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 Objections
Claim 2 is objected to because of the following informalities: In line 2, it states “using the state data and reward data”. The term “the state data” is creating confusion because it has not been defined before. For examination purpose it is assumed that “the state data” is referring to the current state data. Appropriate correction is required.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or no obviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1, 3-7, 15, and 17-20 are rejected under 35 U.S.C. 103 as being unpatentable over Chau et al. (US 20220171373 A1.), and in view of Kulkarni et al. (US 20110286656 A1.).
As per claim 1, Chau et al. teach
A method comprising (para 90):
identifying current state data associated with a substrate processing facility comprising one or more higher-yield tools and (para 168, Fig. 1, The discrete event simulator 1202 receives data indicating the current state of the tool. “For example, the state information of the tool may include status of tool resources (e.g., PMs, airlocks, etc.) in the processing status of the wafers.” Also see para 108, Fig. 3, a fab data collection system 300 for a semiconductor system contain N substrate processing tools 320-1 to 320-T, and each of the substrate tools include one or more processing chamber, controller etc. “Examples of data collected by the processing chambers include substrate or wafer data logs, auto preventive maintenance, high-speed data, optical emission spectroscopy”. The fab data collection system teaches a substrate processing facility with higher yields tools and these can be controlled by the substrate processing tools 320 and the chamber controllers.);
providing the current state data as input to a trained reinforcement learning agent (para 26, in a semiconductor substrate the plurality of states of a current state are sent to a model for determining a best operation with instructions configured for each of the plurality of states. Also see para 6, the instructions are configured to receive first data and second data and based on various parameters and plurality of processing scenario, the data then fit into a simulation process to train a model);
receiving, from the trained reinforcement learning agent, output associated with parameters (Para 124-para 125, In the system 400, the model generator 408 is utilizing a machine learning model using historical data to generate a throughput predictive model. The model will utilize the parameters given to the model and will extract the best scheduling parameter values from the selected patterns, also see para 126, the model generator 408 trains the model to predict optimum scheduling parameters values by using the data collected from the preventive maintenance operations(PMs), recipe times, and wafer loss auto clean(WAC); therefore ,the model is generating the optimum scheduling parameter as output using the associated parameters or collected data); and
causing, based on the parameters, maximizing of lot processing on the one or more higher-yield tools while meeting one or more threshold production values (para 30, the optimum scheduling parameters minimize the idle times for the semiconductor substrate during processing of the plurality of processing chambers and maximize throughput of the tool. Therefore, minimizing the idle time and maximizing the throughput for plurality of processing chambers will produce higher yields in less time.).
However, Chau et al. do not teach
one or more lower-yield tools that have a lower yield than the one or more higher-yield tools.
In the same field of endeavor Kulkarni et al. teach
one or more lower-yield tools that have a lower yield than the one or more higher-yield tools (Kulkarni et al., para 535, Kulkarni et al. teach a model for determining yield changes as the values of one or more attributes of the design data changes the model, and also teaches determining the defect or yield sensitivity of the design. If the yield sensitivity is relatively high, then the DCI for the defect can be modified. Para 694, once the defect related to the parametric yield losses are determined, the data can be input for a simulation to upgrade the model and prevent parametric yield losses).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the optimum scheduling method for a substrate processing facility taught by Chau et al. and to include the determining the yield losses model taught by Kulkarni et al. into the system. This would have been obvious because both Chau et al. and Kulkarni et al. teach maximizing the yields by optimizing the schedule. By adding yield loss determination model, the system can determine the sensitivity yield and defect of the design, and also identify the tools that have lower yields than the higher- yields tools and implement the strategy into the upgraded model (Kulkarni et al., paras 535, 694).
As per claim 3, the combination of Chau et al. and Kulkarni et al. teach
The method of claim 1, wherein the one or more threshold production values comprise an on-time delivery threshold value or a production quantity threshold value (Chau et al., para 88, the smart scheduler can optimize the scheduling parameter values by taking into account preventive maintenance that may have to be skipped or delayed to meet manufacturing deadlines, the smart scheduler can ensure that manufacturing efficiency (accrual/ theoretical time) can be greater than 97% for each recipe).
As per claim 4, the combination of Chau et al. and Kulkarni et al. teach
The method of claim 1, wherein the parameters comprise one or more of:
a maximum waiting lot amount before lot processing via the one or more lower-yield tools;
a maximum lot wait time before lot processing via the one or more lower-yield tools;
per-part wait time before lot processing via the one or more lower-yield tools;
per-process wait time before lot processing via the one or more lower-yield tools; or
maximum lot wait time for each work-in-progress lot (Chau et al., para 88, the smart scheduler can ensure that a wafer idle time can be less than 2% of the total processing time and a manufacturing efficiency (accrual/ theoretical time) can be greater than 97% for each recipe).
As per claim 5, the combination of Chau et al. and Kulkarni et al. teach
The method of claim 1, wherein the state data comprises one or more of lot wait data, lot processing data, lot deadline data, tool data, or preventative maintenance data (Chau et al., para 27, A state of the tools includes indications of resources of the tool and a processing status of the semiconductor substrate; therefore, the data carries the tool and instruction information).
As per claim 6, the combination of Chau et al. and Kulkarni et al. teach
The method of claim 1, wherein the parameters are associated with one or more of dispatching decisions or scheduling decisions (Chau et al., para 81, using reinforcement learning a neural model has been trained to predict optimum scheduling parameter values. Also, the model is trained to use the collective data to design a preventive maintenance operation, recipe times etc.).
As per claim 7, the combination of Chau et al. and Kulkarni et al. teach
The method of claim 1, wherein the causing of the maximizing of the lot processing on the one or more higher-yield tools while meeting the one or more threshold production values comprises providing the parameters to one or more of a dispatching system or a scheduling system (Chau et al., para 30, the optimum scheduling parameters minimize the idle times for the semiconductor substrate during processing of the plurality of processing chambers and maximize throughput of the tool. Therefore, minimizing the idle time and maximizing the throughput for plurality of processing chambers will produce higher yields in less time. Also see para 81, using reinforcement learning a neural model has been trained to predict optimum scheduling parameter values. Also, the model is trained to use the collective data to design a preventive maintenance operation, recipe times etc.).
As per claim 15, Chau et al. teach
A non-transitory computer readable medium having instructions stored thereon, which, when executed by a processing device, cause the processing device perform operations comprising (para 113,119):
identifying current state data associated with a substrate processing facility comprising one or more higher-yield tools and (para 168, Fig. 1, The discrete event simulator 1202 receives data indicating the current state of the tool. “For example, the state information of the tool may include status of tool resources (e.g., PMs, airlocks, etc.) in the processing status of the wafers. Also see para 108, Fig. 3, a fab data collection system 300 for a semiconductor system contain N substrate processing tools 320-1 to 320-T, and each of the substrate tools include one or more processing chamber, controller etc.” Examples of data collected by the processing chambers include substrate or wafer data logs, auto preventive maintenance, high-speed data, optical emission spectroscopy”. The fab data collection system teaches a substrate processing facility with higher yields tools and these can be controlled by the substrate processing tools 320 and the chamber controllers.);
providing the current state data as input to a trained reinforcement learning agent (para 26, in a semiconductor substrate the plurality of states of a current state are sent to a model for determining a best operation with instructions configured for each of the plurality of states. Also see para 6, the instructions are configured to receive first data and second data and based on various parameters and plurality of processing scenario, the data then fit into a simulation process to train a model);
receiving, from the trained reinforcement learning agent, output associated with parameters (Para 124-para 125, In the system 400, the model generator 408 is utilizing a machine learning model using historical data to generate a throughput predictive model. The model will utilize the parameters given to the model and will extract the best scheduling parameter values from the selected patterns, also see para 126, the model generator 408 trains the model to predict optimum scheduling parameters values by using the data collected from the preventive maintenance operations(PMs), recipe times, and wafer loss auto clean(WAC); therefore ,the model is generating the optimum scheduling parameter as output using the associated parameters or collected data); and
causing, based on the parameters, maximizing of lot processing on the one or more higher-yield tools while meeting one or more threshold production values (para 30, the optimum scheduling parameters minimize the idle times for the semiconductor substrate during processing of the plurality of processing chambers and maximize throughput of the tool. Therefore, minimizing the idle time and maximizing the throughput for plurality of processing chambers will produce higher yields in less time.).
However, Chau et al. do not teach
one or more lower-yield tools that have a lower yield than the one or more higher-yield tools.
In the same field of endeavor Kulkarni et al. teach
one or more lower-yield tools that have a lower yield than the one or more higher-yield tools (Kulkarni et al., para 535, Kulkarni et al. teach a model for determining yield changes as the values of one or more attributes of the design data changes the model, and also teaches determining the defect or yield sensitivity of the design. If the yield sensitivity is relatively high, then the DCI for the defect can be modified. Para 694, once the defect related to the parametric yield losses are determined, the data can be input for a simulation to upgrade the model and prevent parametric yield losses).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention to modify the optimum scheduling method for a substrate processing facility taught by Chau et al. and to include the determining the yield losses model taught by Kulkarni et al. into the system. This would have been obvious because both Chau et al. and Kulkarni et al. teach maximizing the yields by optimizing the schedule. By adding yield loss determination model, the system can determine the sensitivity yield and defect of the design, and also identify the tools that have lower yields than the higher- yields tools and implement the strategy into the upgraded model (Kulkarni et al., paras 535, 694).
As per claim 17, please refer to the analysis of claim 3 above, as they recite the same limitations.
As per claim 18, please refer to the analysis of claim 4 above, as they recite the same limitations.
As per claim 19, please refer to the analysis of claim 5 above, as they recite the same limitations.
As per claims 20 please refer to the analysis of claim 6 and 7 above, as they recite the same limitations.
Claims 2,8-14, and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Chau et al. (US 20220171373 A1.), in view of Kulkarni et al. (US 20110286656 A1.), and further in view of (Pinskiy et al. (US 20230391016 A1.).
As per claim 2, Chau et al. teach
The method of claim 1, the trained reinforcement learning agent being trained using the state data (Para 124-para 125, In the system 400, the model generator 408 is utilizing a machine learning model using historical data to generate a throughput predictive model. The model will utilize the parameters given to the model and will extract the best scheduling parameter values from the selected patterns) and reward data (para 165, a reward data obtaining from state-to-state transitions using reinforcement learning),
However, the combination of Chau et al. and Kulkarni et al. do not teach
the reward data being associated with the maximizing of the lot processing on the one or more higher-yield tools while meeting the one or more threshold production values.
In the same field of endeavor Pinskiy et al. teach
the reward data being associated with the maximizing of the lot processing on the one or more higher-yield tools while meeting the one or more threshold production values (Pinskiy et al., claim 1 states reward calculation associated with the state information and also states determining an expected reward calculation corresponding to the current state system. If the reward amount exceeds a threshold amount, the system will update the system and implement the reinforcement learning model accordingly. Also see para 148 describes a decision process, Markov Decision Process (MDP) in which based on plurality of states future state condition are determined. A reward calculation process shown in para 148, para 170-para 197).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention to modify the optimum scheduling method for a substrate processing facility taught by Chau et al. and to include the reward data calculation method taught by Pinskiy et al. into the system. This would have been obvious because the combination of Chau et al., Kulkarni et al., and Pinskiy et al. teach maximizing the yields by optimizing the schedule. By adding the reward data into the trained model, the schedule system will be optimized with consideration of any changes can be input to the trained model and the schedule will be optimized by considering the adjustment. (Pinskiy et al., claim1, para 148).
As per claim 8, Chau et al. teach
A method comprising:
identifying state data associated with a substrate processing facility comprising one or more higher-yield tools and (please refer to the analysis of claim 1);
training a reinforcement learning agent using the state data (please refer to the analysis of claim 1) and
However, Chau et al. do not teach
one or more lower-yield tools that have a lower yield than the one or more higher-yield tools.
In the same field of endeavor, Kulkarni et al. teach
one or more lower-yield tools that have a lower yield than the one or more higher-yield tools (please refer to the analysis of claim 1).
The combination of Chau et al. and Kulkarni et al. do not teach
identifying reward data associated with maximizing lot processing on the one or more higher-yield tools while meeting one or more threshold production values: and
the reward data to generate a trained reinforcement learning agent, wherein the trained reinforcement learning agent is to output parameters to maximize the lot processing on the one or more higher-yield tools while meeting the one or more threshold production values
In the same field of endeavor, Pinskiy et al. teach
identifying reward data associated with maximizing lot processing on the one or more higher-yield tools while meeting one or more threshold production values (Pinskiy et al, claim 1 states reward calculation associated with the state information, and also states determining an expected reward calculation corresponding to the current state system. If the reward amount exceeds a threshold amount, the system will update the system and implement the reinforcement learning model accordingly); and
the reward data to generate a trained reinforcement learning agent, wherein the trained reinforcement learning agent is to output parameters to maximize the lot processing on the one or more higher-yield tools while meeting the one or more threshold production values (Pinskiy et al., para 148 describes a decision process, Markov Decision Process (MDP) in which based on plurality of states, future state condition are determined. A reward calculation process shown in para 148, para 170-para 197, also see para 140, a machine learning method has been used to maximize accumulation of positive results or reward and adjusting the condition based on the reward).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention to modify the optimum scheduling method for a substrate processing facility taught by Chau et al. and to include the reward data calculation method taught by Pinskiy et al. into the system. This would have been obvious to take the same initiative as described in claim 2 above.
As per claim 9, please refer to the analysis of claim 3 above, as they recite the same limitations.
As per claim 10, please refer to the analysis of claim 4 above, as they recite the same limitations.
As per claim 11, please refer to the analysis of claim 5 above, as they recite the same limitations.
As per claim 12, please refer to the analysis of claim 7 above, as they recite the same limitations.
As per claim 13, the combination of Chau et al., Kulkarni et al., and Pinskiy et al. teach
The method of claim 8, wherein the state data comprises one or more of:
current state data associated with current processing of current lots in the substrate processing facility (Chau et al., para 26 -27, using current state of a plurality of states progress to the next steps to simulate an execution of the model based on the current state); or
historical state data associated with historical processing of historical lots in the substrate processing facility (Chau et al., para 112, the tool system controllers 348 record history, detailed event logs, lot-based alarms, time-based alarms, tool controller health, parts tracking, component history, material scheduling, and other data).
As per claim 14, the combination of Chau et al., Kulkarni et al., and Pinskiy et al. teach
The method of claim 8, wherein the state data comprises perturbed state data formed by one or more of lot duplication, lot removal, or lot location adjustment along a route (Chau et al., para 46, an optimum route to transfer the semiconductor substate are determined by following an instruction of a simulation process. Therefore, an adjustment has been done based on the decision made by the model. The modified data to obtain the optimum route teaches the perturbed state data.).
As per claim 16, Chau et al. teach
The non-transitory computer readable medium (para 113,119)
Please refer to the analysis of claim 2 above as they recite the same limitations.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Please refer to the form PTO-892.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Rokeya Alam whose telephone number is (571)272-0083. The examiner can normally be reached on 7:30am - 4:30pm.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Mr. Scott Baderman can be reached at telephone number (571-272-3644). The fax phone number for the organization where this application or proceeding is assigned is (571) 273-8300.
Information regarding the status of an application may be obtained from Patent Center. Status information for published applications may be obtained from Patent Center. Status information for unpublished applications is available through Patent Center for authorized users only. Should you have questions about access to Patent Center, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
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) Form at https://www.uspto.gov/patents/uspto-automated- interview-request-air-form.
/ROKEYA SHAWALI ALAM/Examiner, Art Unit 2118
/SCOTT T BADERMAN/Supervisory Patent Examiner, Art Unit 2118
Prosecution Insights