DETAILED ACTION
This nonfinal rejection is responsive to the amendment filed on February 26, 2026. Claims 1-3, 5-6, 9-11, 14-17, and 20. Claims 1, 9, and 15 are independent. Claims 4, 7, 8, 12, 13, 18, and 19 are canceled.
Claim rejections under 35 USC §103 are maintained. See sections Claim Rejections – 35 USC §103 and Response to Arguments below.
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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on February 26, 2026 has been entered.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on July 7, 2026 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 103
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 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.
Claims 1, 3, 5-6, 9-11, and 15-17 are rejected under 35 U.S.C. 103 as being unpatentable over Firth et al. (US20030055524), hereinafter Firth, in view of Biafore et al. (US20180275523), hereinafter Biafore.
Regarding claim 1, Firth teaches the method:
applying input data comprising an input set of process recipe settings for a process used to process a component in order to manufacture an electronic device, (Firth, paragraph 0009: “The system includes a controller which is adapted to receive context data, process data and numeric data regarding the semiconductor fabrication process,”) output data representative of defect impact with respect to at least one defect type of a set of defect types, wherein the output data comprises an output set of process recipe settings for the process that minimizes at least one defect impact metric with respect to the at least one defect type of the set of defect types in accordance with the input data, and (Firth, paragraph 0019: “As described more fully and completely below, the observer portion or module 12 is adapted to estimate the actual state of the fabrication process, to estimate errors that are correlated with a plurality of contributing context items (i.e., components that contribute to the overall error, such as tool error, reticle error, reference tool error and reference reticle error), and to generate updated process state information… The “control law” of portion 14 utilizes the error information from observer portion 12 to determine how the fabrication process or “recipe” should be modified in order to keep the process (e.g., the output of the process or the fabricated devices) on target.” And paragraph 0044: “In the foregoing manner, the present invention accurately estimates the error within a semiconductor manufacturing process and attributes each portion of the error to a specific contributing context item. Hence, the present system and method offer insight into which context items are varying, thereby providing specific guidance for improving the manufacturing process and eliminating variation” – The estimated errors is analogous to the defect impact, wherein the error estimates correlated with a plurality of contributing context items is analogous to the defect type. The updated process state information being generated based on the estimated errors is analogous to the output minimizing the defect impact metric, e.g., the estimated errors for the context items.) wherein the input data further comprises at least one of: a set of desired characteristics to be achieved by processing the component, or a set of constraints specifying one or more allowable recipe setting ranges for the input set of recipe settings; (Firth, paragraph 0019: “The “process model” of portion 14 relates measurable inputs and states to the desired product qualities. Particularly, the process model is used in the control law to determine which recipe parameters, or process inputs, should be adjusted to provide the desired output.” – The measurable inputs and desired product qualities is analogous to the set of desired characteristics to be achieved.)
receiving, by the processing device, the input data; (Firth, paragraph 0025: “In functional block or step 36, the observer portion 12 receives data describing the particular processes that the lot of wafers encounters during fabrication. This process data may include information regarding tool and/or machine settings and operating points existing during the fabrication process, such as temperature and pressure settings, spin speeds and other tool and/or machine operating parameters.” – The data being received is analogous to the input data.)
obtaining, by the processing device, the output data by applying the input data to the machine learning model; (Firth, paragraph 0039: “Equation (8) is solved at each iteration of the control algorithm and in functional block or step 42, observer portion 12 communicates this error information to portion 14 of the control system 10.” And paragraph 0040: “The control law calculates the recommended process input which is utilized within the manufacturing process 16.” – The equation solving for the error information and then utilizing the error information to calculate new process inputs is analogous to the output data being obtained by applying the input data to the machine learning model, where the control algorithm is analogous to the machine learning model.)
generating, by the processing device based on the output data, a process recipe for performing the process that accounts for the defect impact with respect to the at least one defect type of the set of defect types; and causing, by the processing device, a process tool to process the component in accordance with the process recipe. (Firth, paragraph 0002: “The manufacturing equipment and tools used within these fabs, such as steppers and scanners, have a multitude of parameters which must be examined and controlled in order to minimize these variations and errors.” And paragraph 0020: “The output from the control law and process model portion 14 (i.e., the process inputs) is communicated to the fabrication process 16 (e.g., to the manufacturing tools performing the fabrication process). The fabrication process 16 is then modified in accordance with the information or process inputs received from portion 14.” – Firth teaches the output data being the defect impact and the process recipe above, therefore, the parameters being controlled in order to minimize the errors is analogous to the process recipe that accounts for the defect impact with respect to at least one defect type. While the output being communicated to the fabrication process is analogous to causing the process tool to process the component in accordance with the process recipe.)
Firth does not explicitly teach:
identifying, by a processing device, a machine learning model selected from a plurality of trained machine learning models, wherein the machine learning model is trained for obtaining,
However, Biafore teaches:
identifying, by a processing device, a machine learning model selected from a plurality of trained machine learning models, (Biafore, paragraphs 0071 and 0082 discuss different algorithms that can be used and therefore teaches selecting a model from a plurality of models.) wherein the machine learning model is trained for obtaining, (Biafore, paragraph 0005: “In another illustrative embodiment, the controller receives a production recipe including at least a pattern of elements to be fabricated on a sample and one or more exposure parameters for exposing the pattern of elements with illumination during fabrication of the sample” – The controller is a part of the semiconductor device system, see e.g. Fig. 1A, and therefore is analogous to the processing device. The controller receiving the production recipe is analogous to the input data comprising the input set of recipe settings.)
Biafore is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Firth, which already teaches the method of estimating defects in a process and generating a process recipe to minimize the impact of the defects but does not explicitly teach that the machine learning model is selected from a plurality of trained machine learning models, to include the teachings of Biafore which does teach that the machine learning model is selected from a plurality of trained machine learning models in order to meet “a desired balance of precision and computational requirements.” (Biafore, paragraph 0071)
Regarding claim 3, Firth and Biafore teach the method of claim 1, as cited above.
Firth does not explicitly teach:
receiving, by the processing device, an initially trained machine learning model from the plurality of trained machine learning models; receiving, by the processing device, tuning input data; and tuning, based on the tuning input data, the initially trained machine learning model to obtain a tuned machine learning model.
Biafore further teaches:
receiving, by the processing device, an initially trained machine learning model from the plurality of trained machine learning models; receiving, by the processing device, tuning input data; and tuning, based on the tuning input data, the initially trained machine learning model to obtain a tuned machine learning model. (Biafore, paragraph 0085: “In another embodiment, the care areas selected in step 206 may be fed back to improve and/or train a stochastic simulation model (e.g. a stochastic simulation model used in step 204 and/or step 206) to improve predictive accuracy and/or efficiency.” – The stochastic simulation model used in step 204 and/or step 206 indicates that the initially trained model was received, the care areas selected in step 206 being fed back into the model to improve the model is analogous to the care areas being tuning input data and the model being improved is analogous to tuning the initially trained model based on the tuning input data.)
Regarding claim 5, Firth and Biafore teach the method of claim 1, as cited above.
Firth further teaches:
wherein the output data further comprises the at least one defect impact metric, and wherein the at least one defect impact metric comprises at least one of: an estimated defect count for the at least one defect type of the set of defect types, or a probability that the at least one defect type of the set of defect types will impact performance. (Firth, paragraph 0008: “which estimates values for each of the error components based upon manufacturing feedback data; and a second portion which communicates with the first portion and which receives the estimated values from the first portion and utilizes the estimated values to control the manufacturing process.” – The estimated values for the errors being received indicates that the estimated values are output data where the estimated values are analogous to the estimated defect count, see e.g., Equation 2 in paragraph 0029 which shows the estimated values to be a summation.)
Regarding claim 6, Firth and Biafore teach the method of claim 1, as cited above.
Firth does not explicitly teach:
at least one defect impact metric comprises at least one of: an estimated defect count for the at least one defect type, or a probability that the at least one defect type of the set of defect types will impact performance.
However, Biafore further teaches:
at least one defect impact metric comprises at least one of: an estimated defect count for the at least one defect type, or a probability that the at least one defect type of the set of defect types will impact performance. (Firth, paragraph 0008: “which estimates values for each of the error components based upon manufacturing feedback data; and a second portion which communicates with the first portion and which receives the estimated values from the first portion and utilizes the estimated values to control the manufacturing process.” – The estimated values for the errors are analogous to the estimated defect count, see e.g., Equation 2 in paragraph 0029 which shows the estimated values to be a summation.)
Regarding claim 9, Claim 9 has all the same limitations of claim 1 which are taught by Firth and Biafore – see claim 1 above.
Firth does not explicitly teach:
a memory and a processing device, operatively coupled to the memory, to perform operations comprising: (Biafore, paragraph 0034: “In another embodiment, the controller 106 includes one or more processors 108 configured to execute program instructions maintained on a memory device 110.”)
Regarding claim 10, Firth and Biafore teach the system of claim 9, as cited above.
Claim 10 additionally has the same limitations of claim 5 which are taught by Firth and Biafore – see claim 5 above.
Regarding claim 11, Firth and Biafore teach the system of claim 9, as cited above.
Claim 11 additionally has the same limitations of claim 6 which are taught by Firth and Biafore – see claim 6 above.
Regarding claim 15, Claim 15 has all the same limitations of claim 1 which are taught by Firth and Biafore – see claim 1 above.
Firth does not explicitly teach:
A non-transitory machine-readable storage medium storing instructions which, when executed by a processing device, cause the processing device to perform operations comprising:
However, Biafore further teaches:
A non-transitory machine-readable storage medium storing instructions which, when executed by a processing device, cause the processing device to perform operations comprising: (Biafore, paragraph 0036: “The memory device 110 may include any storage medium known in the art suitable for storing program instructions executable by the associated one or more processors 108. For example, the memory device 110 may include a non-transitory memory medium.”)
Regarding claim 16, Firth and Biafore teach the non-transitory machine-readable storage medium of claim 15, as cited above.
Claim 16 additionally has the same limitations of claim 5 which are taught by Firth and Biafore – see claim 5 above.
Regarding claim 17, Firth and Biafore teach the non-transitory machine-readable storage medium of claim 15, as cited above.
Claim 17 additionally has the same limitations of claim 6 which are taught by Firth and Biafore – see claim 6 above.
Claims 2, 14, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Firth in view of Biafore in view of David (US20160148850), hereinafter David.
Regarding claim 2, Firth and Biafore teach the method of claim 1, as cited above.
Firth does not explicitly teach:
receiving, by the processing device, training input data associated with the process, wherein the training input data comprises a set of experimental data related to the process;
obtaining, by the processing device, target output data for the training input data, wherein the target output data identifies the set of defect types; and
providing, by the processing device, the training input data and the target output data to train the plurality of trained machine learning models, wherein providing the training input data comprises converting the training input data into defect model training data having a machine learning format for training the plurality of trained machine learning models.
However, Biafore further teaches:
the training input data comprises a set of experimental data related to the process; (Biafore, paragraph 0082: “Accordingly, a probability of stochastic repeaters in a candidate care area may be analyzed based on a comparison to training data that may be, but is not required to be, generated by simulations and/or measurements of fabricated samples.” – The training data being generated by simulations and/or measurements of fabricated samples is analogous to the training input data comprising a set of experimental data.)
Firth and Biafore do not explicitly teach:
receiving, by the processing device, training input data associated with the process, …;
obtaining, by the processing device, target output data for the training input data, wherein the target output data identifies the set of defect types; and
providing, by the processing device, the training input data and the target output data to train the plurality of trained machine learning models, wherein providing the training input data comprises converting the training input data into defect model training data having a machine learning format for training the plurality of trained machine learning models.
However, David teaches:
receiving, by the processing device, training input data associated with the process, (David, paragraph 0043: “Based on sensor data from production equipment and actual metrology values of sampled wafers to train the algorithm, virtual metrology can predict metrology values for all wafers. The algorithm can be a supervised learning algorithm, where a model can be trained using a set of input data and measurement targets.” – The input data being used to train the model is analogous to receiving training input data.)
obtaining, by the processing device, target output data for the training input data, wherein the target output data identifies the set of defect types; and (David, paragraph 0065: “The target could be parametric data such as on/off current of the transistor, transistor thresholds, or some other parameter that quantifies the health of the transistor. The target could also be yield information, such as the functionality of a given die or area on the wafer (sometimes measured as either pass or fail). The target could also be semiconductor device performance data.” And paragraph 0066: “Every set of input data is associated with a specific output or target.” – The target is analogous to the target output data with the functionality or performance data being indicative of the target including a set of defect types.)
providing, by the processing device, the training input data and the target output data to train the plurality of trained machine learning models, (David, paragraph 0082: “In step 612, the data is then fed into the algorithm for training. The algorithm could be one of many different types of algorithms.” And paragraph 0113: “The training data can be partitioned into training, testing, and validation portions to ensure a robust model is build that is not over-fit or over-biased.” – The data used for training the algorithm or model necessarily includes both input data as well as target output data as that is what the model is trained on (see above). Thus, the training dataset includes both the input data and target output data.) wherein providing the training input data comprises converting the training input data into defect model training data having a machine learning format for training the plurality of trained machine learning models. (David, paragraph 0045: “Data transformation or feature engineering can be performed on in-situ spectral data or other sensor data that is collected during a particular process such as etch, deposition, or CMP.” And paragraph 0066: “Those values would be an input vector to the model, and would be associated with the target, e.g., the measured offset. If there are n input variables, then the input vector size for each target would be 1xn. Therefore, if there are m targets, there would be an input data matrix of size mxn, with each row of the input data matrix associated with a target. This is a typical training set in matrix format for a machine learning algorithm.” – Transforming the training data and presenting it in matrix format for the machine learning algorithm is analogous to converting the training input data into a format for training the models.)
David is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Firth and Biafore, which already teaches using a machine learning model to predict defects and generate a process recipe based off the predicted defects but does not explicitly teach training the machine learning model using training input data and target output data, to include the teachings of David which does teach training the machine learning model using training input data and target output data in order to apply predictive analytics "to increase yield, improve device performance, and reduce costs like never before." (David, paragraph 0041)
Regarding claim 14, Firth and Biafore teach the system of claim 9, as cited above.
Firth does not explicitly teach:
the operations further comprise training the machine learning model based on training input data and target output data, and wherein the training input data comprises a set of experimental data related to the process.
However, Biafore further teaches:
wherein the training input data comprises a set of experimental data related to the process. (Biafore, paragraph 0082: “Accordingly, a probability of stochastic repeaters in a candidate care area may be analyzed based on a comparison to training data that may be, but is not required to be, generated by simulations and/or measurements of fabricated samples.” – The training data being generated by simulations and/or measurements of fabricated samples is analogous to the training input data comprising a set of experimental data.)
Firth and Biafore do not explicitly teach:
the operations further comprise training the machine learning model based on training input data and target output data,
However, David teaches:
the operations further comprise training the machine learning model based on training input data and target output data, (David, paragraph 0082: “In step 612, the data is then fed into the algorithm for training. The algorithm could be one of many different types of algorithms.” And paragraph 0113: “The training data can be partitioned into training, testing, and validation portions to ensure a robust model is build that is not over-fit or over-biased.” – The data used for training the algorithm or model necessarily includes both input data as well as target output data as that is what the model is trained on, see e.g. paragraph 0065. Thus, the training dataset includes both the input data and target output data.)
David is considered analogous to the claimed invention as it is in the same field of endeavor, machine learning. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to have modified Firth and Biafore, which already teaches using a machine learning model to predict defects and generate a process recipe based off the predicted defects but does not explicitly teach training the machine learning model using training input data and target output data, to include the teachings of David which does teach training the machine learning model using training input data and target output data in order to apply predictive analytics "to increase yield, improve device performance, and reduce costs like never before." (David, paragraph 0041)
Regarding claim 20, Firth and Biafore teach the non-transitory machine-readable storage medium of claim 15, as cited above.
Claim 20 additionally has the same limitations of claim 14 which are taught by Firth, Biafore, and David – see claim 14 above.
Response to Arguments
Applicant's arguments filed February 26, 2026 have been fully considered but they are not persuasive. Applicant’s arguments regarding Biafore are moot as Biafore is no longer relied upon to teach the limitations specifically challenged in the argument. Regarding applicant’s arguments that Firth is a different technical problem, examiner disagrees. Firth states in paragraph 0019: “As described more fully and completely below, the observer portion or module 12 is adapted to estimate the actual state of the fabrication process, to estimate errors that are correlated with a plurality of contributing context items (i.e., components that contribute to the overall error, such as tool error, reticle error, reference tool error and reference reticle error), and to generate updated process state information.” Thus, Firth teaches a method that uses inputs to generate the predicted defects (errors) and further updating the process recipe (process state information) in order to minimize the defects. This is done using the process model/algorithm (see e.g. paragraph 0027) which is analogous to the machine learning model. Further, examiner notes that there is nothing in the claimed invention that limits it to being performed “proactively” before the production process rather than during the production process.
Examiner disagrees that one of ordinary skill in the art would have no motivation to combine the teachings of Firth and Biafore as they are directed to the same area of reducing errors/defects in the manufacturing of semiconductors. Firth’s error estimation is predictive as it is estimating errors in future iterations, see paragraphs 0031-0032. Thus, it is providing the estimated errors and process parameter adjustment to minimize those errors. Likewise, Biafore is predicting defects and providing adjustments to account for those defects. The motivation to combine the references was provided in the previous office action for the rejection of Biafore in view of Firth and is provided above for the rejection of Firth in view of Biafore. Paragraph 0071 of Biafore discusses using different algorithms/models in order to account for the “desired balance of precision and computational requirements.”
Claim rejections under 35 USC §103 of claims 1-3, 5-6, 9-11, 14-17, and 20 are maintained, see section Claim Rejections – 35 USC §103 above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Sawlani et al. (US20200226742)
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JACQUELINE MEYER whose telephone number is (703)756-5676. The examiner can normally be reached M-F 8:00 am - 4:30 pm 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, Tamara Kyle can be reached at 571-272-4241. 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.
/J.C.M./Examiner, Art Unit 2144
Prosecution Insights