DETAILED ACTION
This office action addresses Applicant’s response filed on 27 July 2026. Claims 1, 4, 6, 9, 12, 14, 16, 19, and 22-31 are pending.
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Objections
Claim 19 is objected to because of the following informalities: the “is trained” following “wherein the first machine learning model” appears to have been inadvertently deleted, such that the claim reads “wherein the first machine learning model using a dataset” instead of “wherein the first machine learning model is trained using a dataset”, as in analogous claims 4 and 12. Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 24, 27, and 29 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
New claims 24, 27, and 29 recite that “a model type associated with the second machine learning model is different from a model type associated with the first machine learning model”, which is not adequately supported by the original disclosure. The closest disclosure is ¶106 of the Specification, which states that the first and second models are different because they use different data, rather than differing in ‘model type’.
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, 6, 9, 14, 16, 22-24, 26, 27, 29, and 30 is/are rejected under 35 U.S.C. 103 as being unpatentable over Yennie (US 2019/0286075) and/or Werkman (US 2021/0405544) in view of Xiong (CN 109543731).
Regarding claim 1, Yennie discloses a method comprising: obtaining spectral data associated with a first portion of a first prior substrate at a manufacturing system and additional spectral data associated with at least one of a second portion of the first prior substrate or a third portion of a second prior substrate at the manufacturing system and identifying one or more measured metrology measurement values associated with the at least one of the second portion of the first prior substrate or the third portion of the second prior substrate (¶68);
providing the obtained spectral data associated with the first portion of the first prior substrate and contextual data associated with the first prior substrate as an input to a first machine learning model obtaining one or more outputs of the first machine learning model, the one or more outputs comprising a plurality of candidate metrology measurement values (¶¶86-88, 90, 93, 96);
generating training data for training a second machine learning model to predict metrology measurement values of a current substrate processed at the manufacturing system (¶¶86, 96), wherein generating the training data comprises: generating a first training data set comprising the spectral data associated with the first portion of the first prior substrate and the selected candidate metrology measurement value (¶¶86, 96); and generating a second training data set comprising the additional spectral data associated with the at least one of a second portion of the first prior substrate or a third portion of a second prior substrate at the manufacturing system and the one or more measured metrology measurement values (¶¶68, 77, 94); and
providing the training data to train the second machine learning model on the first training data set and the second training data set (¶97), wherein upon receiving current spectral data associated with the current substrate, the trained second machine learning model is configured to predict a current metrology measurement value for the current substrate (¶¶98-99).
In the interest of compact prosecution, Werkman also provides further disclosure of obtaining spectral data associated with a first portion of a first prior substrate at a manufacturing system and additional spectral data associated with at least one of a second portion of the first prior substrate or a third portion of a second prior substrate at the manufacturing system and identifying one or more measured metrology measurement values associated with the at least one of the second portion of the first prior substrate or the third portion of the second prior substrate (¶¶4, 59, 65, 75);
providing the obtained spectral data associated with the first portion of the first prior substrate and contextual data associated with the first prior substrate as an input to a first machine learning model obtaining one or more outputs of the first machine learning model, the one or more outputs comprising a plurality of candidate metrology measurement values (¶¶4, 64-73);
generating training data for training a second machine learning model to predict metrology measurement values of a current substrate processed at the manufacturing system (¶80), wherein generating the training data comprises: generating a first training data set comprising the spectral data associated with the first portion of the first prior substrate and the selected candidate metrology measurement value (¶¶65, 80); and generating a second training data set comprising the additional spectral data associated with the at least one of a second portion of the first prior substrate or a third portion of a second prior substrate at the manufacturing system and the one or more measured metrology measurement values (¶¶65, 80); and
providing the training data to train the second machine learning model on the first training data set and the second training data set (¶80), wherein upon receiving current spectral data associated with the current substrate, the trained second machine learning model is configured to predict a current metrology measurement value for the current substrate (¶¶59, 79).
It would have been obvious to persons having ordinary skill in the art before the effective filing date of the application to combine the teachings of Yennie and Werkman, because doing so would have involved merely the routine combination of known elements according to known techniques, or the routine use of a known technique to improve similar devices in the same way, to produce merely the predictable results of generating additional training data from spectral and contextual information. KSR Int’l Co. v. Teleflex Inc., 82 U.S.P.Q.2d 1385, 1395-1396. Yennie discloses training data including spectral data and corresponding metrology measurements, and generating additional training data using spectral and contextual data. Werkman provides additional disclosure of a machine learning model for generating synthetic training data from spectral data and additional contextual data so that the training data includes both the synthetic training data and the original training data that was also used to generate the synthetic training data. As noted above, this architecture is also disclosed by Xiong. The teachings of Werkman are directly applicable to Yennie in the same way, so that Yennie would similarly use a machine learning model to generate synthetic training data from spectral and contextual data, in order to add improved synthetic training data to the training data set.
Yennie and/or Werkman do not appear to explicitly disclose the first machine learning model output comprising, for each respective candidate metrology measurement value, an indication of a level of confidence that the respective candidate metrology measurement value is associated with the first portion of the first prior substrate; and selecting a candidate metrology measurement value of the plurality of candidate metrology measurement values having the level of confidence that satisfies one or more confidence criteria; first training data set comprising the selected candidate metrology measurement value. Xiong discloses these limitations (Fig. 1, first model f1 generates pseudo-labeled sample set S1, which is confidence-filtered to pseudo-labeled sample set S2, which is combined with the labeled data set S to train final model f; ¶¶11, 15). Furthermore, Xiong also provides additional disclosure of the overall topology of using a first machine learning model (¶11, Gaussian process regression model f1) to generate additional training data for a second machine learning model (Fig. 1, final model f).
It would have been obvious to persons having ordinary skill in the art before the effective filing date of the application to combine the teachings of Yennie and/or Werkman and Xiong, because doing so would have involved merely the routine combination of known elements according to known techniques, or the routine use of a known technique to improve similar devices in the same way, to produce merely the predictable results of augmenting training data using high-confidence synthetic data. KSR Int’l Co. v. Teleflex Inc., 82 U.S.P.Q.2d 1385, 1395-1396. Yennie and Werkman disclose generating additional training data for a machine learning model. Xiong teaches that the additional training data should be the data generated by a machine learning model that has high confidence. The teachings of Xiong are directly applicable to Yennie and/or Werkman in the same way, so that Yennie and/or Werkman would similarly use high-confidence data generated by a machine learning model as additional training data, to augment existing training data with high-confidence synthetic training data.
Regarding claim 6, Yennie discloses that the contextual data associated with the first prior substrate comprises at least one of one or more first coordinates associated with the first portion of the first prior substrate, a substrate process performed for the first prior substrate, a time period during which the substrate process was performed for the first prior substrate, a time period during which the spectral data for the first portion of the first prior substrate was collected, or an indication of one or more types of equipment used to perform the substrate process (¶¶75, 88, 91, 93). Werkman also discloses the same (¶¶66-73). Motivation to combine remains consistent with claim 1.
Regarding claim 22, Yennie discloses obtaining contextual data associated with the at least one of the first prior substrate or the second prior substrate, the obtained contextual data comprising one or more of an indication of a substrate process previously performed for the at least one of the first prior substrate or the second prior substrate, an indication of a time period during which the substrate process was performed, an indication of a time period during which the spectral data associated with the first portion of the first prior substrate or the at least one of the second portion of the first prior substrate or the third portion of the second prior substrate was obtained, or an indication of one or more types of equipment associated with the substrate process that was performed, wherein the first training data set further comprises the obtained contextual data (¶¶75, 88, 91, 93). Werkman also discloses the same (¶¶66-73). Motivation to combine remains consistent with claim 1.
Regarding claim 23, Yennie and/or Werkman do not appear to explicitly disclose that the one or more confidence criteria comprise the level of confidence exceeding a threshold level of confidence; Xiong discloses these limitations (¶15). Motivation to combine remains consistent with claim 1.
Regarding claim 24, Yennie discloses that a model type associated with the second machine learning model is different from a model type associated with the first machine learning model (¶¶82, 87, 96, 97). Werkman also discloses the same (¶¶73, 78). Motivation to combine remains consistent with claim 1.
Claims 9, 14, 26, and 27 are directed to a system comprising a memory and a processing device coupled to the memory, the processing device to perform the methods of claims 1, 6, 23, and 24, and are rejected under the same reasoning. Yennie (¶102) and/or Werkman (¶9) disclose a system comprising a memory and a processing device coupled to the memory for performing the claimed methods.
Claims 16, 29, and 30 are directed to non-transitory computer readable storage media for performing the methods of claims 1, 23, and 24, and are rejected under the same reasoning. Yennie (¶102) and/or Werkman (¶9) disclose a non-transitory computer readable storage media for performing the claimed methods.
Claim(s) 4, 12, and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Yennie in view of Werkman, Xiong, and Xie (“Self-training with Noisy Student improves ImageNet classification”).
Regarding claims 4, 12, and 19, Yennie discloses that the first machine learning model is trained using a dataset comprising the additional spectral data associated with the at least one of the second portion of the first prior substrate or the third portion of the second prior substrate and the one or more measured metrology measurement values (¶¶68, 87, 88, 90). These limitations are also taught by Werkman (¶¶4, 64, 65, 73, 80), and Xiong (Fig. 1, training model f1 using labeled data to generate synthetic data S1/S2; ¶11) as applied to Yennie and/or Werkman. Specifically, Werkman and Xiong disclose training regression models using the original labeled data set, which in Yennie’s or Werkman’s context is the spectral data and associated metrology measurement values. Xiong’s and Werkman’s final training data is a hybrid data set including both the synthetic data generated by the regression model and the original labeled data that the regression model was trained on; the synthetic data in Yennie and Werkman are the spectral data without corresponding metrology measurements (unlabeled) + generated (synthetic) metrology measurements, and the original labeled data are the spectral data + measured metrology values.
Nevertheless, if Yennie, Werkman, and Xiong are found to be unclear regarding these limitations, Xie discloses the same (p. 1, col. 2, ¶1; teacher is trained on labeled data, teacher generates pseudo labels on unlabeled data, and student is trained on combination of labeled data and pseudo-labeled data). The application of Xie to Yennie and/or Werkman suggests training the first machine learning model on Yennie’s and/or Werkman’s labeled data (spectral + corresponding metrology measurement), using the first machine learning model to generate pseudo labels (predicted metrology measurement) for unlabeled data (spectral data with no corresponding metrology measurement), and then training the second machine learning model using the labeled (spectral + metrology measurement) and pseudo-labeled (spectral + metrology prediction) data.
It would have been obvious to persons having ordinary skill in the art before the effective filing date of the application to combine the teachings of Yennie and/or Werkman, Xiong, and Xie, because doing so would have involved merely the routine combination of known elements according to known techniques, or the routine use of a known technique to improve similar devices in the same way, to produce merely the predictable results of training a model to produce usable synthetic training data. KSR Int’l Co. v. Teleflex Inc., 82 U.S.P.Q.2d 1385, 1395-1396. Yennie, Xiong, and Werkman disclose a first machine learning model that generates predicted metrology values for spectral data, which is used as additional training data for a second machine learning model. Persons having ordinary skill in the art would understand Yennie, Xiong, and Werkman as disclosing the training the first machine learning model using spectral and metrology data; Xie provides additional explicit disclosure of the same. The teachings of Xie are directly applicable to Yennie and/or Werkman in the same way, so that Yennie and/or Werkman would train the first machine learning model using labeled (spectral + metrology measurement) data, to build a model that produces usable synthetic training data for a metrology model.
Claim(s) 25, 28, and 31 is/are rejected under 35 U.S.C. 103 as being unpatentable over Yennie in view of Werkman, Xiong, and Lindeman (US 2021/0081698).
Regarding claims 25, 28, and 31, Yennie does not appear to explicitly disclose updating the trained machine learning model such that one or more model weights associated with the one or more measured metrology measurement values are higher than one or more additional weights associated with the selected candidate metrology measurement value. Lindeman discloses these limitations (¶80). Lindeman also discloses first and second data sets (¶¶79-80). It would have been obvious to persons having ordinary skill in the art before the effective filing date of the application to combine the teachings of Yennie and/or Werkman, Xiong, and Lindeman, because doing so would have involved merely the routine use of a known technique to improve similar devices in the same way to achieve the predictable results of emphasizing measured data over synthetic data. KSR Int’l Co. v. Teleflex Inc., 82 U.S.P.Q.2d 1385, 1396. Yennie, Xiong, and Werkman disclose training data for a machine learning model including labeled (spectral + metrology measurement) data and pseudo-labeled/synthetic (spectral + metrology prediction) data. Lindeman teaches that real and synthetic datasets are weighted in the machine learning model to emphasize one over the other. The teachings of Lindeman are directly applicable to Yennie, Xiong, and Werkman, so that Yennie would similarly weight real and synthetic datasets to emphasize measured data over synthetic data.
Response to Arguments
Applicant’s arguments have been considered but are moot in view of the new grounds of rejection. Applicant asserts that the prior art fails to teach newly-added limitations, which are addressed above using newly-cited prior art.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ARIC LIN whose telephone number is (571)270-3090. The examiner can normally be reached M-F 07:30-17:00 ET.
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, Jack Chiang can be reached at 571-272-7483. 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.
3 October 2026
/ARIC LIN/ Examiner, Art Unit 2851 /JACK CHIANG/ Supervisory Patent Examiner, Art Unit 2851
