Apparatus and method for inferring parameter for controlling plasma equipment

Abstract

The present invention relates to an apparatus and method for inferring a parameter for controlling plasma equipment, the apparatus comprising: a reverse inference module that reversely infers a final plasma state value for deriving an achievement process result to be achieved on the basis of a pre-generated learning performance result, reversely infers final sensing data to be obtained from a plurality of sensors in response to the final plasma state value, and reversely infers a final parameter value of the plasma equipment, corresponding to the final sensing data; and a control module for controlling the plasma equipment on the basis of the final parameter value. The apparatus and method are applicable to other embodiments.

Description

Parameter inference device and method for plasma equipment control

[0001] The present invention relates to a parameter inference device and method for controlling plasma equipment.

[0002] Plasma generated by high-frequency power is widely used in etching and deposition processes for wafers and display panel substrates. This plasma is sensitive to changes in the chamber's internal environment. This plasma response alters the plasma state (e.g., electron density, electron temperature, and ion content), which in turn alters the product process results, potentially leading to defects in the process and product.

[0003] To solve these problems, a method of maintaining a constant internal state of the chamber through a cleaning process (e.g., cleaning) as an intermediate step in the plasma-based process is currently common.

[0004] However, since only some data is used to determine whether the cleaning process has been performed accurately, the cleaning process may end before all impurities in the chamber are removed, and the plasma state may change before the cleaning process is performed, so there is still a high probability that defects will occur in the process and product.

[0005] In addition, since stopping the process every time a change in the plasma state occurs will result in a decrease in product production, it is necessary to decide whether to continue the process or to stop the process, inspect the equipment, and then restart the process based on the amount of change in the plasma state each time the plasma state changes.

[0006] Embodiments of the present invention for solving these conventional problems provide a parameter inference device and method for controlling plasma equipment capable of setting parameters to change the plasma state to achieve a desired process result by checking when a change in the plasma state occurs and the influence of the change in the plasma state on the process.

[0007] In addition, embodiments of the present invention provide a parameter inference device and method for plasma equipment control that can apply sensing data related to plasma equipment to an artificial intelligence algorithm to check whether the predicted plasma state, process state, and process result fall within a set range, and control a power device, gas inlet device, or vacuum pump installed in the plasma equipment in real time based on the check result.

[0008] A parameter inference device for controlling plasma equipment according to an embodiment of the present invention is characterized by including a reverse inference module that reversely infers a final plasma state value for deriving an achievement process result to be achieved based on a pre-generated learning performance result, reversely infers final sensing data to be acquired from a plurality of sensors corresponding to the final plasma state value, and reversely infers a final parameter value of the plasma equipment corresponding to the final sensing data, and a control module that controls the plasma equipment based on the final parameter value.

[0009] In addition, it is characterized by including a collection module that collects sensing data acquired from the plurality of sensors when operating the plasma equipment; a prediction module that applies the sensing data to an artificial intelligence algorithm to predict a plasma state value, applies at least one of the sensing data and the plasma state value to the artificial intelligence algorithm to predict a process state, and predicts a process result using the predicted process state; and a planning module that applies the plasma state value, the process state, and the process result to an artificial intelligence algorithm to confirm an equipment parameter value requiring control and generates the learning performance result.

[0010] In addition, the control module is characterized by controlling the parameters of the plasma equipment based on the equipment parameter values ​​confirmed in the planning module.

[0011] In addition, it is characterized by including a message queue that loads information derived from the collection module, the prediction module, the planning module, the control module, and the reverse inference module.

[0012] In addition, the prediction module includes a plasma state prediction module that predicts the plasma state value, a process state prediction module that predicts the process state, and a process result prediction module that predicts the process result, and each of the prediction modules is characterized in that it uses a different artificial intelligence algorithm.

[0013] In addition, the plasma state prediction module, the process state prediction module, and the process result prediction module are characterized in that they operate independently by collecting information loaded in the message queue as needed.

[0014] In addition, the reverse inference module is characterized in that it applies the above-described achievement process result to an artificial intelligence algorithm for reverse inference, compares the predicted plasma state value from the message queue with the plasma state value called from the learning performance result generated from the planning module, and infers the final plasma state value.

[0015] In addition, the reverse inference module is characterized in that it applies the final plasma state value to the artificial intelligence algorithm for the reverse inference, compares the sensing data predicted from the message queue with the sensing data called from the learning performance result generated from the planning module, and infers the final sensing data.

[0016] In addition, the reverse inference module is characterized in that it applies the final sensing data to the artificial intelligence algorithm for the reverse inference, compares the predicted equipment parameter values ​​from the message queue with the equipment parameter values ​​called from the learning performance results generated from the planning module, and infers the final parameter values.

[0017] In addition, the plasma state value and the process state are characterized in that they are predicted through a plasma simulator.

[0018] In addition, a parameter inference method for controlling plasma equipment according to an embodiment of the present invention is characterized by including a step of receiving an achievement process result to be achieved by an electronic device, a step of inferring a final plasma state value for deriving the achievement process result inversely based on a pre-generated learning performance result by the electronic device, a step of inferring final sensing data to be acquired from a plurality of sensors in response to the final plasma state value inversely based on the pre-generated learning performance result by the electronic device, a step of inferring a final parameter value of the plasma equipment corresponding to the final sensing data inversely based on the pre-generated learning performance result by the electronic device, and a step of controlling the plasma equipment in response to the final parameter value by the electronic device.

[0019] As described above, the parameter inference device and method for controlling plasma equipment according to the present invention can predict the plasma state, process state, and process result according to the set parameters by setting parameters to change the plasma state to a state that can achieve a desired process result by checking the case where a change in the plasma state occurs and the influence of the change in the plasma state on the process, thereby shortening the time for identifying process conditions and parameter input values ​​for newly required process states and process results.

[0020] In addition, the parameter inference device and method for controlling plasma equipment according to the present invention applies sensing data related to the plasma equipment to an artificial intelligence algorithm to check whether the predicted plasma state, process state, and process result fall within a set range, and based on the check result, controls a power device, a gas inlet device, or a vacuum pump installed in the plasma equipment in real time, thereby adjusting the parameters so that the plasma state, process state, and process result fall within the set range, thereby minimizing process byproducts and process defects due to component damage that may occur as process time accumulates, and has the effect of improving product productivity.

[0021] FIG. 1 is a diagram showing the main configuration of a system that performs parameter inference for plasma equipment control according to an embodiment of the present invention.

[0022] FIG. 2 is a flowchart illustrating a method for performing learning for parameter inference according to an embodiment of the present invention.

[0023] FIG. 3 is a diagram showing a table in which information on plasma state changes according to an embodiment of the present invention is entered.

[0024] FIG. 4 is a diagram showing a table for confirming a process control combination according to an embodiment of the present invention.

[0025] FIG. 5 is a flowchart illustrating a method for inferring parameters in the reverse direction according to an embodiment of the present invention.

[0026] Figure 6 is an example screen diagram showing a screen configuration for parameter inference according to an embodiment of the present invention.

[0027] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The detailed description set forth below, together with the accompanying drawings, is intended to explain exemplary embodiments of the present invention and is not intended to represent the only embodiments in which the present invention may be practiced. In the drawings, portions irrelevant to the description may be omitted for clarity in describing the present invention, and the same reference numerals may be used throughout the specification for identical or similar components.

[0028] FIG. 1 is a diagram showing the main configuration of a system that performs parameter inference for plasma equipment control according to an embodiment of the present invention.

[0029] Referring to FIG. 1, a system (10) for performing parameter inference according to the present invention may include a sensor (100) and an electronic device (200).

[0030] The sensor (100) is a sensor equipped in plasma equipment (not shown), and obtains sensing data such as spectrum data and electrical characteristic data related to light generated during a plasma process and provides the data to the collection module (210). To this end, the sensor (100) can communicate with the collection module (210) 1:1 or can be connected to N:1 via LAN to perform communication. In addition, the sensor (100) may include an OES sensor, a VI sensor, a Matcher sensor, a SPOES sensor, an MFC (mass flow controller) sensor, a Gage sensor, a Temperature sensor, a Vacuum pumps sensor, etc.

[0031] The electronic device (200) collects sensing data acquired from the sensor (100), and derives a learning performance result by performing learning of a plurality of artificial intelligence algorithms for parameter inference based on the sensing data. In addition, the electronic device (200) can infer a final plasma state value for deriving a desired process result based on the derived learning performance result, final sensing data to be acquired from a plurality of sensors corresponding to the final plasma state value, and final parameter values ​​of the plasma equipment corresponding to the final sensing data, and control the plasma equipment based on the inference. To this end, the electronic device (200) may include a collection module (210), a message queue (220), a prediction module (230), a planning module (240), a control module (250), and a reverse inference module (260).

[0032] First, the electronic device (200) receives a start signal for learning from a user of the electronic device (200), learns the plasma state, process state, and process result based on the sensing data received from the sensor (100), and can perform learning of an artificial intelligence algorithm.

[0033] The collection module (210) collects sensing data obtained from the sensor (100). At this time, the sensing data may be sensing data obtained from an OES sensor, a VI sensor, a Matcher sensor, a SPOES sensor, an MFC (mass flow controller) sensor, a Gage sensor, a Temperature sensor, a Vacuum pumps sensor, etc.

[0034] The collection module (210) can transmit sensing data to a message queue (220) or to a prediction module (230). At this time, the sensing data transmitted to the message queue (220) can be loaded into the message queue (220).

[0035] The sensing data loaded in the message queue (220) is provided to the prediction module (230). Although not shown, the prediction module (230) may include a plasma state prediction module that predicts a plasma state value, a process state prediction module that predicts a process state, and a process result prediction module that predicts a process result.

[0036] The prediction module (230) can predict plasma state values, process states, and process results using sensing data. More specifically, the plasma state prediction module can predict electron density, electron temperature, and ion amount using sensing data. The process state prediction module can predict deposition thickness, etching profile, and the like using sensing data. The process result prediction module can predict process result values, such as deposition thickness and etching profile after the process is completed, using sensing data, plasma state values, and process states.

[0037] For example, the plasma state prediction module applies a set of sensing data loaded in the message queue (220) to an artificial intelligence algorithm to predict electron density, electron temperature, or ion amount, and performs learning of the artificial intelligence algorithm by comparing the electron density, electron temperature, or ion amount acquired from the plasma diagnostic device at the time the sensing data was acquired. The plasma state prediction module can provide the plasma state value to the message queue (220) in real time so that the message queue (220) can load the plasma state value predicted by the artificial intelligence algorithm.

[0038] The process state prediction module can predict the process state, for example, deposition thickness or etching profile, by applying a set of sensing data loaded in the message queue (220) to an artificial intelligence algorithm, and can predict the process state using the sensing data and the plasma state value predicted from a plasma simulator (not shown).

[0039] In this way, when the process state prediction module predicts the process state using the sensing data or predicts the process state using the sensing data and the plasma state value predicted from the plasma simulator, the process state prediction module can operate independently from the plasma state prediction module. In addition, when the process state prediction module predicts the process state using the sensing data and the plasma state value loaded into the message queue (220), the process state prediction module does not operate after the operation of the plasma state prediction module is completed, but rather the plasma state value predicted by the plasma state prediction module is loaded into the message queue (220) in real time, so it can operate independently from the plasma state prediction module almost simultaneously.

[0040] The process status prediction module can provide the process status to the message queue (220) in real time so that the message queue (220) can load the process status predicted by the artificial intelligence algorithm.

[0041] The process result prediction module can predict the process result using the sensing data, plasma state value, and process state loaded into the message queue (220), and can predict the process result using the sensing data and the plasma state value and process state predicted by the plasma simulator. In this way, when the process result prediction module predicts the process result using the sensing data and the plasma state value and process state predicted by the plasma simulator, the process result prediction module can operate independently from the plasma state prediction module and the process state prediction module. In addition, when the process result prediction module predicts the process result using the sensing data, plasma state value, and process state loaded into the message queue (220), rather than operating after the operations of the plasma state prediction module and the process state prediction module are completed, the plasma state value and process state predicted by the plasma state prediction module and the process state prediction module are loaded into the message queue (220) in real time, so that it can operate independently from the plasma state prediction module and the process state prediction module almost simultaneously.

[0042] The process result prediction module can provide the process result to the message queue (220) in real time so that the message queue (220) can load the process result predicted by the artificial intelligence algorithm.

[0043] A plurality of artificial intelligence algorithms stored in the prediction module (230) are trained based on the sensing data and the predicted results predicted from each module, and the electronic device (200) performs planning by calling the planning module (240) to set parameter values ​​to be controlled when the process is performed in the plasma equipment in the future based on the predicted process results.

[0044] The planning module (240) is called when the plasma state value predicted by the plasma state prediction module has a difference greater than or equal to a threshold from a preset range, when the process state predicted by the process state prediction module has a difference greater than or equal to a threshold from a preset process state range, and when the process result predicted by the process result prediction module has a difference greater than or equal to a threshold from a range of process results measured after the actual process is completed.

[0045] If the difference between the predicted plasma state value and the preset range exceeding the threshold value occurs in the plasma state prediction module, the planning module (240) can receive input from the user about the change items, change amount, process parameters, number of processes, process execution time, etc. that have caused the plasma state to change.

[0046] If the process state predicted by the process state prediction module differs by a threshold or more from the range of the preset process state, the planning module (240) can receive from the user an input of a process influence related to a change item in which the plasma state has changed.

[0047] The planning module (240) can use an artificial intelligence algorithm to plan which parameters of the plasma equipment should be adjusted to which values ​​when a difference occurs between the process results predicted by the process result prediction module and the range of the actual measured process results that exceeds a threshold value.

[0048] The planning module (240) can generate a control combination based on an adjustable control factor in a parameter control combination condition having information on an adjustable item in at least one process among a process in which a predicted plasma state value deviates from a range of preset plasma state values ​​by a threshold or more, a process in which a predicted process state deviates from a range of preset process states by a threshold or more, and a process in which a predicted process result deviates from a range of an actually measured process result by a threshold or more.

[0049] The process control combination can be performed by applying the sensing data acquired from the performed process and the predicted plasma state value, process state, and process result corresponding to the sensing data to the reinforcement learning algorithm, which is an artificial intelligence algorithm. The control module (250) for controlling parameter values ​​based on the process control combination derived as a learning performance result in the planning module (240) can be called, and the learning performance result can be loaded into the message queue (220).

[0050] The control module (250) controls the parameter values ​​of the plasma equipment based on the parameter values ​​provided by the planning module (240). In addition, the control module (250) can load the corresponding parameter values ​​into the message queue (220) as learning performance results.

[0051] In this way, based on the sensing data received from the sensor (100), the plasma state, process state, and process result are learned, and after the learning of the artificial intelligence algorithm is completed, when the electronic device (200) receives a signal for inferring parameters in the reverse direction from the user, the reverse inference module (260) receives the process result to be achieved (hereinafter, “achieved process result”) from the user. At this time, the achieved process result input from the user may be a deposition thickness and an etching profile, etc. to be confirmed as the result of the process after the process is completed.

[0052] The reverse inference module (260) predicts the final plasma state value for deriving the achievement process result through data transmission and reception with the message queue (220) or the prediction module (230). More specifically, the reverse inference module (260) applies the achievement process result to a regression-based artificial intelligence algorithm stored in the reverse inference module (260). The reverse inference module (260) applies the achievement process result to predict the plasma state value for deriving the achievement process result among the plasma state values ​​loaded in the message queue (220) or the plasma state values ​​learned and predicted in the prediction module (230).

[0053] The reverse inference module (260) calls the plasma state value corresponding to the achieved process result based on the amount of change according to the learning performance result. At this time, if the plasma state value corresponding to the achieved process result does not exist in the learning performance result, the reverse inference module (260) can predict the plasma state value corresponding to the achieved process result through learning.

[0054] The reverse inference module (260) can select an appropriate plasma state value as the final plasma state value to derive the achieved process result by applying the predicted plasma state value and the called plasma state value to an artificial intelligence algorithm.

[0055] The reverse inference module (260) predicts the final sensing data for deriving the final plasma state value. More specifically, the reverse inference module (260) applies the final plasma state value to a regression-based artificial intelligence algorithm. The reverse inference module (260) applies the final plasma value to predict the sensing data for deriving the final plasma state value among the sensing data loaded in the message queue (220) or the sensing data learned and predicted in the prediction module (230).

[0056] The reverse inference module (260) calls the sensing data measured at the same time as the final plasma state value among the plasma state values ​​corresponding to the learning performance results. At this time, if the sensing data measured at the same time as the final plasma state value does not exist, the reverse inference module (260) can predict the sensing data corresponding to the final plasma state value through learning.

[0057] The reverse inference module (260) can select appropriate sensing data as the final sensing data to derive the final plasma state value by applying the predicted sensing data and the called sensing data to an artificial intelligence algorithm.

[0058] The reverse inference module (260) predicts the final equipment parameter values ​​for deriving the final sensing data. More specifically, the reverse inference module (260) applies the final equipment parameter values ​​to a regression-based artificial intelligence algorithm. The reverse inference module (260) applies the final equipment parameter values ​​to predict the parameter values ​​for deriving the final equipment parameter values ​​from among the parameter values ​​loaded in the message queue (220) or the parameter values ​​learned and predicted in the prediction module (230).

[0059] The reverse inference module (260) calls the parameter values ​​measured and stored at the same or similar time as the final sensing data among the parameter values ​​corresponding to the learning performance results. At this time, if the parameter values ​​measured and stored at the same or similar time as the final sensing data do not exist in the table, the reverse inference module (260) can predict the parameter values ​​corresponding to the final sensing data through learning.

[0060] The reverse inference module (260) can select an appropriate parameter value as the final equipment parameter value to derive the final sensing data by applying the predicted parameter value and the called parameter value.

[0061] The reverse inference module (260) calls the control module (250) to provide the final equipment parameter values ​​selected by the control module (250), and the control module (250) controls the parameter values ​​of the plasma equipment using the final equipment parameter values. Through this, the present invention has the effect of controlling a power device, a gas inlet device, or a vacuum pump installed in the plasma equipment in real time to achieve the process result to be achieved by reversely inferring the equipment parameter values ​​that must be set in the plasma equipment to derive the process result to be achieved.

[0062] FIG. 2 is a flowchart illustrating a method for performing learning for parameter inference according to an embodiment of the present invention. FIG. 3 is an exemplary diagram illustrating a table in which information on plasma state changes is input according to an embodiment of the present invention. FIG. 4 is an exemplary diagram illustrating a table for confirming process control combinations according to an embodiment of the present invention.

[0063] Referring to FIGS. 2 to 4, in step 201, the electronic device (200) checks whether a start signal for learning is received from the user of the electronic device (200). At this time, learning means learning the plasma state, process state, and process result matching the sensing data collected from the sensor (100) and the parameter values ​​input at that point in time using an artificial intelligence algorithm, and through such learning, the electronic device (200) can predict the plasma state, process state, and process result from the sensing data.

[0064] In step 203, the collection module (210) of the electronic device (200) collects sensing data obtained from the sensor (100). At this time, the sensing data may be sensing data obtained from an OES sensor, a VI sensor, a Matcher sensor, a SPOES sensor, an MFC (mass flow controller) sensor, a Gage sensor, a Temperature sensor, a Vacuum pumps sensor, etc.

[0065] In step 205, the collection module (210) can transmit the sensing data to the message queue (220) or to the prediction module (230). At this time, the sensing data transmitted to the message queue (220) can be loaded into the message queue (220).

[0066] In step 207, the electronic device (200) provides the sensing data loaded in the message queue (220) to the prediction module (230). Although not shown, the prediction module (230) may include a plasma state prediction module that predicts a plasma state value, a process state prediction module that predicts a process state, and a process result prediction module that predicts a process result.

[0067] In step 209, the prediction module (230) can predict plasma state values, process states, and process results using sensing data. More specifically, the plasma state prediction module can predict electron density, electron temperature, and ion amount using sensing data. The process state prediction module can predict deposition thickness, etching profile, and the like using sensing data. The process result prediction module can predict process result values, such as deposition thickness and etching profile after the process is completed, using sensing data, plasma state values, and process states.

[0068] At this time, the type of sensor (100) and the sensing data used in the plasma state prediction module, process state prediction module, and process result prediction module will be described in more detail using Table 1 below.

[0069] Sensor1 Set2 Set3 Set4 Set5 SetOESOOOOVIOOOOOMatcherOOSPOESOOOMFCOGageTempOVacuum pumpsO

[0070] In Table 1, the first to fifth sets may refer to sets of sensing data, and the sets of sensing data may be applied to an artificial intelligence algorithm to learn the artificial intelligence algorithm as needed. The plasma state prediction module applies the first set to the artificial intelligence algorithm to predict the electron density, and performs learning of the artificial intelligence algorithm by comparing the predicted electron density with the electron density acquired from the cutoff probe, which is a plasma diagnostic device, at the time when the sensing data included in the first set was acquired. In addition, the plasma state prediction module applies the second set to the artificial intelligence algorithm to predict the electron density, and performs learning of the artificial intelligence algorithm by comparing the predicted electron density with the electron density acquired from the cutoff probe, which is a plasma diagnostic device, at the time when the sensing data included in the second set was acquired. The plasma state prediction module applies the second set to the artificial intelligence algorithm to predict the electron temperature, and performs learning of the artificial intelligence algorithm by comparing the predicted electron temperature with the electron temperature acquired from the Langmuir probe, which is a plasma diagnostic device, at the time when the sensing data included in the second set was acquired. In addition, the plasma state prediction module applies the second set and the predicted electron density to an artificial intelligence algorithm to predict the electron temperature, and performs learning of the artificial intelligence algorithm by comparing the predicted electron temperature with the electron temperature acquired from a Langmuir probe, which is a plasma diagnostic device, at the time when the sensing data included in the second set was acquired.

[0071] The plasma state prediction module predicts the ion quantity by applying the third set to the artificial intelligence algorithm, and performs learning of the artificial intelligence algorithm by comparing the predicted ion quantity with the ion quantity acquired from the EQP, which is a plasma diagnostic device, at the time when the sensing data included in the third set was acquired. In addition, the plasma state prediction module predicts the ion quantity state of the plasma by inputting the electron density predicted as the first set and the electron temperature predicted as the second set into a plasma simulator (not shown), and performs learning of the artificial intelligence algorithm by comparing the predicted ion quantity state with the ion quantity acquired from the EQP, which is a plasma diagnostic device, at the time when the sensing data included in the first and second sets were acquired.

[0072] The plasma state prediction module can provide the plasma state value to the message queue (220) in real time so that the message queue (220) can load the plasma state value predicted by the artificial intelligence algorithm.

[0073] In addition, it is clarified that the combination of data applied to the artificial intelligence algorithm included in the plasma state prediction module is not limited to the above-described embodiment, and can be expanded and applied depending on the type of sensor and the type of plasma diagnostic device.

[0074] The process state prediction module can predict the process state using the sensing data loaded into the message queue (220), can predict the process state using the sensing data and plasma state value loaded into the message queue (220), and can predict the process state using the sensing data and the plasma state value predicted by the plasma simulator. In this way, when the process state prediction module predicts the process state using the sensing data or predicts the process state using the sensing data and the plasma state value predicted by the plasma simulator, the process state prediction module can operate independently from the plasma state prediction module. In addition, when the process state prediction module predicts the process state using the sensing data and plasma state value loaded into the message queue (220), it does not operate after the operation of the plasma state prediction module is completed, but since the plasma state value predicted by the plasma state prediction module is loaded into the message queue (220) in real time, it can operate independently from the plasma state prediction module almost simultaneously.

[0075] The process state prediction module applies the fifth set to the artificial intelligence algorithm to predict the deposition thickness, and performs learning of the artificial intelligence algorithm by comparing the predicted deposition thickness with the deposition thickness acquired from a scanning electron microscope (SEM), which is a process inspection device, at the time when the sensing data included in the fifth set was acquired. In addition, the process state prediction module applies the electron density predicted by the fourth set and the plasma state prediction module and the ion quantity state predicted by the plasma simulator to the artificial intelligence algorithm to predict the deposition thickness, and performs learning of the artificial intelligence algorithm by comparing the predicted deposition thickness with the deposition thickness acquired from a SEM, which is a process inspection device, at the time when the sensing data included in the fourth set was acquired.

[0076] The process state prediction module applies the fifth set to the artificial intelligence algorithm to predict the etching profile, and performs learning of the artificial intelligence algorithm by comparing the predicted etching profile with the etching profile acquired from the SEM, which is a process inspection device, at the time when the sensing data included in the fifth set was acquired. In addition, the process state prediction module applies the electron temperature predicted by the fourth set and the plasma state prediction module and the ion quantity state predicted by the plasma simulator to the artificial intelligence algorithm to predict the etching profile, and performs learning of the artificial intelligence algorithm by comparing the predicted etching profile with the etching profile acquired from the SEM, which is a process inspection device, at the time when the sensing data included in the fourth set was acquired.

[0077] The process status prediction module can provide the process status to the message queue (220) in real time so that the message queue (220) can load the process status predicted by the artificial intelligence algorithm.

[0078] In addition, it is clarified that the combination of data applied to the artificial intelligence algorithm included in the process status prediction module is not limited to the above-described embodiment, and can be expanded and applied depending on the type of sensor and the type of process inspection device.

[0079] The process result prediction module can predict the process result using the sensing data, plasma state value, and process state loaded into the message queue (220), and can predict the process result using the sensing data and the plasma state value and process state predicted by the plasma simulator. In this way, when the process result prediction module predicts the process result using the sensing data and the plasma state value and process state predicted by the plasma simulator, the process result prediction module can operate independently from the plasma state prediction module and the process state prediction module. In addition, when the process result prediction module predicts the process result using the sensing data, plasma state value, and process state loaded into the message queue (220), rather than operating after the operations of the plasma state prediction module and the process state prediction module are completed, the plasma state value and process state predicted by the plasma state prediction module and the process state prediction module are loaded into the message queue (220) in real time, so that it can operate independently from the plasma state prediction module and the process state prediction module almost simultaneously.

[0080] The process result prediction module can provide the process result to the message queue (220) in real time so that the message queue (220) can load the process result predicted by the artificial intelligence algorithm.

[0081] In step 211, multiple artificial intelligence algorithms stored in the prediction module (230) are trained based on the sensing data and the predicted results predicted by each module, and step 213 is performed. In step 213, the electronic device (200) performs planning by calling the planning module (240) to set parameter values ​​to be controlled when the process is performed in the plasma equipment in the future based on the predicted process results.

[0082] More specifically, the planning module (240) is called when the plasma state value predicted by the plasma state prediction module has a difference greater than or equal to a threshold from a preset range, when the process state predicted by the process state prediction module has a difference greater than or equal to a threshold from a preset process state range, and when the process result predicted by the process result prediction module has a difference greater than or equal to a threshold from a range of process results measured after the actual process is completed.

[0083] More specifically, if the predicted plasma state value from the plasma state prediction module differs from a preset range by a threshold value or more, the planning module (240) can receive input from the user, as shown in FIG. 3, of change items, change amounts, process parameters, number of processes, process execution time, etc. that have changed in the plasma state. Through this, the electronic device (200) can use FIG. 3 as basic data for determining the inspection time of the plasma equipment and whether the plasma equipment is in operation.

[0084] If the process state predicted by the process state prediction module differs by a threshold or more from the range of the preset process state, the planning module (240) can receive input from the user of the process influence related to the change item in which the plasma state changed, as shown in FIG. 3.

[0085] The planning module (240) can use an artificial intelligence algorithm to plan which parameters of the plasma equipment should be adjusted to which values ​​when a difference occurs between the process results predicted by the process result prediction module and the range of the actual measured process results that exceeds a threshold value.

[0086] The planning module (240) can generate a control combination based on a control factor that can be adjusted under parameter control combination conditions that have information on an adjustable item in at least one process among a process in which a predicted plasma state value deviates from a range of preset plasma state values ​​by a threshold or more, a process in which a predicted process state deviates from a range of preset process states by a threshold or more, and a process in which a predicted process result deviates from a range of an actually measured process result by a threshold or more, as shown in FIG. 4.

[0087] The process control combination can perform learning by applying the sensing data acquired from the performed process and the predicted plasma state value, process state, and process result corresponding to the sensing data to the reinforcement learning algorithm, which is an artificial intelligence algorithm.

[0088] The planning module (240) performs a process impact evaluation on the predicted control combination when the prediction of the process control combination is completed using a reinforcement learning algorithm. The process impact evaluation determines whether the control combination exists within the range defined in the parameter control constraints defined in FIG. 4, and if it is within the defined range, the control information is stored in the recommended parameter control combination and parameter adjustment amount to be used as learning data for more stable control of the plasma equipment, and FIGS. 3 and 4 can be loaded into the message queue (220) as learning performance results.

[0089] For example, when performing process control in a state where the electron density state changes at 2024.02.16. 14:10:20 and the process influence is determined to be 1.1% as in FIG. 4, the planning module (240) can adjust the Power item to an importance of 10, the Pressure item to an importance of 3, the MFC value position to an importance of 3, and the Pumping Speed ​​to an importance of 2 in the parameter control recommendation combination. In addition, the planning module (240) can set the control to increase the Power by 5 Watts according to the importance determination, and store the corresponding value in the parameter adjustment amount.

[0090] In step 215, the control module (250) of the electronic device (200) controls the parameter values ​​based on the process control combination set in the planning module (240). For example, if the planning module (240) is set to control in the direction of increasing the power to 5 Watt, the control module (250) controls to increase the power to 5 Watt according to the setting of the planning module (240). Then, if the process result is completed normally, the control module (250) marks the process control result of FIG. 4 as success, and if it fails, it marks it as failure. If the equipment control parameter value extracted from the reinforcement learning algorithm of the planning module (240) satisfies the condition of the parameter control constraint, the control module (250) can stop the process and record it in the process stop of FIG. 4. In addition, the control module (250) can load the learning performance result of Fig. 4 into the message queue (220).

[0091] In step 217, the electronic device (200) terminates learning when a learning termination signal for the artificial intelligence algorithm included in the modules is received from the user, and if the learning termination signal is not received, the electronic device returns to step 203 and re-performs steps 203 to 215 to continuously perform learning of the artificial intelligence algorithm.

[0092] Figure 5 is a flowchart illustrating a method for reverse parameter inference according to an embodiment of the present invention. The reverse parameter inference method refers to a method for reversely inferring equipment parameters that must be set in the plasma equipment to derive the desired process results. The electronic device (200) can utilize the artificial intelligence algorithm learned using Figures 2 to 4 to reverse parameter inference.

[0093] Referring to FIG. 5, in step 501, the electronic device (200) checks whether a start signal for reverse inference is received from the user. If the start signal is received in step 501, the electronic device (200) calls the reverse inference module (260) and performs step 503. The reverse inference module (260) can exchange data with the message queue (220), the prediction module (230), the planning module (240), and the control module (250) as needed to infer equipment parameters.

[0094] In step 503, the reverse inference module (260) receives a process result to be achieved (hereinafter referred to as the "achieved process result") from the user. At this time, the achieved process result input by the user may be a deposition thickness and etching profile, etc. to be confirmed as a result of the process after the process is completed.

[0095] In step 505, the reverse inference module (260) predicts the final plasma state value for deriving the achievement process result through data transmission and reception with the message queue (220) or the prediction module (230). More specifically, the reverse inference module (260) applies the achievement process result to a regression-based artificial intelligence algorithm stored in the reverse inference module (260). The reverse inference module (260) applies the achievement process result to predict the plasma state value for deriving the achievement process result among the plasma state values ​​loaded in the message queue (220) or the plasma state values ​​learned and predicted by the prediction module (230).

[0096] In step 507, the reverse inference module (260) calls the plasma state value corresponding to the achieved process result based on the amount of change stored in the table as shown in FIGS. 3 and 4. At this time, if the plasma state value corresponding to the achieved process result does not exist in the table, the reverse inference module (260) can predict the plasma state value corresponding to the achieved process result through learning.

[0097] In step 509, the reverse inference module (260) can select an appropriate plasma state value as the final plasma state value to derive the achievement process result by applying the predicted plasma state value and the called plasma state value to the artificial intelligence algorithm.

[0098] In step 511, the reverse inference module (260) predicts the final sensing data for deriving the final plasma state value. More specifically, the reverse inference module (260) applies the final plasma state value to a regression-based artificial intelligence algorithm. The reverse inference module (260) applies the final plasma value to predict the sensing data for deriving the final plasma state value among the sensing data loaded in the message queue (220) or the sensing data learned and predicted in the prediction module (230).

[0099] In step 513, the reverse inference module (260) calls the sensing data measured at the same time as the final plasma state value among the plasma state values ​​stored in the table as shown in FIGS. 3 and 4. At this time, if the sensing data measured at the same time as the final plasma state value does not exist, the reverse inference module (260) can predict the sensing data corresponding to the final plasma state value through learning.

[0100] In step 515, the reverse inference module (260) can select appropriate sensing data as the final sensing data to derive the final plasma state value by applying the predicted sensing data and the called sensing data to an artificial intelligence algorithm.

[0101] In step 517, the reverse inference module (260) predicts the final equipment parameter values ​​for deriving the final sensing data. More specifically, the reverse inference module (260) applies the final sensing data to a regression-based artificial intelligence algorithm. The reverse inference module (260) applies the final sensing data to predict the parameter values ​​for deriving the final sensing data among the parameter values ​​loaded in the message queue (220) or the parameter values ​​learned and predicted by the prediction module (230).

[0102] In step 519, the reverse inference module (260) calls the parameter values ​​measured and stored at the same or similar time as the final sensing data among the parameter values ​​stored in the table as shown in FIGS. 3 and 4. At this time, if the parameter values ​​measured and stored at the same or similar time as the final sensing data do not exist in the table, the reverse inference module (260) can predict the parameter values ​​corresponding to the final sensing data through learning.

[0103] In step 521, the reverse inference module (260) can select an appropriate parameter value as the final equipment parameter value to derive the final sensing data by applying the predicted parameter value and the called parameter value.

[0104] In step 523, the reverse inference module (260) calls the control module (250) to provide the final equipment parameter values ​​selected by the control module (250), and the control module (250) controls the parameter values ​​of the plasma equipment using the final equipment parameter values. Through this, the present invention has the effect of controlling in real time a power device, a gas inlet device, or a vacuum pump installed in the plasma equipment to achieve the process result to be achieved by reverse inferring the equipment parameter values ​​that must be set in the plasma equipment to derive the process result to be achieved.

[0105] Figure 6 is an example screen diagram showing a screen configuration for parameter inference according to an embodiment of the present invention.

[0106] Referring to FIG. 6, FIG. 6a shows a plasma state value to be obtained for a process result to be achieved, for example, a plasma electron density profile value, FIG. 6b shows a plasma optical diagnostic data generation result corresponding to the plasma state value, for example, sensing data, and FIG. 6c shows equipment power parameters and equipment pressure parameters to be input for the sensing data, for example, equipment parameter values.

[0107] At this time, the user can check in real time that the sensing data and equipment parameter values ​​change while adjusting the plasma electron density profile value in Fig. 6a. Through this, the present invention has the effect of controlling the power device, gas inlet device, or vacuum pump installed in the plasma equipment in real time to achieve the desired process result by inferring the equipment parameter values ​​that must be set in the plasma equipment to derive the desired process result.

[0108] The embodiments of the present invention disclosed in this specification and drawings are merely specific examples intended to facilitate understanding and easily explain the technical content of the present invention, and are not intended to limit the scope of the present invention. Therefore, the scope of the present invention should be interpreted to include all modifications or variations derived based on the technical concept of the present invention, in addition to the embodiments disclosed herein.

Claims

1. A reverse inference module that reversely infers the final plasma state value to derive the desired achievement process result based on the generated learning performance result, reversely infers the final sensing data to be acquired from multiple sensors corresponding to the final plasma state value, and reversely infers the final parameter value of the plasma equipment corresponding to the final sensing data; and A control module that controls the plasma equipment based on the final parameter values; A parameter inference device characterized by including:

2. In paragraph 1, A collection module that collects sensing data obtained from the plurality of sensors during operation of the plasma equipment; A prediction module that applies the sensing data to an artificial intelligence algorithm to predict a plasma state value, applies at least one of the sensing data and the plasma state value to an artificial intelligence algorithm to predict a process state, and predicts a process result using the predicted process state; and A planning module that applies the plasma state value, the process state, and the process result to an artificial intelligence algorithm to identify equipment parameter values ​​that require control and generate the learning performance results; A parameter inference device characterized by including:

3. In paragraph 2, The above control module, A parameter inference device characterized in that it controls the parameters of the plasma equipment based on the equipment parameter values ​​confirmed in the above planning module.

4. In paragraph 3, A message queue that loads information derived from the collection module, the prediction module, the planning module, the control module, and the reverse inference module; A parameter inference device characterized by including:

5. In paragraph 4, The above prediction module, It includes a plasma state prediction module that predicts the plasma state value, a process state prediction module that predicts the process state, and a process result prediction module that predicts the process result. A parameter inference device characterized in that each of the above prediction modules uses a different artificial intelligence algorithm.

6. In paragraph 5, The above plasma state prediction module, the process state prediction module and the process result prediction module, A parameter inference device characterized in that it operates independently by collecting information loaded in the above message queue as needed.

7. In paragraph 6, The above reverse inference module is, A parameter inference device characterized in that the final plasma state value is inferred by comparing the plasma state value predicted from the message queue with the plasma state value called from the learning performance result generated from the planning module by applying the above-described achievement process result to the artificial intelligence algorithm for reverse inference.

8. In paragraph 7, The above reverse inference module is, A parameter inference device characterized in that the final sensing data is inferred by comparing the sensing data loaded in the message queue with the sensing data called from the learning performance result generated in the planning module by applying the final plasma state value to the artificial intelligence algorithm for the reverse inference.

9. In paragraph 8, The above reverse inference module is, A parameter inference device characterized in that the final sensing data is applied to the artificial intelligence algorithm for the reverse inference, the final parameter value is inferred by comparing the equipment parameter value loaded in the message queue with the equipment parameter value called from the learning performance result generated in the planning module.

10. In paragraph 2, A parameter inference device characterized in that the above plasma state value and the above process state are predicted through a plasma simulator.

11. A step of receiving the results of the process that the electronic device is trying to achieve; A step of inferring the final plasma state value for deriving the achievement process result based on the learning performance result generated by the electronic device; A step in which the electronic device reversely infers the final sensing data to be acquired from a plurality of sensors in response to the final plasma state value based on the parasitic learning performance result; A step in which the electronic device reversely infers the final parameter value of the plasma equipment corresponding to the final sensing data based on the parasitic learning performance result; and A step in which the electronic device controls the plasma equipment based on the final parameter value; A parameter inference method characterized by including:

Images