second part of the ML 500 is a metrology interpretation function, meaning a calibration of the network output (the post-process scatterometric data) to the wafer parameters to be controlled. This calibration, typically by an OCD model 570, uses a second loss term and makes use of labelled data.”], and a trained weighting model [Paragraph [0050] – “A second loss term 545 may be set as a difference between the predicted parameters (the output of the OCD model 570) and measured target parameters.”Paragraph [0051] – “The two loss terms compete on gradient direction during training convergence, meaning that an additional hyperparameter in the form of these loss weights must be tuned.”].
Regarding Claim 18, Bringoltz discloses that the first amount of raw measurement data is provided as input to the trained, pre-process measurement model [Paragraph [0049] – “FIG. 5 depicts a process of training a machine learning model 500 that directly transforms a representation of wafer patterns, before and after a process step, and then calibrates the post-process signal (i.e., scatterometric data, {right arrow over (S)}.sub.Post) to the target (controlled) parameter, {right arrow over (P)}.sub.Post. The model 500 includes an encoder-decoder neural network 510, which has encoder layers 520 that compress the dimensionality of the pre-process signal (i.e., the pre-process scatterometric data), a bottleneck layer 540, and decoder layers 550.”], wherein the second amount of raw measurement data is provided as input to the trained, post-process measurement model [Paragraph [0050] – “The second part of the ML 500 is a metrology interpretation function, meaning a calibration of the network output (the post-process scatterometric data) to the wafer parameters to be controlled. This calibration, typically by an OCD model 570, uses a second loss term and makes use of labelled data.”], and wherein an output of the trained, pre-process measurement model and an output of the trained, post-process measurement model is provided as input to the trained weighting model, and wherein the output of the trained weighting model is the estimated value of the parameter of interest [Paragraph [0050] – “A second loss term 545 may be set as a difference between the predicted parameters (the output of the OCD model 570) and measured target parameters.”Paragraph [0051] – “The two loss terms compete on gradient direction during training convergence, meaning that an additional hyperparameter in the form of these loss weights must be tuned.”].
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1, 3, 6, 8, 10, 11, 15, and 19-23 is/are rejected under 35 U.S.C. 103 as being unpatentable over Bringoltz et al. (US 20230124431 A1)[hereinafter “Bringoltz”] and David (US 20170109646 A1).
Regarding Claim 1, Bringoltz discloses a measurement system [Fig. 1] comprising:
an illumination subsystem configured to illuminate a structure with an amount of radiation at each of a plurality of measurement sites disposed on a first plurality of semiconductor wafers [See Fig. 1 and Paragraph [0028] – “The system 10 includes a light source 20, which generates a beam of light 22 of a predetermined wavelength range. The beam of light 22 is reflected from wafer patterns at a wafer site 14 (indicated as reflected, or “scattered,” light 24) towards a spectrophotometric detector 26.”] in a pre-process state [Abstract – “A system and methods for Advance Process Control (APC) in semiconductor manufacturing include: for each of a plurality of waiter sites, receiving a pre-process set of scatterometric training data, measured before implementation of a processing step”];
a detector configured to detect an amount of raw measurement data associated with the measurements of the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in response to the amount of radiation [See Fig. 1 and Paragraph [0028] – “The system 10 includes a light source 20, which generates a beam of light 22 of a predetermined wavelength range. The beam of light 22 is reflected from wafer patterns at a wafer site 14 (indicated as reflected, or “scattered,” light 24) towards a spectrophotometric detector 26.”Paragraph [0029] – “The scatterometric data 32 generated by the metrology system 30 typically includes various types of plotted data 34, which may be represented in vector form (e.g., a spectrogram, whose data points are measures of reflected light intensity “I” at different light wavelengths, or a mapping of reflected irradiance vs. incident angle). As described above, variations between sets of scatterometric data are indicative of variations in pattern parameters at the respective wafer sites.”]; and
a computing system [See Fig. 1 and Paragraph [0031] – “In embodiments of the present invention, a computer system including machine learning (ML) tools known in the art, referred to herein as an ML modeling system 40, may be configured for training an ML model for OCD metrology.”] configured to:
receive first estimated values of a first parameter of interest characterizing the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in the pre-process state, the first estimated values of the first parameter of interest generated based on the measurements of the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers [See Fig. 6 and Paragraph [0052] – “FIG. 6 depicts application of the ML model 500 in production, i.e., for inference, to generate a knob recommendation, {right arrow over (k)}.sub.recommended. An optimization step may be applied to force the model output to be as close as possible to the target parameter. A recommendation {right arrow over (k)}.sub.recommended may then be achieved by minimizing a distance metric, D. The metric D is a difference between 1) a prediction of pattern parameters made by the model (indicated below as {circumflex over (P)}), which is based on the pre-process spectra {right arrow over (S)}.sub.Pre and the knob setting, {right arrow over (k)}, and 2) the target value of the pattern parameter {right arrow over (P)}.sub.target. The value of {right arrow over (k)}.sub.recommended may then be deduced by the equation:
PNG
media_image1.png
48
611
media_image1.png
Greyscale
”].
Bringoltz discloses estimating values of a second parameter of interest characterizing the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in a post-process state [See Fig. 5 and Paragraph [0050] – “The second part of the ML 500 is a metrology interpretation function, meaning a calibration of the network output (the post-process scatterometric data) to the wafer parameters to be controlled. This calibration, typically by an OCD model 570, uses a second loss term and makes use of labelled data. A second loss term 545 may be set as a difference between the predicted parameters (the output of the OCD model 570) and measured target parameters.”], but fails to disclose that the computing system is configured to map the first estimated values of the first parameter of interest characterizing the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in the pre-process state to first estimated values of a second parameter of interest characterizing the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in a post-process state because Bringoltz fails to disclose performing machine learning using the relationship between the estimated values of the first/second parameters.
However, David discloses mapping upstream/downstream semiconductor manufacturing feature predictions using machine learning [Abstract – “Techniques for measuring and/or compensating for process variations in a semiconductor manufacturing processes. Machine learning algorithms are used on extensive sets of input data, including upstream data, to organize and pre-process the input data, and to correlate the input data to specific features of interest. The correlations can then be used to make process adjustments. The techniques may be applied to any feature or step of the semiconductor manufacturing process, such as overlay, critical dimension, and yield prediction.”Paragraph [0153] – “In an embodiment, the algorithm utilizes calculated propensity from an upstream test which contains more failures to determine the failure rate of the final test, which may contain much fewer failures. For example, at the end of an upstream testing process, the failure rate may be higher, which would make it easier to produce a model that gives more accurate predictions (e.g., a CHAID decision tree). A model can be built to determine the failure rate of this upstream process, and produce a pass/fail prediction along with a confidence and propensity metric. The failure prediction, confidence and propensity metric can then be used as inputs to predict the failure of a test further downstream. This may be particularly useful when the test downstream has a lower number of failures, making it more difficult to build an accurate model.”].
It would have been obvious to map the pre and post process predictions because the determination of their correlation would have improved the performance of the machine learning in determining resulting wafer features.
Bringoltz, as modified, would disclose that the pre-process state and the post-process state are separated by one or more intervening semiconductor manufacturing process steps [Abstract – “receiving a pre-process set of scatterometric training data, measured before implementation of a processing step, receiving a corresponding post-process set of scatterometric training data measured after implementation of the process step”]; and
the computing system is configured to train a post-process measurement model based on the first estimated values of the second parameter of interest characterizing the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in the post-process state [See Fig. 5 and Paragraph [0050] – “The second part of the ML 500 is a metrology interpretation function, meaning a calibration of the network output (the post-process scatterometric data) to the wafer parameters to be controlled. This calibration, typically by an OCD model 570, uses a second loss term and makes use of labelled data. A second loss term 545 may be set as a difference between the predicted parameters (the output of the OCD model 570) and measured target parameters.”] and an amount of raw measurement data associated with measurements of the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in the post-process state [See Fig. 5 and Paragraph [0049] – “FIG. 5 depicts a process of training a machine learning model 500 that directly transforms a representation of wafer patterns, before and after a process step, and then calibrates the post-process signal (i.e., scatterometric data, {right arrow over (S)}.sub.Post) to the target (controlled) parameter, {right arrow over (P)}.sub.Post. … The output is set to the set of post-process scatterometric data {right arrow over (S)}.sub.Post. A loss function 540 may be set as the difference between scatterometric data predicted by the network and a measured set of post-process scatterometric data.”].
Regarding Claim 3, Bringoltz discloses that the first parameter of interest and the second parameter of interest are the same parameter [The pattern parameters].
Regarding Claim 6, Bringoltz discloses a method [See Fig. 6 and Paragraph [0052] – “FIG. 6 depicts application of the ML model 500 in production”] comprising:
receiving first estimated values of a first parameter of interest characterizing a structure at each of a plurality of measurement sites disposed on a first plurality of semiconductor wafers in a pre-process state, the first estimated values of the first parameter of interest generated based on measurements of the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers [See Fig. 6 and Paragraph [0052] – “FIG. 6 depicts application of the ML model 500 in production, i.e., for inference, to generate a knob recommendation, {right arrow over (k)}.sub.recommended. An optimization step may be applied to force the model output to be as close as possible to the target parameter. A recommendation {right arrow over (k)}.sub.recommended may then be achieved by minimizing a distance metric, D. The metric D is a difference between 1) a prediction of pattern parameters made by the model (indicated below as {circumflex over (P)}), which is based on the pre-process spectra {right arrow over (S)}.sub.Pre and the knob setting, {right arrow over (k)}, and 2) the target value of the pattern parameter {right arrow over (P)}.sub.target. The value of {right arrow over (k)}.sub.recommended may then be deduced by the equation:
PNG
media_image1.png
48
611
media_image1.png
Greyscale
”] by one or more in-line measurement systems [Fig. 1].
Bringoltz discloses estimating values of a second parameter of interest characterizing the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in a post-process state [See Fig. 5 and Paragraph [0050] – “The second part of the ML 500 is a metrology interpretation function, meaning a calibration of the network output (the post-process scatterometric data) to the wafer parameters to be controlled. This calibration, typically by an OCD model 570, uses a second loss term and makes use of labelled data. A second loss term 545 may be set as a difference between the predicted parameters (the output of the OCD model 570) and measured target parameters.”], but fails to disclose mapping the first estimated values of the first parameter of interest characterizing the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in the pre-process state to first estimated values of a second parameter of interest characterizing the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in a post-process state because Bringoltz fails to disclose performing machine learning using the relationship between the estimated values of the first/second parameters.
However, David discloses mapping upstream/downstream semiconductor manufacturing feature predictions using machine learning [Abstract – “Techniques for measuring and/or compensating for process variations in a semiconductor manufacturing processes. Machine learning algorithms are used on extensive sets of input data, including upstream data, to organize and pre-process the input data, and to correlate the input data to specific features of interest. The correlations can then be used to make process adjustments. The techniques may be applied to any feature or step of the semiconductor manufacturing process, such as overlay, critical dimension, and yield prediction.”Paragraph [0153] – “In an embodiment, the algorithm utilizes calculated propensity from an upstream test which contains more failures to determine the failure rate of the final test, which may contain much fewer failures. For example, at the end of an upstream testing process, the failure rate may be higher, which would make it easier to produce a model that gives more accurate predictions (e.g., a CHAID decision tree). A model can be built to determine the failure rate of this upstream process, and produce a pass/fail prediction along with a confidence and propensity metric. The failure prediction, confidence and propensity metric can then be used as inputs to predict the failure of a test further downstream. This may be particularly useful when the test downstream has a lower number of failures, making it more difficult to build an accurate model.”].
It would have been obvious to map the pre and post process predictions because the determination of their correlation would have improved the performance of the machine learning in determining resulting wafer features.
Bringoltz, as modified, would disclose that the pre-process state and the post-process state are separated by one or more intervening semiconductor manufacturing process steps [Abstract – “receiving a pre-process set of scatterometric training data, measured before implementation of a processing step, receiving a corresponding post-process set of scatterometric training data measured after implementation of the process step”]; and
training a post-process measurement model based on the first estimated values of the second parameter of interest characterizing the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in the post-process state [See Fig. 5 and Paragraph [0050] – “The second part of the ML 500 is a metrology interpretation function, meaning a calibration of the network output (the post-process scatterometric data) to the wafer parameters to be controlled. This calibration, typically by an OCD model 570, uses a second loss term and makes use of labelled data. A second loss term 545 may be set as a difference between the predicted parameters (the output of the OCD model 570) and measured target parameters.”] and an amount of raw measurement data associated with measurements of the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in the post-process state by the one or more in-line measurement systems [See Fig. 5 and Paragraph [0049] – “FIG. 5 depicts a process of training a machine learning model 500 that directly transforms a representation of wafer patterns, before and after a process step, and then calibrates the post-process signal (i.e., scatterometric data, {right arrow over (S)}.sub.Post) to the target (controlled) parameter, {right arrow over (P)}.sub.Post. … The output is set to the set of post-process scatterometric data {right arrow over (S)}.sub.Post. A loss function 540 may be set as the difference between scatterometric data predicted by the network and a measured set of post-process scatterometric data.”].
Regarding Claim 8, Bringoltz discloses that the first parameter of interest and the second parameter of interest are the same parameter [The pattern parameters].
Regarding Claim 10, Bringoltz discloses that the post-process measurement model is a machine learning based measurement model [See Fig. 5 and Paragraph [0050] – “The second part of the ML 500 is a metrology interpretation function, meaning a calibration of the network output (the post-process scatterometric data) to the wafer parameters to be controlled. This calibration, typically by an OCD model 570, uses a second loss term and makes use of labelled data. A second loss term 545 may be set as a difference between the predicted parameters (the output of the OCD model 570) and measured target parameters.”].
Regarding Claim 11, Bringoltz discloses illuminating the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers [See Fig. 1 and Paragraph [0028] – “The system 10 includes a light source 20, which generates a beam of light 22 of a predetermined wavelength range. The beam of light 22 is reflected from wafer patterns at a wafer site 14 (indicated as reflected, or “scattered,” light 24) towards a spectrophotometric detector 26.”] in the pre-process state [Abstract – “A system and methods for Advance Process Control (APC) in semiconductor manufacturing include: for each of a plurality of waiter sites, receiving a pre-process set of scatterometric training data, measured before implementation of a processing step”];
detecting the amount of raw measurement data associated with the measurements of the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in response to the illumination [See Fig. 1 and Paragraph [0028] – “The system 10 includes a light source 20, which generates a beam of light 22 of a predetermined wavelength range. The beam of light 22 is reflected from wafer patterns at a wafer site 14 (indicated as reflected, or “scattered,” light 24) towards a spectrophotometric detector 26.”Paragraph [0029] – “The scatterometric data 32 generated by the metrology system 30 typically includes various types of plotted data 34, which may be represented in vector form (e.g., a spectrogram, whose data points are measures of reflected light intensity “I” at different light wavelengths, or a mapping of reflected irradiance vs. incident angle). As described above, variations between sets of scatterometric data are indicative of variations in pattern parameters at the respective wafer sites.”]; and
estimating the first estimated values of the first parameter of interest based on the amount of raw measurement data [See Fig. 6 and Paragraph [0052] – “FIG. 6 depicts application of the ML model 500 in production, i.e., for inference, to generate a knob recommendation, {right arrow over (k)}.sub.recommended. An optimization step may be applied to force the model output to be as close as possible to the target parameter. A recommendation {right arrow over (k)}.sub.recommended may then be achieved by minimizing a distance metric, D. The metric D is a difference between 1) a prediction of pattern parameters made by the model (indicated below as {circumflex over (P)}), which is based on the pre-process spectra {right arrow over (S)}.sub.Pre and the knob setting, {right arrow over (k)}, and 2) the target value of the pattern parameter {right arrow over (P)}.sub.target. The value of {right arrow over (k)}.sub.recommended may then be deduced by the equation:
PNG
media_image1.png
48
611
media_image1.png
Greyscale
”].
Regarding Claim 15, Bringoltz discloses receiving an amount of raw measurement data associated with a measurement of the structure at a measurement site disposed on a third semiconductor wafer in the post-process state by one of the one or more in-line measurement systems [See Fig. 5 and Paragraph [0049] – “FIG. 5 depicts a process of training a machine learning model 500 that directly transforms a representation of wafer patterns, before and after a process step, and then calibrates the post-process signal (i.e., scatterometric data, {right arrow over (S)}.sub.Post) to the target (controlled) parameter, {right arrow over (P)}.sub.Post. … The output is set to the set of post-process scatterometric data {right arrow over (S)}.sub.Post.”]; and
estimating a value of the second parameter of interest characterizing the structure at the measurement site disposed on the third semiconductor wafer in the post-process state based on the received amount of raw measurement data and the trained post-process measurement model [Paragraph [0027] – “Manufacturing variations cause slight variations in pattern parameters, such that these pattern parameters vary between wafers and at sites across a single wafer after each process step. As described further hereinbelow, embodiments of the present invention provide methods and systems for determining changes that should be made to knob settings in order to reduce variations in parameters at sites within wafers and between wafers.” The ongoing process discloses determining wafer parameters including that for a “third semiconductor wafer.].
Regarding Claim 19, Bringoltz discloses a method comprising:
receiving a first amount of raw measurement data associated with a measurement of a structure at a measurement site disposed on a first semiconductor wafer by a measurement system, the first semiconductor wafer in a pre-process state [See Fig. 1 and Abstract – “A system and methods for Advance Process Control (APC) in semiconductor manufacturing include: for each of a plurality of waiter sites, receiving a pre-process set of scatterometric training data, measured before implementation of a processing step”];
receiving a second amount of raw measurement data associated with a measurement of the structure at the measurement site disposed on the first semiconductor wafer by a measurement system, the first semiconductor wafer in a post-process state, wherein the pre-process state and the post-process state are separated by one or more intervening semiconductor manufacturing process steps [Abstract – “receiving a corresponding post-process set of scatterometric training data measured after implementation of the process step”].
Bringoltz fails to disclose that the measurements are taken using a first in-line measurement system and a second in-line measurement system. However, David discloses taking measurements both upstream and downstream of a semiconductor manufacturing process [Paragraph [0153] – “In an embodiment, the algorithm utilizes calculated propensity from an upstream test which contains more failures to determine the failure rate of the final test, which may contain much fewer failures. For example, at the end of an upstream testing process, the failure rate may be higher, which would make it easier to produce a model that gives more accurate predictions (e.g., a CHAID decision tree). A model can be built to determine the failure rate of this upstream process, and produce a pass/fail prediction along with a confidence and propensity metric. The failure prediction, confidence and propensity metric can then be used as inputs to predict the failure of a test further downstream. This may be particularly useful when the test downstream has a lower number of failures, making it more difficult to build an accurate model.”].
It would have been obvious to take the pre and post process measurements in such a manner because David teaches that this is useful in modeling the semiconductor manufacturing process.
Bringoltz further discloses estimating a value of a parameter of interest characterizing the structure at the measurement site disposed on the first semiconductor wafer based on a trained, combined machine learning based measurement model and the first and second amounts of raw measurement data [Abstract – “generating a machine learning model correlating variations in the pre-process sets of scatterometric training data and the corresponding process control knob training data with the corresponding post-process sets of scatterometric training data, to train the machine learning model to recommend changes to process control knob settings to compensate for variations in the pre-process scatterometric data.].
Regarding Claim 20, Bringoltz discloses that the trained, combined machine learning based measurement model includes a trained, pre-process measurement model [Paragraph [0049] – “FIG. 5 depicts a process of training a machine learning model 500 that directly transforms a representation of wafer patterns, before and after a process step, and then calibrates the post-process signal (i.e., scatterometric data, {right arrow over (S)}.sub.Post) to the target (controlled) parameter, {right arrow over (P)}.sub.Post. The model 500 includes an encoder-decoder neural network 510, which has encoder layers 520 that compress the dimensionality of the pre-process signal (i.e., the pre-process scatterometric data), a bottleneck layer 540, and decoder layers 550.”], a trained, post-process measurement model [Paragraph [0050] – “The second part of the ML 500 is a metrology interpretation function, meaning a calibration of the network output (the post-process scatterometric data) to the wafer parameters to be controlled. This calibration, typically by an OCD model 570, uses a second loss term and makes use of labelled data.”], and a trained weighting model [Paragraph [0050] – “A second loss term 545 may be set as a difference between the predicted parameters (the output of the OCD model 570) and measured target parameters.”Paragraph [0051] – “The two loss terms compete on gradient direction during training convergence, meaning that an additional hyperparameter in the form of these loss weights must be tuned.”].
Regarding Claim 21, Bringoltz discloses that the first amount of raw measurement data is provided as input to the trained, pre-process measurement model [Paragraph [0049] – “FIG. 5 depicts a process of training a machine learning model 500 that directly transforms a representation of wafer patterns, before and after a process step, and then calibrates the post-process signal (i.e., scatterometric data, {right arrow over (S)}.sub.Post) to the target (controlled) parameter, {right arrow over (P)}.sub.Post. The model 500 includes an encoder-decoder neural network 510, which has encoder layers 520 that compress the dimensionality of the pre-process signal (i.e., the pre-process scatterometric data), a bottleneck layer 540, and decoder layers 550.”], wherein the second amount of raw measurement data is provided as input to the trained, post-process measurement model [Paragraph [0050] – “The second part of the ML 500 is a metrology interpretation function, meaning a calibration of the network output (the post-process scatterometric data) to the wafer parameters to be controlled. This calibration, typically by an OCD model 570, uses a second loss term and makes use of labelled data.”], and wherein an output of the trained, pre-process measurement model and an output of the trained, post-process measurement model is provided as input to the trained weighting model, and wherein the output of the trained weighting model is the estimated value of the parameter of interest [Paragraph [0050] – “A second loss term 545 may be set as a difference between the predicted parameters (the output of the OCD model 570) and measured target parameters.”Paragraph [0051] – “The two loss terms compete on gradient direction during training convergence, meaning that an additional hyperparameter in the form of these loss weights must be tuned.”].
Regarding Claim 22, Bringoltz discloses that the trained, pre-process measurement model, the trained, post-process measurement model, and the trained weighting model are trained simultaneously [Paragraph [0049] – “FIG. 5 depicts a process of training a machine learning model 500 that directly transforms a representation of wafer patterns, before and after a process step, and then calibrates the post-process signal (i.e., scatterometric data, {right arrow over (S)}.sub.Post) to the target (controlled) parameter, {right arrow over (P)}.sub.Post.”].
Regarding Claim 23, the combination would disclose that the first in-line measurement system and the second in-line measurement system are the same measurement system [Taking measurements both upstream and downstream of a semiconductor manufacturing process per David to accomplish the modeling of Bringoltz].
Claim(s) 2, 4, 5, 7, 9, and 12-14 is/are rejected under 35 U.S.C. 103 as being unpatentable over Bringoltz et al. (US 20230124431 A1)[hereinafter “Bringoltz”], David (US 20170109646 A1), and Banna et al. (US 20210175104 A1)[hereinafter “Banna”]
Regarding Claim 2, Bringoltz fails to disclose the use of measured/estimated data from one or more reference measurement sites disposed on a second plurality of semiconductor wafers in the machine learning process.
However, Banna discloses the comparison of estimated values for wafer features with information from one or more reference measurement sites disposed on a second plurality of semiconductor wafers [Paragraph [0079] – “In addition to improving wafer-to-wafer (and/or lot-to-lot) device performance variability control, integrating in-line metrology data to the adaptive model allows process engineers to understand potential points of failure. VM and OBM are used to qualitatively provide a spatial map of the dimension of interest, which is then compared to a golden reference to identify potential points of failure. Selected wafers are then taken to the in-line metrology tool (such as e-beam for customized metrology) and the areas of interest where the potential failure sites are identified are scanned in detail to identify the root cause of failure.”Paragraph [0082] – “The run-to-run process control flow 1010 starts at block 1060, where on-wafer metrology (i.e. spatial map of dimensions of interest) is predicted based on VM and OBM data. The prediction may be compared to golden references and/or statistical references.”]. It would have been obvious to make use of reference measurement sites disposed on a second plurality of semiconductor wafers in order to improve the machine learning process through use of appropriate ground truth information.
The combination would disclose that the illumination subsystem is further configured to illuminate the structure with a second amount of radiation at each of one or more reference measurement sites disposed on a second plurality of semiconductor wafers in the pre-process state [Abstract and Fig. 1 (light source 20) of Bringoltz, per application to reference wafers in light of Banna – “A system and methods for Advance Process Control (APC) in semiconductor manufacturing include: for each of a plurality of waiter sites, receiving a pre-process set of scatterometric training data, measured before implementation of a processing step”];
wherein the detector is further configured to detect an amount of raw measurement data associated with the measurements of the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers in response to the second amount of radiation [Abstract and Fig. 1 (spectrophotometric detector 26) of Bringoltz, per application to reference wafers in light of Banna – “A system and methods for Advance Process Control (APC) in semiconductor manufacturing include: for each of a plurality of waiter sites, receiving a pre-process set of scatterometric training data, measured before implementation of a processing step”]; and
wherein the computing system is further configured to: receive second estimated values of a first parameter of interest characterizing the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers in the pre-process state, the second estimated values of the first parameter of interest generated based on the measurements of the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers [See Fig. 6 and Paragraph [0052] of Bringoltz, per application to reference wafers in light of Banna – “FIG. 6 depicts application of the ML model 500 in production, i.e., for inference, to generate a knob recommendation, {right arrow over (k)}.sub.recommended. An optimization step may be applied to force the model output to be as close as possible to the target parameter. A recommendation {right arrow over (k)}.sub.recommended may then be achieved by minimizing a distance metric, D. The metric D is a difference between 1) a prediction of pattern parameters made by the model (indicated below as {circumflex over (P)}), which is based on the pre-process spectra {right arrow over (S)}.sub.Pre and the knob setting, {right arrow over (k)}, and 2) the target value of the pattern parameter {right arrow over (P)}.sub.target. The value of {right arrow over (k)}.sub.recommended may then be deduced by the equation:
PNG
media_image1.png
48
611
media_image1.png
Greyscale
”];
receive second estimated values of the second parameter of interest characterizing the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers in the post-process state, the second estimated values of the second parameter of interest generated based on measurements of the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers by one or more reference measurement systems [Per David and per application to reference wafers in light of Banna]; and
generate a mapping model that maps the second estimated values of the first parameter of interest to the second estimated values of the second parameter of interest [Per David and per application to reference wafers in light of Banna].
Regarding Claim 4, the combination would disclose training the post-process measurement model based on the second estimated values of the second parameter of interest characterizing the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers in the post-process state [See Fig. 5 and Paragraph [0050] of Bringoltz, per application to reference wafers in light of Banna – “The second part of the ML 500 is a metrology interpretation function, meaning a calibration of the network output (the post-process scatterometric data) to the wafer parameters to be controlled. This calibration, typically by an OCD model 570, uses a second loss term and makes use of labelled data. A second loss term 545 may be set as a difference between the predicted parameters (the output of the OCD model 570) and measured target parameters.”] and an amount of raw measurement data associated with measurements of the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers in the post-process state [See Fig. 5 and Paragraph [0049] of Bringoltz, per application to reference wafers in light of Banna – “FIG. 5 depicts a process of training a machine learning model 500 that directly transforms a representation of wafer patterns, before and after a process step, and then calibrates the post-process signal (i.e., scatterometric data, {right arrow over (S)}.sub.Post) to the target (controlled) parameter, {right arrow over (P)}.sub.Post. … The output is set to the set of post-process scatterometric data {right arrow over (S)}.sub.Post. A loss function 540 may be set as the difference between scatterometric data predicted by the network and a measured set of post-process scatterometric data.”].
Regarding Claim 5, Bringoltz fails to disclose that the computing system further configured to: weigh the second estimated values of the second parameter of interest characterizing the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers in the post-process state differently than the first estimated values of the second parameter of interest characterizing the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in the post-process state during the training.
However, David discloses appropriately weighting estimated values based their confidence [Paragraph [0090] – “The score generated by the model will correspond to whatever metric was used as a target for training the algorithm that generated the model. For example, if a DBO measurement was used for the target to train the algorithm, then the score will be a predicted DBO measurement. If the target was a parametric test value, then the score will be a prediction of that parametric test value.”Paragraph [0091] – “In addition to the score, the model can also output a confidence metric that describes how reliable the score prediction is. This can be useful in determining whether or not to employ the score, or weight the use of that prediction in conjunction with other traditional measurements.”]. It would have been obvious to weight the reference predictions higher than the first estimated values because the reference wafers of Banna represent “golden” ground truth values, doing so would have improved the machine learning process.
Regarding Claim 7, Bringoltz fails to disclose the use of measured/estimated data from one or more reference measurement sites disposed on a second plurality of semiconductor wafers in the machine learning process.
However, Banna discloses the comparison of estimated values for wafer features with information from one or more reference measurement sites disposed on a second plurality of semiconductor wafers [Paragraph [0079] – “In addition to improving wafer-to-wafer (and/or lot-to-lot) device performance variability control, integrating in-line metrology data to the adaptive model allows process engineers to understand potential points of failure. VM and OBM are used to qualitatively provide a spatial map of the dimension of interest, which is then compared to a golden reference to identify potential points of failure. Selected wafers are then taken to the in-line metrology tool (such as e-beam for customized metrology) and the areas of interest where the potential failure sites are identified are scanned in detail to identify the root cause of failure.”Paragraph [0082] – “The run-to-run process control flow 1010 starts at block 1060, where on-wafer metrology (i.e. spatial map of dimensions of interest) is predicted based on VM and OBM data. The prediction may be compared to golden references and/or statistical references.”]. It would have been obvious to make use of reference measurement sites disposed on a second plurality of semiconductor wafers in order to improve the machine learning process through use of appropriate ground truth information.
The combination would disclose receiving second estimated values of a first parameter of interest characterizing the structure at each of one or more reference measurement sites disposed on a second plurality of semiconductor wafers in the pre-process state, the second estimated values of the first parameter of interest generated based on measurements of the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers [See Fig. 6 and Paragraph [0052] of Bringoltz, per application to reference wafers in light of Banna – “FIG. 6 depicts application of the ML model 500 in production, i.e., for inference, to generate a knob recommendation, {right arrow over (k)}.sub.recommended. An optimization step may be applied to force the model output to be as close as possible to the target parameter. A recommendation {right arrow over (k)}.sub.recommended may then be achieved by minimizing a distance metric, D. The metric D is a difference between 1) a prediction of pattern parameters made by the model (indicated below as {circumflex over (P)}), which is based on the pre-process spectra {right arrow over (S)}.sub.Pre and the knob setting, {right arrow over (k)}, and 2) the target value of the pattern parameter {right arrow over (P)}.sub.target. The value of {right arrow over (k)}.sub.recommended may then be deduced by the equation:
PNG
media_image1.png
48
611
media_image1.png
Greyscale
”] by the one or more in-line measurement systems [Fig. 1 of Bringoltz];
receiving second estimated values of the second parameter of interest characterizing the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers in the post-process state, the second estimated values of the second parameter of interest generated based on measurements of the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers by one or more reference measurement systems [Per David and per application to reference wafers in light of Banna]; and
generating a mapping model that maps the second estimated values of the first parameter of interest to the second estimated values of the second parameter of interest [Per David and per application to reference wafers in light of Banna].
Regarding Claim 9, Banna discloses that the one or more reference measurement systems employ a different metrology technique than the one or more in-line measurement systems [Paragraph [0052] – “The output of module 106 is multi-input metrology data 107a (derived from on-tool and off-tool metrology data) from the currently used DoE wafers. The output of module 106 may also have some reference measurements 107b obtained from previous trustworthy measurements. These reference measurements (sometimes called “golden profile”)[.]”].
Regarding Claim 12, the combination would disclose illuminating the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers in the pre-process state [See Fig. 1 and Paragraph [0028] of Bringoltz, per application to reference wafers in light of Banna – “The system 10 includes a light source 20, which generates a beam of light 22 of a predetermined wavelength range. The beam of light 22 is reflected from wafer patterns at a wafer site 14 (indicated as reflected, or “scattered,” light 24) towards a spectrophotometric detector 26.”] in the pre-process state [Abstract – “A system and methods for Advance Process Control (APC) in semiconductor manufacturing include: for each of a plurality of waiter sites, receiving a pre-process set of scatterometric training data, measured before implementation of a processing step”];
detecting an amount of raw measurement data in response to the illumination [See Fig. 1 and Paragraph [0028] of Bringoltz, per application to reference wafers in light of Banna – “The system 10 includes a light source 20, which generates a beam of light 22 of a predetermined wavelength range. The beam of light 22 is reflected from wafer patterns at a wafer site 14 (indicated as reflected, or “scattered,” light 24) towards a spectrophotometric detector 26.”Paragraph [0029] – “The scatterometric data 32 generated by the metrology system 30 typically includes various types of plotted data 34, which may be represented in vector form (e.g., a spectrogram, whose data points are measures of reflected light intensity “I” at different light wavelengths, or a mapping of reflected irradiance vs. incident angle). As described above, variations between sets of scatterometric data are indicative of variations in pattern parameters at the respective wafer sites.”]; and
estimating the second estimated values of the first parameter of interest based on the amount of raw measurement data [See Fig. 6 and Paragraph [0052] of Bringoltz, per application to reference wafers in light of Banna – “FIG. 6 depicts application of the ML model 500 in production, i.e., for inference, to generate a knob recommendation, {right arrow over (k)}.sub.recommended. An optimization step may be applied to force the model output to be as close as possible to the target parameter. A recommendation {right arrow over (k)}.sub.recommended may then be achieved by minimizing a distance metric, D. The metric D is a difference between 1) a prediction of pattern parameters made by the model (indicated below as {circumflex over (P)}), which is based on the pre-process spectra {right arrow over (S)}.sub.Pre and the knob setting, {right arrow over (k)}, and 2) the target value of the pattern parameter {right arrow over (P)}.sub.target. The value of {right arrow over (k)}.sub.recommended may then be deduced by the equation:
PNG
media_image1.png
48
611
media_image1.png
Greyscale
”].
Regarding Claim 13, the combination would disclose training the post-process measurement model based on the second estimated values of the second parameter of interest characterizing the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers in the post-process state [See Fig. 5 and Paragraph [0050] of Bringoltz, per application to reference wafers in light of Banna – “The second part of the ML 500 is a metrology interpretation function, meaning a calibration of the network output (the post-process scatterometric data) to the wafer parameters to be controlled. This calibration, typically by an OCD model 570, uses a second loss term and makes use of labelled data. A second loss term 545 may be set as a difference between the predicted parameters (the output of the OCD model 570) and measured target parameters.”] and an amount of raw measurement data associated with measurements of the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers in the post-process state by the one or more in-line measurement systems [See Fig. 5 and Paragraph [0049] of Bringoltz, per application to reference wafers in light of Banna – “FIG. 5 depicts a process of training a machine learning model 500 that directly transforms a representation of wafer patterns, before and after a process step, and then calibrates the post-process signal (i.e., scatterometric data, {right arrow over (S)}.sub.Post) to the target (controlled) parameter, {right arrow over (P)}.sub.Post. … The output is set to the set of post-process scatterometric data {right arrow over (S)}.sub.Post. A loss function 540 may be set as the difference between scatterometric data predicted by the network and a measured set of post-process scatterometric data.”].
Regarding Claim 14, Bringoltz fails to disclose weighing the second estimated values of the second parameter of interest characterizing the structure at each of the one or more reference measurement sites disposed on the second plurality of semiconductor wafers in the post-process state differently than the first estimated values of the second parameter of interest characterizing the structure at each of the plurality of measurement sites disposed on the first plurality of semiconductor wafers in the post-process state during the training.
However, David discloses appropriately weighting estimated values based their confidence [Paragraph [0090] – “The score generated by the model will correspond to whatever metric was used as a target for training the algorithm that generated the model. For example, if a DBO measurement was used for the target to train the algorithm, then the score will be a predicted DBO measurement. If the target was a parametric test value, then the score will be a prediction of that parametric test value.”Paragraph [0091] – “In addition to the score, the model can also output a confidence metric that describes how reliable the score prediction is. This can be useful in determining whether or not to employ the score, or weight the use of that prediction in conjunction with other traditional measurements.”]. It would have been obvious to weight the reference predictions higher than the first estimated values because the reference wafers of Banna represent “golden” ground truth values, doing so would have improved the machine learning process.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Restaino et al., Optimization of Pre and Post Recipe Sensitivity for Unpatterned Wafer Defectivity Inspection, IEEE, 2016
US 20190004504 A1 – SYSTEMS AND METHODS FOR PREDICTING DEFECTS AND CRITICAL DIMENSION USING DEEP LEARNING IN THE SEMICONDUCTOR MANUFACTURING PROCESS
US 20230196189 A1 – MEASUREMENT METHOD AND APPARATUS FOR SEMICONDUCTOR FEATURES WITH INCREASED THROUGHPUT
US 20230197534 A1 – Neural Network Based Prediction Of Semiconductor Device Response
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KYLE ROBERT QUIGLEY whose telephone number is (313)446-4879. The examiner can normally be reached 9AM-5PM EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Arleen Vazquez can be reached at (571) 272-2619. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/KYLE R QUIGLEY/Primary Examiner, Art Unit 2857
Prosecution Insights