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 .
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 03/24/2026 has been entered.
Claim Status
Claims 1, 4, 10, 13, and 16-17 have been amended. Claims 1-20 remain pending and are ready for examination.
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.
Claim(s) 1-2, 10-11, and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chau et al. (US20220171373A1 -hereinafter Chau) in view of Denome et al. (US20190155265A1 -hereinafter Denome).
Regarding Claim 1, Chau teaches a method comprising:
receiving a first process recipe comprising first process recipe setpoint data; (see [0094]; Chau: “The processing chamber controllers 130 associated with the processing chambers 104 generally follow a recipe that specifies the timing of steps, process gases to be supplied, temperature, pressure, RF power, and so on”)
inputting the first process recipe into a set of chained models; (see [0006]; Chau: “A system for processing semiconductor substrates in a tool comprising a plurality of processing chambers configured to process the semiconductor substrates according to a recipe… The instructions are configured to train a model using the first data and data generated by the simulation to predict optimum scheduling parameters for processing the semiconductor substrates in the plurality of processing chambers according to the recipe.” See [0189] and Fig. 14: “The scheduler level neural network is shown as 1412 and receives outputs of the neural networks 1410 as inputs.”) [The nested neural network reads on ‘a set of chained models’]
receiving, as output from the set of chained models, predicted first ecological efficiency data… (see Fig. 14 and [0182]; Chau: “The method can recommend recipe/wafer assignment mix with maximum tool utilization.”)
wherein a first model of the set of chained models outputs predicted measurement data associated with processing the substrate in the process chamber according to the first process recipe responsive to receiving the first process recipe as input (see [0189]; Chau: “The plurality of module level neural networks include one neural network for each processing module (e.g., processing modules 1602 shown in FIG. 16) in the tool 1406 and one neural network for each robot (e.g., for robots 1610 and 1614 shown in FIG. 16) in the tool 1406. These neural networks are shown as 1410-1, ..., and 1410-N, where N is an integer greater than 1, may be collectively called neural networks 1410.”) [The module level neural network reads on ‘the first model’], and wherein a second model of the set of chained models receives the predicted measurement data as input and outputs the predicted first ecological efficiency data based on the predicted measurement data; (see [0189]; Chau: “The scheduler level neural network is shown as 1412 and receives outputs of the neural networks 1410 as inputs.” See [0185]: “. That is, the nested neural network based dynamic scheduler, which is trained using different tool hardware configurations and different recipe types, can now recommend optimum tool hardware configuration for a given recipe or recipes.”) [The scheduler level neural network reads on ‘the second model’]
However, Chau does not explicitly teach: receiving …predicted first ecological efficiency data indicative of a first ecological efficiency associated with processing a substrate in a process chamber according to the first process recipe; and outputting a recommendation associated with the first process recipe based at least in part on the predicted first ecological efficiency data.
Denome the same or similar field of endeavor teaches:
Receiving …predicted first ecological efficiency data indicative of a first ecological efficiency associated with processing a substrate in a process chamber according to the first process recipe; and (see [0080]; Denome: “eco-efficiency characterizer 216 performs the per-unit eco-efficiency characterization functionality of characterizers 202-208 of FIG. 2A.” See [0034]: “The manufacturing equipment 112, 116 may be semiconductor wafer manufacturing equipment that includes one or more processing chambers.” See [0082]: “Using the utility and utilization information of 210 and 212, eco-efficiency characterizer 216 may determine the eco-efficiency characteristics of the associated manufacturing equipment.”)
outputting a recommendation associated with the first process recipe based at least in part on the predicted first ecological efficiency data. (see [0077]; Denome: “eco-efficiency analyzer may provide design change recommendations based on the comparison, in view of the second eco-efficiency characterization being associated with a higher eco-efficiency characterization than the first characterization.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to modify the teaching of Chau to include Denome’s features of receiving, as output from the one or more models, predicted first ecological efficiency data indicative of a first ecological efficiency associated with processing a substrate in a process chamber according to the first process recipe; and outputting a recommendation associated with the first process recipe based at least in part on the predicted first ecological efficiency data. Doing so would optimize the manufacturing equipment for eco-efficiency. (Denome, [0041])
Regarding Claim 2, the combination of Chau and Denome teaches all the limitations of claim 1 above, Denome further teaches:
determining the recommendation based on a comparison of the predicted first ecological efficiency data and a predicted second ecological efficiency data (see [0077]; Denome: “eco-efficiency analyzer may provide design change recommendations based on the comparison, in view of the second eco-efficiency characterization being associated with a higher eco-efficiency characterization than the first characterization.” See Abstract: “The method further includes comparing the first eco-efficiency characterization to a second eco-efficiency characterization that is associated with a second design of the manufacturing equipment.”), wherein the predicted second ecological efficiency data is indicative of a second ecological efficiency associated with processing the substrate in the process chamber according to a second process recipe. (see [0062]; Denome: “process recipes may be input to the per-unit eco-efficiency characterizations. A process recipe may specify time, power, flow, temperature, etc. to produce the desired result. Such process recipes may affect the eco-efficiency characterization of the manufacturing equipment.” See [0011]: “The processing device of the manufacturing equipment may further implement the adjustment to the one or more settings associated with the second eco-efficiency characterization on the manufacturing equipment.”)
The same motivation to combine Chau and Denome a set forth for Claim 1 equally applies to Claim 2.
Regarding Claim 10, the limitations in this claim is taught by the combination of Chau and Denome as discussed connection with claim 1.
Regarding Claim 11, the limitations in this claim is taught by the combination of Chau and Denome as discussed connection with claim 2.
Regarding Claim 20, the limitations in this claim is taught by the combination of Chau and Denome as discussed connection with claim 1.
Claim(s) 3-5 and 12-14 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chau in view of Denome in view of Dobashi et al. (US20230012173A1 -hereinafter Dobashi)
Regarding Claim 3, the combination of Chau and Denome teaches all the limitations of claim 2 above; however, it does not teach further comprising: receiving target data comprising a target substrate condition for a processed substrate; inputting the target data into one or more additional models; and receiving, as output from the one or more additional models, the first process recipe and the second process recipe.
Dobashi from the same or similar field of endeavor teaches:
receiving target data comprising a target substrate condition for a processed substrate; (see [0009]; Dobashi: “The process recipe search apparatus has a target shape decision unit that decides a target shape that defines the desired shape by a plurality of shape elements,”)
inputting the target data into one or more additional models; and (see [0009]; Dobashi: “a machine learning model creation unit that creates a machine learning model that predicts a process shape of the sample processed by the plasma processing apparatus from the parameter of the plasma processing apparatus, a recipe search unit that uses the machine learning model to search for a candidate etching recipe that becomes a candidate of the etching recipe,”)
receiving, as output from the one or more additional models, the first process recipe and the second process recipe. (see [0066]; Dobashi: “In the choice region 707, the candidate etching recipes (predicted recipes) chosen by the machine learning model and the target shape are displayed.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to modify the teaching of the combination of Chau and Denome to include Dobashi’s features of receiving target data comprising a target substrate condition for a processed substrate; inputting the target data into one or more additional models; and receiving, as output from the one or more additional models, the first process recipe and the second process recipe. Doing so would determine an enormous number of parameters in the etching processing at high accuracy and at high speed. (Dobashi, [0003])
Regarding Claim 4, the combination of Chau, Denome, and Dobashi teaches all the limitations of claim 3 above, Chau further teaches wherein the set of chained models comprise one or more first trained machine learning models, (see 0184; Chau: “In an initial layer of the model, a module level neural network (i.e., a neural network for a processing module) is trained to predict program execution times for different processes.”) and wherein the one or more additional models comprise one or more second trained machine learning model. (see [0185]; Chau: “The input for the scheduler level neural network is expanded to other tool configurations, mix of recipe types, process times, multiple layers to be processed on the wafers, scheduler modes, etc. Coupled (i.e., nested) with the module level neural networks, the scheduler level neural network with expanded inputs provides recommendations for best product/recipe/wafer mix to achieve highest tool/fleet utilization to reduce cost-of-ownership for the tools.”)
Regarding Claim 5, the combination of Chau, Denome, and Dobashi teaches all the limitations of claim 3 above, Dobashi further teaches further comprising:
predicting, via a first additional model of the one or more additional models, one or more first measurements corresponding to the first process recipe; and (see [0077]; Dobashi: “IG. 11 illustrates an example of a GUI displayed on the display device 1322 by the target shape decision unit 1313 for adding the shape element to redefine the target shape and remaking the dataset. At the upper area of a GUI 1101, a choice region 1102 for choosing the etching recipe in which the etching 204 is performed and the process result is acquired by the measurement 205 is disposed.” See [0078]: “FIG. 11 illustrates an example in which the process shapes corresponding to the redefined target shape are displayed in the shape display region 1103. Here, it is assumed that only an opening width W1 and the depth D1 are defined in the unredefined target shape. At the lower area, an existing measurement value table 1107 is displayed, but as is apparent from here, the opening width W1 and the depth D1 of the process results of all the predicted recipes are equal to those of the target shape.” See [0066]: “In the choice region 707, the candidate etching recipes (predicted recipes) chosen by the machine learning model and the target shape are displayed.”)
predicting, via a second additional model of the one or more additional models, one or more second measurements based on the first process recipe and the one or more first measurements output from the first additional model. (see [0077]; Dobashi: “IG. 11 illustrates an example of a GUI displayed on the display device 1322 by the target shape decision unit 1313 for adding the shape element to redefine the target shape and remaking the dataset. At the upper area of a GUI 1101, a choice region 1102 for choosing the etching recipe in which the etching 204 is performed and the process result is acquired by the measurement 205 is disposed.” See [0078]: “FIG. 11 illustrates an example in which the process shapes corresponding to the redefined target shape are displayed in the shape display region 1103. Here, it is assumed that only an opening width W1 and the depth D1 are defined in the unredefined target shape. At the lower area, an existing measurement value table 1107 is displayed, but as is apparent from here, the opening width W1 and the depth D1 of the process results of all the predicted recipes are equal to those of the target shape.” See Fig. 9: the choice region 707 includes Predicted Recipe No. 2)
The same motivation to combine of Chau, Denome, and Dobashi a set forth for Claim 3 equally applies to Claim 5.
Regarding Claim 12, the limitations in this claim is taught by the combination of Chau, Denome, and Dobashi as discussed connection with claim 3.
Regarding Claim 13, the limitations in this claim is taught by the combination of Chau, Denome, and Dobashi as discussed connection with claim 4.
Regarding Claim 14, the limitations in this claim is taught by the combination of Chau, Denome, and Dobashi as discussed connection with claim 5.
Claim(s) 6-7 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chau in view of Denome in view of Dobashi in view of Funk et al. (US20100036514A1 -hereinafter Funk).
Regarding Claim 6, the combination of Chau, Denome, and Dobashi teaches all the limitations of claim 5 above; however, it does not explicitly teach: wherein the one or more first measurements and the one or more second measurements comprise predicted measurements of at least one of current, voltage, power, flow, pressure, concentration, speed, acceleration, or temperature.
Funk from the same or similar field of endeavor teaches wherein the one or more first measurements and the one or more second measurements comprise predicted measurements of at least one of current, voltage, power, flow, pressure, concentration, speed, acceleration, or temperature. (see [0077]; Funk: “The sensors 250 can include both sensors that are intrinsic to the plasma processing chamber 210 and sensors extrinsic to the plasma-processing chamber 210. Intrinsic sensors can include those sensors pertaining to the functionality of plasma processing chamber 210 such as the measurement of the Helium backside gas pressure, Helium backside flow, electrostatic clamping (ESC) voltage, ESC current, wafer holder 220 temperature (or lower electrode (LEL) temperature), coolant temperature, upper electrode (UEL) temperature, forward RF power, reflected RF power, RF self-induced DC bias, RF peak-to-peak voltage, chamber wall temperature, process gas flow rates, process gas partial pressures, chamber pressure, capacitor settings (i.e., C1 and C2 positions), a focus ring thickness, RF hours, focus ring RF hours, and any statistic thereof.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to modify the teaching of the combination of Chau, Denome, and Dobashi to include Funk’s features of the one or more first measurements and the one or more second measurements comprise predicted measurements of at least one of current, voltage, power, flow, pressure, concentration, speed, acceleration, or temperature. Doing so would optimize a process recipe and/or process time and improve the accuracy. (Funk, [0066]-[0067])
Regarding Claim 7, the combination of Chau and Denome teaches all the limitations of claim 1 above, Denome teaches wherein the predicted first ecological efficiency data… (see Abstract; Denome: “A method includes determining, by a processing device, a first eco-efficiency characterization associated with a first design of manufacturing equipment based on one or more of water eco-efficiency characterization, emissions eco-efficiency characterization, or electrical energy eco-efficiency characterization.”)
However, it does not teach: …comprises predicted time series data associated with a predicted behavior of the process chamber during execution of the first process recipe.
Funk from the same or similar field of endeavor teaches …comprises predicted time series data associated with a predicted behavior of the process chamber during execution of the first process recipe. (see Abstract; Funk: “The MLMIMO process control uses dynamically interacting behavioral modeling between multiple layers and/or multiple process steps.” See [0104]: “adaptive feedback can be used when copying a MLMIMO model from chamber to chamber and allowing the MLMIMO model to adapt to the new chamber behavior.”)
The same motivation to combine of Chau, Denome, and Funk a set forth for Claim 6 equally applies to Claim 7.
Claim(s) 8-9 and 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chau in view of Denome in view of Torikoshi et al. (US20220234164A1 -hereinafter Torikoshi).
Regarding Claim 8, the combination of Chau and Denome teaches all the limitations of claim 1 above; however, it does not explicitly teach: wherein the recommendation comprises a modification to the first process recipe to form a modified first process recipe, and wherein processing the substrate according to the modified first process recipe has a reduced environmental resource consumption compared to processing the substrate according to the first process recipe.
Torikoshi from the same or similar field of endeavor teaches wherein the recommendation comprises a modification to the first process recipe to form a modified first process recipe, and wherein processing the substrate according to the modified first process recipe has a reduced environmental resource consumption compared to processing the substrate according to the first process recipe. (see [0012]; Torikoshi: “since the polishing end point timing can be automatically predicted, the time and cost required for predicting the polishing end point timing are reduced, and when an abnormality is present in a time-series change in the physical quantity, the polishing end point timing is automatically corrected by updating the processing condition (recipe). Therefore, since it is not necessary to go to the site to update the recipe, labor, energy, and/or cost can be saved.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to modify the teaching of the combination of Chau and Denome to include Torikoshi’s features of a modification to the first process recipe to form a modified first process recipe, and wherein processing the substrate according to the modified first process recipe has a reduced environmental resource consumption compared to processing the substrate according to the first process recipe. Doing so would provide a substrate processing system capable of saving labor, energy, and/or cost for a substrate processing apparatus. (Torikoshi, [0008])
Regarding Claim 9, the combination of Chau and Denome teaches all the limitations of claim 1 above, Denome further teaches wherein the ecological efficiency data comprises …data for at least one of an energy consumption, a gas consumption, or a water consumption associated with substrate processing in the process chamber. (see [0024]; Denome: “Eco-efficiency may be the amount of environmental resource (e.g., electrical energy, water, gas, etc.) consumed per unit of equipment production.”)
Torikoshi from the same or similar field of endeavor teaches …comprises time series data for at least one of an energy consumption, a gas consumption, or a water consumption associated with substrate processing in the process chamber. (see [0067]; Torikoshi: “As illustrated in FIG. 4, the table T1 stores a record of a set of a lot of a wafer, time-series data of the motor current, time-series data of the flow rate of water or slurry, time-series data of the polishing pressure, time-series data of the rotation speed of the polishing table, time-series data of the rotation speed of the top ring, and the like. As described above, the storage 53 stores at least one piece of past time-series data of the target physical quantity (for example, the motor current, the flow rate of water or slurry, the polishing pressure, the rotation speed of the polishing table) during the processing of the substrate in association with the lot of the substrate.”)
The same motivation to combine of Chau, Denome, and Torikoshi a set forth for Claim 8 equally applies to Claim 9.
Regarding Claim 15, the limitations in this claim is taught by Chau and Torikoshi as discussed connection with claim 8.
Claim(s) 16 and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chau in view of Denome.
Regarding Claim 16, Chau teaches a non-transitory machine-readable storage medium comprising instructions that, when executed by a processing device, cause the processing device to:
train a first machine learning model to form a first trained machine learning model of a set of chained models (see 0184; Chau: “In an initial layer of the model, a module level neural network (i.e., a neural network for a processing module) is trained to predict program execution times for different processes.”), wherein the first trained machine learning model is trained to output predicted measurement data associated with processing a substrate in a process chamber according to a process recipe input into the first trained machine learning model responsive to receiving the process recipe as input; and (see [0189]; Chau: “The plurality of module level neural networks include one neural network for each processing module (e.g., processing modules 1602 shown in FIG. 16) in the tool 1406 and one neural network for each robot (e.g., for robots 1610 and 1614 shown in FIG. 16) in the tool 1406. These neural networks are shown as 1410-1, ..., and 1410-N, where N is an integer greater than 1, may be collectively called neural networks 1410.”) [The module level neural network reads on ‘the first trained machine learning model’]
train a second machine learning model with training data comprising the predicted measurement data output from the first trained machine learning model to form a second trained machine learning model of the set of chained models (see [0185]; Chau: “The input for the scheduler level neural network is expanded to other tool configurations, mix of recipe types, process times, multiple layers to be processed on the wafers, scheduler modes, etc. Coupled (i.e., nested) with the module level neural networks, the scheduler level neural network with expanded inputs provides recommendations for best product/recipe/wafer mix to achieve highest tool/fleet utilization to reduce cost-of-ownership for the tools.”), wherein the second trained machine learning model is trained to output predicted first ecological efficiency data …based on the predicted measurement data input into the second trained machine learning model. (see [0189]; Chau: “The scheduler level neural network is shown as 1412 and receives outputs of the neural networks 1410 as inputs.” See [0185]: “. That is, the nested neural network based dynamic scheduler, which is trained using different tool hardware configurations and different recipe types, can now recommend optimum tool hardware configuration for a given recipe or recipes.”) [The scheduler level neural network reads on ‘the second trained machine learning model’]
However, Talukder does not explicitly teach: … indicative of an ecological efficiency associated with processing the substrate in the process chamber…
Denome the same or similar field of endeavor teaches: … indicative of an ecological efficiency associated with processing the substrate in the process chamber… (see Abstract; Denome: “A method includes determining, by a processing device, a first eco-efficiency characterization associated with a first design of manufacturing equipment based on one or more of water eco-efficiency characterization, emissions eco-efficiency characterization, or electrical energy eco-efficiency characterization.”)
The same motivation to combine of the combination of Chau and Denome a set forth for Claim 1 equally applies to Claim 16.
Regarding Claim 19, the combination of Chau and Denome teaches all the limitations of claim 16 above, Chau further teaches:
receive measurement data associated with a plurality of process recipes, wherein the measurement data comprises measurement of at least one of current, voltage, power, flow, pressure, concentration, speed, acceleration, or temperature; (see [0005]; Chau: “For example, a recipe defines sequencing, operating temperatures, pressures, gas chemistry, plasma usage, parallel modules, periods for each operation or sub-operation, substrate routing path, and/or other parameters.”)
wherein the processing device is further to: train one or more of the first machine learning model or the second machine learning model with one or more of the measurement data or the ecological efficiency data. (see 0184; Chau: “In an initial layer of the model, a module level neural network (i.e., a neural network for a processing module) is trained to predict program execution times for different processes.” See [0185]: “The input for the scheduler level neural network is expanded to other tool configurations, mix of recipe types, process times, multiple layers to be processed on the wafers, scheduler modes, etc. Coupled (i.e., nested) with the module level neural networks, the scheduler level neural network with expanded inputs provides recommendations for best product/recipe/wafer mix to achieve highest tool/fleet utilization to reduce cost-of-ownership for the tools.”)
However, Chau does not explicitly teach: receive ecological efficiency data corresponding to the plurality of process recipes, wherein the ecological efficiency data is indicative of environmental resource consumption associated with the plurality of process recipes;
Denome further teaches receive ecological efficiency data corresponding to the plurality of process recipes (see [0104]; Denome: “Component data may include the output vs. input power ratio efficiency of the components that are controlled by recipes on the basis of their output powers. The component efficiency data may have been previously generated for one or more of the sub-components.”), wherein the ecological efficiency data is indicative of environmental resource consumption associated with the plurality of process recipes; (see [0051]; Denome: “Examples of inputs may include: how much of each respective utility/resource/waste the manufacturing equipment consumes or generates, equipment configuration information, component efficiency, process recipes and materials, throughput and/or product information, and utilization (up-time vs. down-time of manufacturing equipment).”)
The same motivation to combine of the combination of Chau and Denome a set forth for Claim 1 equally applies to Claim 19.
Claim(s) 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chau in view of Denome in view of Talukder et al. (US20240047248A1 -hereinafter Talukder).
Regarding Claim 17, the combination of Chau and Denome teaches all the limitations of claim 16 above; Denome the same or similar field of endeavor teaches:
…the predicted first ecological efficiency… (see Abstract; Denome: “A method includes determining, by a processing device, a first eco-efficiency characterization associated with a first design of manufacturing equipment based on one or more of water eco-efficiency characterization, emissions eco-efficiency characterization, or electrical energy eco-efficiency characterization.”)
…output predicted second ecological efficiency data indicative of the ecological efficiency associated with processing a substrate in a process chamber… (see Abstract; Denome: “The method further includes comparing the first eco-efficiency characterization to a second eco-efficiency characterization that is associated with a second design of the manufacturing equipment.”)
The same motivation to combine of the combination of Chau and Denome a set forth for Claim 1 equally applies to Claim 17.
However, it does not explicitly teach: wherein the processing device is further to: train a third machine learning model with training data comprising the predicted measurement data output from the first trained machine learning model and the predicted …data output from the second machine learning model to form a third trained machine learning model, wherein the third machine learning model is trained to output …data …according to the process recipe input into the third trained machine learning model.
Talukder further teaches wherein the processing device is further to: train a third machine learning model with training data comprising the predicted measurement data output from the first trained machine learning model and the predicted first …data output from the second machine learning model to form a third trained machine learning model (see [0027]; Talukder: “the training set used to generate the third machine learning model comprises newer ex situ data and in situ measurements than the training set used to generate the second machine learning model.”), wherein the third machine learning model is trained to output …data …according to the process recipe input into the third trained machine learning model. (see [0139]; Talukder: “During Iteration 2 (i.e., training and evaluation of the third library, as shown in and described above in connection with blocks 220-224 of FIG. 2B), training set 456 can be used to train the third library, and test set 458 can be used for evaluation. Note that test set 458 can be used for evaluation of the third library, as well as for evaluation of the second library when compared to the third library (e.g., to determine if the third library is an improvement over the second library).” See [0069]: “training sets of successive iterations can be expanded such that libraries are trained on additional training data. By modifying allocation of training sets and test sets over successive library training iterations, an optimal library can be more quickly trained. In particular, by expanding training sets when a library does not satisfy deployment criteria, libraries can be more quickly and efficiently trained.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to modify the teaching of Chau and Denome to include Talukder’s features of training a third machine learning model with training data comprising the predicted measurement data output from the first trained machine learning model and the predicted data output from the second machine learning model to form a third trained machine learning model, wherein the third machine learning model is trained to output data according to the process recipe input into the third trained machine learning model. Doing so would optimize computational resources and minimize an error. (Talukder, [0191] and [0059])
Claim(s) 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Chau in view of Denome in view of Funk et al. (US20100036514A1 -hereinafter Funk).
Regarding Claim 18, the combination of Chau and Denome teaches all the limitations of claim 16 above; however, it does not explicitly teach wherein the processing device is further to: train an additional machine learning model with training input data comprising historical process target data and training target output data comprising historical process recipes to form an additional trained machine learning model, wherein the additional trained machine learning model is trained to output one or more predicted process recipes associated with a process target input into the additional trained machine learning model.
Funk from the same or similar field of endeavor teaches train an additional machine learning model with training input data comprising historical process target data and training target output data comprising historical process recipes to form an additional trained machine learning model (see [0063]; Funk: “An alternative procedure for generating data for a MLMIMO-related library can include using a machine learning system (MLS). For example, prior to generating the MLMIMO-related library data, the MLS can be trained using known input and output data, and the MLS may be trained with a subset of the MLMIMO-related library data.” See [0095]: “A first set of target parameters 441 can be provided to the first calculation element 440, and a first set of filter outputs 471 can be provided to the first calculation element 440. Output data items 442 from the first calculation element 440 can be provided to one or more MLMIMO model Optimizers 450.” See [0068]: “MLMIMO model-related limits can be obtained by performing the MLMIMO model-related procedure in a “golden” processing chamber, can be historical data that is stored in a library, can be obtained by performing a verified deposition procedure, can be obtained from the MES 180, can be simulation data, and can be predicted data.”), wherein the additional trained machine learning model is trained to output one or more predicted process recipes associated with a process target input into the additional trained machine learning model. (see [0096]-[0097]; Funk: “One or more of the MLMIMO model Optimizers 450 can determine one or more sets of recipe/chamber parameters 456 that can be sent to one or more of the tool controller/models (420, 421, and 422). One or more of the tool controller/models (420, 421, and 422) can be used to calculate predicted data items 427 that can include one or more predicted etch biases, one or more predicted SWA biases, one or more predicted step times for one or more etch recipes, and one or more predicted process gas flows for one or more etch recipes.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to modify the teaching of the combination of Chau and Denome to include Funk’s features of training an additional machine learning model with training input data comprising historical process target data and training target output data comprising historical process recipes to form an additional trained machine learning model, wherein the additional trained machine learning model is trained to output one or more predicted process recipes associated with a process target input into the additional trained machine learning model. Doing so would optimize a process recipe and/or process time and improve the accuracy. (Funk, [0066]-[0067])
Response to Arguments
Applicant’s arguments with respect to the claim rejection(s) of the independent claim(s) have been fully considered and are persuasive because of the amendments. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Nakata (US11694099B2) discloses the third trained model may be acquired through training in which the input data corresponding to the output data of the first trained model is used as additional training data.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to VI N TRAN whose telephone number is (571)272-1108. The examiner can normally be reached Mon-Fri 9:00-5:00.
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, ROBERT FENNEMA can be reached at (571) 272-2748. 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.
/V.N.T./Examiner, Art Unit 2117
/ROBERT E FENNEMA/Supervisory Patent Examiner, Art Unit 2117
Prosecution Insights