Beam uniformity adjustment methods, devices, storage media and electronic equipment

Abstract

This invention provides a method, apparatus, storage medium, and electronic device for beam uniformity adjustment. The method includes: acquiring historical adjustment data; determining the current training state vector from the historical adjustment data and calling an applied adjustment model to determine the device adjustment vector corresponding to the current training state vector; determining the adjustment model loss value of the applied adjustment model based on the device adjustment vector; updating the model parameters in the applied adjustment model in the direction of reducing the adjustment model loss value; iteratively executing the determination of the current training state vector from the historical adjustment data until the target model training conditions are met, thereby completing the model training of the applied adjustment model using historical adjustment data. The applied adjustment model supports guidance for beam uniformity adjustment of a target ion implanter. Embodiments of this invention can improve the accuracy of beam uniformity adjustment, thereby improving the adjustment efficiency of beam uniformity.

Description

Technical Field

[0001] This invention relates to the field of semiconductor technology, and in particular to a method, apparatus, storage medium, and electronic device for beam current uniformity adjustment. Background Technology

[0002] Currently, with the development of the information age, the semiconductor industry has been widely applied in various fields. As the performance requirements for semiconductor chips continue to increase, semiconductor integrated circuit manufacturing processes are rapidly developing, leading to higher demands on the performance of related manufacturing equipment. Among these, the ion implanter, as a core piece of equipment in the semiconductor device manufacturing process, can perform ion doping on the surface region of a semiconductor, changing the carrier concentration and conductivity type. The uniformity of its beam has a crucial impact on the consistency of device performance within the wafer; therefore, controlling beam uniformity during ion implantation is particularly important, and beam uniformity has become a key indicator for evaluating the performance of the ion implanter. Related technologies obtain device adjustment vectors by solving linear equations to adjust the ion implanter, thereby achieving beam uniformity regulation. However, the effect of device adjustment on beam uniformity is non-linear, making it difficult to obtain accurate adjustment values ​​by solving linear equations, resulting in low regulation efficiency. Therefore, there is currently no good solution for improving the accuracy of beam uniformity regulation to increase regulation efficiency. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a beam uniformity adjustment method, apparatus, storage medium, and electronic device to solve the problems of low efficiency in beam uniformity adjustment caused by related technologies. That is, embodiments of the present invention can train the application adjustment model using historical adjustment data to obtain a pre-trained application adjustment model, which can effectively improve the model performance of the application adjustment model. Then, the beam uniformity of the target ion implanter can be adjusted by the application adjustment model with better model performance, which can effectively improve the accuracy of beam uniformity adjustment and thus effectively improve the adjustment efficiency of beam uniformity.

[0004] According to one aspect of the present invention, a beam uniformity adjustment method is provided, the method comprising:

[0005] Acquire historical adjustment data, which includes multiple historical state vectors of the target ion implanter. Each historical state vector includes the beam state vector and device state vector of the target ion implanter at a historical acquisition moment.

[0006] The current training state vector is determined from the historical adjustment data, and the application adjustment model is invoked to determine the device adjustment vector corresponding to the current training state vector.

[0007] Based on the device adjustment vector, the adjustment model loss value of the application adjustment model is determined; and the model parameters in the application adjustment model are updated in the direction of reducing the adjustment model loss value.

[0008] The process iteratively executes the determination of the current training state vector from the historical adjustment data until the target model training conditions are met, thereby completing the model training of the application adjustment model using the historical adjustment data. The application adjustment model supports the guidance of beam uniformity adjustment for the target ion implanter.

[0009] According to another aspect of the present invention, a beam uniformity adjustment device is provided, the device comprising:

[0010] The acquisition unit is used to acquire historical adjustment data, which includes multiple historical state vectors of the target ion implanter. Each historical state vector includes the beam state vector and device state vector of the target ion implanter at a historical acquisition moment.

[0011] The processing unit is used to determine the current training state vector from the historical adjustment data, and call the applied adjustment model to determine the device adjustment vector corresponding to the current training state vector.

[0012] The processing unit is further configured to determine the adjustment model loss value of the application adjustment model based on the device adjustment vector; and update the model parameters in the application adjustment model in the direction of reducing the adjustment model loss value.

[0013] The processing unit is further configured to iteratively execute the process of determining the current training state vector from the historical adjustment data until the target model training conditions are met, so as to complete the model training of the application adjustment model through the historical adjustment data, wherein the application adjustment model supports the use of guiding the beam uniformity adjustment of the target ion implanter.

[0014] According to another aspect of the present invention, an electronic device is provided, the electronic device including a processor and a memory storing a program, wherein the program includes instructions that, when executed by the processor, cause the processor to perform the methods mentioned above.

[0015] According to another aspect of the present invention, a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the methods mentioned above is provided.

[0016] This invention provides embodiments that can acquire historical conditioning data, including multiple historical state vectors of the target ion implanter. Each historical state vector includes the beam state vector and device state vector of the target ion implanter at a historical acquisition moment. Based on this, the current training state vector can be determined from the historical conditioning data, and the applied conditioning model can be invoked to determine the device conditioning vector corresponding to the current training state vector. Correspondingly, the conditioning model loss value of the applied conditioning model can be determined based on the device conditioning vector, and the model parameters in the applied conditioning model are updated in the direction of reducing the conditioning model loss value. Furthermore, the process of determining the current training state vector from the historical conditioning data can be iteratively executed until the target model training conditions are met, thus completing the model training of the applied conditioning model using historical conditioning data. The applied conditioning model supports guidance for beam uniformity adjustment of the target ion implanter. As can be seen, the embodiments of the present invention can train the application adjustment model using historical adjustment data to obtain a pre-trained application adjustment model, which can effectively improve the model performance of the application adjustment model. Then, the beam uniformity of the target ion implanter can be adjusted by the application adjustment model with better model performance, which can effectively improve the accuracy of beam uniformity adjustment. That is, the current device adjustment vector with higher accuracy can be obtained by applying the adjustment model, so as to effectively improve the adjustment efficiency of beam uniformity. Attached Figure Description

[0017] Further details, features, and advantages of the invention are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0018] Figure 1 A schematic flowchart of a beam uniformity adjustment method according to an exemplary embodiment of the present invention is shown;

[0019] Figure 2 A schematic diagram illustrating the effect of a change in magnetic pole position on a beam according to an exemplary embodiment of the present invention is shown;

[0020] Figure 3 A schematic flowchart of another beam uniformity adjustment method according to an exemplary embodiment of the present invention is shown;

[0021] Figure 4 A schematic flowchart of another beam uniformity adjustment method according to an exemplary embodiment of the present invention is shown;

[0022] Figure 5 A schematic diagram of a parallel lens according to an exemplary embodiment of the present invention is shown;

[0023] Figure 6 A schematic block diagram of a beam uniformity adjustment device according to an exemplary embodiment of the present invention is shown;

[0024] Figure 7 A structural block diagram of an exemplary electronic device that can be used to implement embodiments of the present invention is shown. Detailed Implementation

[0025] Embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the invention. It should be understood that the accompanying drawings and embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the invention.

[0026] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0027] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0028] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0029] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0030] It should be noted that the execution subject of the beam uniformity adjustment method provided in this embodiment of the invention can be one or more electronic devices, and this invention does not limit this; wherein, the electronic device can be a terminal (i.e., a client) or a server. Therefore, when the execution subject includes multiple electronic devices, and among the multiple electronic devices includes at least one terminal and at least one server, the beam uniformity adjustment method provided in this embodiment of the invention can be jointly executed by the terminal and the server. Accordingly, the terminal mentioned herein can include, but is not limited to: smartphones, tablets, laptops, desktop computers, smartwatches, smart voice interaction devices, smart home appliances, vehicle terminals, aircraft, etc. The server mentioned herein can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms, etc.

[0031] Based on the above description, this embodiment of the invention proposes a beam uniformity adjustment method, which can be executed by the aforementioned electronic device (terminal or server); or, the beam uniformity adjustment method can be executed jointly by the terminal and the server. For ease of explanation, the following description will use the execution of the beam uniformity adjustment method by an electronic device as an example; such as Figure 1 As shown, the beam uniformity adjustment method may include the following steps S101-S104:

[0032] S101, acquire historical conditioning data. The historical conditioning data includes multiple historical state vectors of the target ion implanter. Each historical state vector includes the beam state vector and device state vector of the target ion implanter at a historical acquisition moment.

[0033] A beam state vector can be used to indicate the distribution of current at different locations. That is, a beam state vector can include the current magnitude at each of multiple sampling locations. Optionally, this embodiment of the invention does not limit the number of sampling locations. For example, the multiple sampling locations can be between -180 mm and 180 mm, and the interval between any two sampling locations can be set empirically or according to actual needs; this embodiment of the invention does not limit this. Optionally, this embodiment of the invention can measure the beam state vector of the target ion implanter at any acquisition time using any measurement device.

[0034] Optionally, a device state vector may include device information of at least one device in the target ion implanter (i.e., device information of each device in at least one device); optionally, at least one device may include, but is not limited to, at least one of the following: at least one coil (e.g., a multi-coil array) and at least one magnetic block (e.g., a multi-pole array), etc., which are not limited in this embodiment of the invention. Optionally, when a device is a coil, the device information of the device may be the current magnitude of the device; when a device is a magnetic block, the device information of the device may be the distance between the device and the beam (e.g., the distance between the device and the final arrival position of the beam), that is, the device information of the device may be the position information of the device. Optionally, a state vector may be formed by concatenating a beam state vector and a device state vector. For example, a historical state vector may be a concatenated vector between the beam state vector and the device state vector at a historical acquisition moment, which can be used to reflect the environment at the corresponding acquisition moment; based on this, the spatial dimension of the state space (i.e., the space in which the state vector formed by concatenating the beam state vector and the device state vector is distributed) may depend on the number of beam sampling points (i.e., sampling positions) and the actual number of physical devices.

[0035] Optionally, the target ion implanter can be any broadband beam ion implanter (which can be used to generate broadband ion beams), and the embodiments of the present invention do not limit this; for example, the target ion implanter can be one or more ion implanters with the same structure (such as the same devices), etc.

[0036] In this embodiment of the invention, the methods for obtaining historical adjustment data may include, but are not limited to, the following:

[0037] The first method of acquisition: The electronic device stores historical adjustment data in its own storage space. In this case, the electronic device can directly obtain the historical adjustment data from its own storage space.

[0038] The second method of acquisition: Electronic devices can obtain historical regulation data download links. In this case, electronic devices can download historical regulation data through the historical regulation data download links to obtain historical regulation data, and so on.

[0039] S102, determine the current training state vector from the historical adjustment data, and call the applied adjustment model to determine the device adjustment vector corresponding to the current training state vector.

[0040] Optionally, the electronic device may sequentially determine a historical state vector from the historical adjustment data and use the currently determined historical state vector as the current training state vector to determine the current training state vector from the historical adjustment data; or, it may randomly determine at least one historical state vector from the historical adjustment data and use the currently randomly determined at least one historical state vector as the current training state vector, etc.; the embodiments of the present invention do not limit this.

[0041] In this embodiment of the invention, the electronic device can input the current training state vector into the application adjustment model (i.e., the current application adjustment model) and output the device adjustment vector corresponding to the current training state vector, so as to realize the calling of the application adjustment model and determine the device adjustment vector corresponding to the current training state vector.

[0042] Optionally, a regulation model can be a neural network model, that is, the application regulation model can be a neural network model; optionally, the model structure of the application regulation model can be set according to experience or according to actual needs, and the embodiments of the present invention do not limit this. It should be understood that before training the application regulation model, that is, before using the application regulation model for the first time, the electronic device can randomly initialize the application regulation model, that is, randomly initialize the model parameters in the application regulation model.

[0043] Optionally, a device adjustment vector may include adjustment information of each device in at least one device in the target ion implanter; optionally, adjustment information may be adjustment action information (which can be used to indicate the magnitude of the device information adjustment (i.e., adjustment amount) for the corresponding device, such as moving 2 mm towards the beam (e.g., towards the beam forward direction or the final position reached by the beam, etc.) or reducing the current by a certain amount), or adjustment result information (which can be used to indicate the device information after adjustment for the corresponding device, such as adjusting to a position 3 mm away from the beam or adjusting to a certain current magnitude, etc.). This embodiment of the present invention does not limit this.

[0044] The space where the device adjustment vector resides can be the device adjustment space, i.e., the space where the device adjustment information is located. This ensures that after adjustment, each device is within its actual adjustable range, and its dimension is consistent with the actual number of devices. Optionally, the adjustment information range of any device indicated by the device adjustment space can be set according to requirements; or, when the adjustment information is an adjustment amount, it can be determined according to the current device state vector (such as the device state vector in the current training state vector), so that after any device is adjusted according to any value within the adjustment information range corresponding to any device, the device information of any device is within the device information range corresponding to any device, etc.; this embodiment of the invention does not limit this. Optionally, the device information range corresponding to any device can be set according to experience or actual needs, and this embodiment of the invention does not limit this.

[0045] S103, based on the device adjustment vector, determine the adjustment model loss value of the applied adjustment model; and update the model parameters in the applied adjustment model in the direction of reducing the adjustment model loss value.

[0046] In this embodiment of the invention, the model parameters in the applied adjustment model can be continuously updated, thereby continuously improving the model performance of the applied adjustment model, which can effectively improve the accuracy of the device adjustment vector output by the applied adjustment model.

[0047] S104, iteratively executes to determine the current training state vector from the historical conditioning data until the target model training conditions are met, so as to complete the model training of the applied conditioning model through the historical conditioning data. The applied conditioning model supports the beam uniformity adjustment of the target ion implanter.

[0048] Optionally, the target model training conditions may include model training conditions under the first model training stage and / or model training conditions under the second model training stage. When the target model training conditions include both the first and second model training stages, the electronic device can perform model training under the first model training stage using historical adjustment data, and then perform model training under the second model training stage using the same historical adjustment data to complete the model training—that is, to complete the model training of the application adjustment model using historical adjustment data. Optionally, the target model training conditions may be set based on experience or actual needs; this embodiment of the invention does not limit this. In other words, both the model training conditions under the first and second model training stages may be set based on experience or actual needs; this embodiment of the invention does not limit this. A model training condition can be used to indicate the termination condition for model training; that is, a model training condition can also be called a model training termination condition or a model training convergence condition.

[0049] Optionally, a model training condition can be that the number of iterations reaches a corresponding preset maximum number of iterations (e.g., the model training condition in the first model training stage can be that the number of iterations reaches a first preset iteration threshold, and the model training condition in the second model training stage can be that the number of iterations reaches a second preset iteration threshold); or, a model training condition can also be that the historical adjustment data is completely traversed (e.g., the model training condition in the first model training stage and the model training condition in the second model training stage can be that the historical adjustment data is completely traversed, etc.); the embodiments of the present invention do not limit this. Optionally, the first preset iteration threshold and the second preset iteration threshold can be set according to experience or according to actual needs, the embodiments of the present invention do not limit this; Optionally, the first preset iteration threshold and the second preset iteration threshold can be the same or different, the embodiments of the present invention do not limit this; Optionally, when performing model training in the second model training stage, the number of iterations can be reset to 1, that is, after completing the model training in the first model training stage, the electronic device can reset the number of iterations to 1.

[0050] Based on this, electronic devices can train the model in the first model training stage using historical adjustment data. After meeting the model training conditions in the first model training stage (such as meeting the first preset iteration threshold), they can start training the model in the second model training stage using historical adjustment data. When the model training conditions in the second model training stage are met, the target model training conditions are met. In other words, the model training process using historical adjustment data meets the target model training conditions, which can also be referred to as the model training process using historical adjustment data meeting the target model training conditions.

[0051] Optionally, the updated beam state vector can be determined according to either the method used in the first model training phase or the method used in the second model training phase. That is, in the first model training phase, the electronic device can determine the updated beam state vector according to the method used in the first model training phase; in the second model training phase, the electronic device can determine the updated beam state vector according to the method used in the second model training phase. The accuracy of the method used in the second model training phase may be higher than that of the method used in the first model training phase, and the first model training phase precedes the second model training phase. Optionally, the efficiency (i.e., convenience) of the method used in the first model training phase may be higher than that of the method used in the second model training phase.

[0052] Optionally, during the first model training phase, an updated beam state vector can be determined based on a uniformity simulation system. That is, the updated beam state vector during the first model training phase can be determined through a uniformity simulation system. In this case, the device adjustment vector can be input into the simulation system to output the corresponding updated beam state vector. Optionally, the electronic device can also obtain beam uniformity change information corresponding to the device adjustment vector (also referred to as the benefit information or beam uniformity change amount corresponding to the updated state vector corresponding to the device adjustment vector) through the simulation system. This embodiment of the invention does not limit this. Optionally, the beam uniformity change information corresponding to the device adjustment vector can be used to indicate the beam uniformity change after adjustment according to the device adjustment vector. Optionally, a beam uniformity change information can be any value. Optionally, beam uniformity can be described by a beam uniformity indication value. Therefore, a beam uniformity change information can be used to indicate the change in the beam uniformity indication value after adjustment according to the corresponding device adjustment vector, i.e., to indicate the change in beam uniformity after taking the action indicated by the device adjustment vector in the current state. Based on this, the simulation system may include, but is not limited to, at least one of the following: a relationship model between the device adjustment vector and the beam state vector (which can be used to determine the updated beam state vector corresponding to any device adjustment vector) and a relationship model between the device adjustment vector and beam uniformity change information (which can be used to determine the beam uniformity change information corresponding to any device adjustment vector), etc., and the embodiments of the present invention do not limit this. Optionally, the relationship model between the device adjustment vector and the beam state vector may be a system of linear equations constructed by multiple device adjustment vectors and the beam state vectors corresponding to each of the multiple device adjustment vectors (the beam state vector corresponding to a device adjustment vector may refer to the beam state vector after adjustment according to the corresponding device adjustment vector, i.e., the updated beam state vector corresponding to a device adjustment vector), and the relationship model between the device adjustment vector and the beam uniformity change information may be a system of linear equations constructed by multiple device adjustment vectors and the beam uniformity change information corresponding to each device adjustment vector (the beam uniformity change information corresponding to a device adjustment vector may refer to the beam uniformity change information between the beam uniformity of the beam state vector after adjustment according to the corresponding device adjustment vector and the beam uniformity before adjustment), etc.; the embodiments of the present invention do not limit this. Based on this, embodiments of the present invention can quickly determine the updated beam state vector corresponding to the adjustment vector of any device through a simulation system, thereby quickly achieving preliminary training of model parameters and effectively improving model training efficiency.

[0053] Optionally, a beam uniformity change information (i.e., gain information) can be the difference between the beam uniformity indicator value corresponding to an updated beam state vector and the beam uniformity indicator value corresponding to the corresponding beam state vector before adjustment (such as the current state vector or the current training state vector). Optionally, the beam uniformity indicator value can be negatively correlated with beam uniformity. In this case, a larger beam uniformity indicator value indicates lower beam uniformity (i.e., worse, more uneven), and a smaller beam uniformity indicator value indicates higher beam uniformity (i.e., better, more uniform). In this case, when a beam uniformity change value is positive, it indicates that the beam uniformity indicator value corresponding to the updated beam state vector is greater than the beam uniformity indicator value before adjustment (such as the beam uniformity indicator value corresponding to the current training state vector), which means that the beam uniformity corresponding to the updated beam state vector is lower. Conversely, when a beam uniformity change value is negative, it indicates that the beam uniformity indicator value corresponding to the updated beam state vector is less than the beam uniformity indicator value before adjustment, which means that the beam uniformity corresponding to the updated beam state vector is higher. Optionally, the beam uniformity indicator value can also be positively correlated with beam uniformity. In this case, a larger beam uniformity indicator value indicates higher beam uniformity, and a smaller beam uniformity indicator value indicates lower beam uniformity. In this situation, when a beam uniformity change value is positive, it indicates that the beam uniformity indicator value corresponding to the updated beam state vector is greater than the beam uniformity indicator value before adjustment (i.e., the beam uniformity change information can be the increase in the beam uniformity indicator value), which indicates that the beam uniformity corresponding to the updated beam state vector is higher. Conversely, when a beam uniformity change value is negative, it indicates that the beam uniformity indicator value corresponding to the updated beam state vector is less than the beam uniformity indicator value before adjustment (i.e., the beam uniformity change information can be the decrease in the beam uniformity indicator value), which indicates that the beam uniformity corresponding to the updated beam state vector is lower, and so on. This embodiment of the invention does not limit this aspect.

[0054] Optionally, the beam uniformity indicator value corresponding to a beam state vector can be the variance (i.e., the variance between the magnitudes of each current in the beam state vector) or standard deviation (i.e., the standard deviation between the magnitudes of each current in the beam state vector), etc. In this case, the beam uniformity indicator value is negatively correlated with the beam uniformity; that is, the larger the variance or standard deviation, the higher the volatility of the corresponding beam state vector, and the lower the beam uniformity. Alternatively, the beam uniformity indicator value corresponding to a beam state vector can be the reciprocal of the variance or standard deviation of the corresponding beam state vector. In this case, the beam uniformity indicator value can be positively correlated with the beam uniformity, and so on. The specific method for determining the beam uniformity indicator value is not limited in the embodiments of the present invention. For ease of explanation, the following description will use the example of a negative correlation between the beam uniformity indicator value and the beam uniformity as an example.

[0055] Optionally, in the second model training phase, an updated beam state vector can be determined based on the measurement of the angled Faraday cup in the target chamber of the target ion implanter. That is, the updated beam state vector in the second model training phase can be determined by measuring the angled Faraday cup in the target chamber of the target ion implanter. In other words, in the second model training phase, after adjusting the target ion implanter according to any device adjustment vector, the electronic device can obtain the measurement result of the beam state vector of the angled Faraday cup in the target ion implanter, and determine it as the updated beam state vector. In this case, the electronic device can output any device adjustment vector or send any device adjustment vector to the target ion implanter to adjust at least one device in the target ion implanter. Based on this, the measurement of the beam state vector is almost real-time, thus obtaining an accurate updated beam state vector, effectively improving the convergence rate of model training, and quickly obtaining a high-performance application adjustment model, etc.

[0056] In summary, the embodiments of the present invention can achieve rapid adjustment of beam uniformity by applying adjustment models, etc.; while the prior art adjusts through linear relationships, resulting in low adjustment efficiency, because the actual change in beam current and the change in device information are not linearly related. For example, assuming that at least one device includes at least one magnetic block, and device information can be the device position, the beam current changes (which may include the percentage change in current magnitude at any sampling position) caused by the movement of the upper magnetic pole of the multi-pole array 10 in the direction of the beam by 1mm, 2mm, 3mm, and 4mm from the reference position can be sequentially as follows: Figure 2 The fluctuations are shown as lines from small to large; therefore, traditional adjustment methods based on linear approximation suffer from inaccurate adjustment and long adjustment times. In this embodiment of the invention, a neural network is used to fit the relationship between the current environment and the device adjustment vector, giving full play to the fitting ability of the neural network and achieving precise adjustment of the device position or device current magnitude. This allows the beam current uniformity to meet the actual requirements (such as target process requirements) after a small number of device adjustments.

[0057] This invention provides embodiments that can acquire historical conditioning data, including multiple historical state vectors of the target ion implanter. Each historical state vector includes the beam state vector and device state vector of the target ion implanter at a historical acquisition moment. Based on this, the current training state vector can be determined from the historical conditioning data, and the applied conditioning model can be invoked to determine the device conditioning vector corresponding to the current training state vector. Correspondingly, the conditioning model loss value of the applied conditioning model can be determined based on the device conditioning vector, and the model parameters in the applied conditioning model are updated in the direction of reducing the conditioning model loss value. Furthermore, the process of determining the current training state vector from the historical conditioning data can be iteratively executed until the target model training conditions are met, thus completing the model training of the applied conditioning model using historical conditioning data. The applied conditioning model supports guidance for beam uniformity adjustment of the target ion implanter. As can be seen, the embodiments of the present invention can train the application adjustment model using historical adjustment data to obtain a pre-trained application adjustment model, which can effectively improve the model performance of the application adjustment model. Then, the beam uniformity of the target ion implanter can be adjusted by the application adjustment model with better model performance, which can effectively improve the accuracy of adjustment. That is, the current device adjustment vector with higher accuracy can be obtained by applying the adjustment model, so as to effectively improve the adjustment efficiency of beam uniformity.

[0058] Based on the above description, this embodiment of the invention also proposes a more specific method for beam uniformity adjustment. Accordingly, this beam uniformity adjustment method can be executed by the aforementioned electronic device (terminal or server); or, the beam uniformity adjustment method can be executed jointly by the terminal and the server. For ease of explanation, the following description will use the execution of this beam uniformity adjustment method by an electronic device as an example; please refer to [link to relevant documentation]. Figure 3 The beam uniformity adjustment method may include the following steps S301-S307:

[0059] S301, acquire historical conditioning data. The historical conditioning data includes multiple historical state vectors of the target ion implanter. Each historical state vector includes the beam state vector and device state vector of the target ion implanter at a historical acquisition time.

[0060] S302 determines the current training state vector from historical adjustment data and calls the application adjustment model to determine the device adjustment vector corresponding to the current training state vector.

[0061] S303 determines the updated beam state vector based on the device adjustment vector, and determines the current updated state vector corresponding to the current training state vector based on the updated beam state vector.

[0062] The updated beam state vector can be the updated beam state vector corresponding to the device adjustment vector, that is, the beam state vector after adjustment according to the device adjustment vector. For example, in the first model training stage, the updated beam state vector can be determined by the simulation system based on the device adjustment vector. In the second model training stage, the device adjustment vector can be output (such as displaying the device adjustment vector or sending the device adjustment vector to the target ion implanter), so that the target ion implanter is adjusted according to the device adjustment vector. After the target ion implanter is adjusted according to the device adjustment vector, the electronic device can obtain the current beam state vector measured by the angle Faraday cup (that is, the beam state vector obtained after adjustment according to the device adjustment vector), and use the currently obtained beam state vector as the updated beam state vector to determine the updated beam state vector, and so on.

[0063] Optionally, the current updated state vector corresponding to the current training state vector may include the updated beam state vector and the updated device state vector. Based on this, when determining the current updated state vector corresponding to the current training state vector based on the updated beam state vector, the electronic device can determine the updated device state vector corresponding to the device adjustment vector. For example, when an adjustment information is an adjustment result (i.e., the adjusted device information), the device adjustment vector can be used as the corresponding updated device state vector. When an adjustment information is an adjustment action, the device state vector in the current training state vector can be adjusted according to the device adjustment vector (e.g., moving the device information of any device 2mm in the beam direction (e.g., the beam forward direction) to obtain the updated device state vector, and so on. Correspondingly, the updated beam state vector and the updated device state vector can be concatenated to obtain the current updated state vector corresponding to the current training state vector, thus realizing the determination of the current updated state vector corresponding to the current training state vector based on the updated beam state vector.

[0064] S304. Based on the current training state vector and the current updated state vector, determine the current training tuple; and based on the current training tuple, determine at least one training sample.

[0065] Optionally, when determining the current training tuple based on the current training state vector and the current updated state vector, the electronic device can also determine the revenue information (i.e., the aforementioned beam uniformity change information) and process state information corresponding to the current updated state vector, and use the current training state vector, the current updated state vector, the device adjustment vector, and the revenue information and process state information corresponding to the current updated state vector to determine the current training tuple. In other words, the current training state vector, the current updated state vector, the device adjustment vector, and the revenue information and process state information corresponding to the current updated state vector can be added to the current training tuple to determine the current training tuple.

[0066] Optionally, when determining the revenue information corresponding to the current updated state vector, the difference between the beam uniformity indicator value corresponding to the beam state vector in the current updated state vector (i.e., the aforementioned updated beam state vector) and the beam uniformity indicator value corresponding to the beam state vector in the current training state vector can be used as the revenue information corresponding to the current updated state vector.

[0067] Optionally, when determining the process state information corresponding to the current updated state vector, the electronic device can determine whether the beam uniformity indication value corresponding to the updated beam state vector is less than or equal to a preset beam uniformity threshold. When the beam uniformity indication value corresponding to the updated beam state vector is less than or equal to the preset beam uniformity threshold, the process state information corresponding to the current updated state vector can be determined to be the target process requirement satisfaction indication information. When the beam uniformity indication value corresponding to the updated beam state vector is greater than the preset beam uniformity threshold, the process state information corresponding to the current updated state vector can be determined to be the target process requirement non-satisfaction indication information. Specifically, the target process requirement satisfaction indication information can be used to indicate that the beam uniformity meets the target process requirement, and the target process requirement non-satisfaction indication information can be used to indicate that the beam uniformity does not meet the target process requirement. The target process requirement can be the process requirement where the beam uniformity indication value meets (i.e., is less than or equal to) the preset beam uniformity threshold. The beam uniformity indication value can be used to indicate beam uniformity; that is, the target process requirement can refer to the process requirement where the beam uniformity meets the beam uniformity indicated by the preset beam uniformity threshold. Optionally, the preset beam uniformity threshold, the indication information for meeting the target process requirements, and the indication information for not meeting the target process requirements can all be set based on experience or based on actual needs. This embodiment of the invention does not limit this.

[0068] Based on this, when the process state information corresponding to the current updated state vector is an indication that the target process requirement is met, the process state information corresponding to the current updated state vector can be used to indicate that the beam uniformity of the beam state vector in the current updated state vector meets the target process requirement; when the process state information corresponding to the current updated state vector is an indication that the target process requirement is not met, the process state information corresponding to the current updated state vector can be used to indicate that the beam uniformity of the beam state vector in the current updated state vector does not meet the target process requirement.

[0069] Furthermore, the electronic device can add the current training tuples to the experience replay set; optionally, before the start of the first model training phase or the second model training phase, the electronic device can initialize the experience replay set to empty, i.e., clear the experience replay set. Based on this, the electronic device can determine at least one training sample from the experience replay set, thereby determining at least one training sample based on the current training tuples; optionally, the number of training samples in the at least one training sample can be m (m is a positive integer). In this case, the electronic device can randomly select m training tuples from the experience replay set, treating each selected training tuple as a training sample, thereby determining at least one training sample. Based on this, a training sample can be a training tuple, that is, a training tuple in the experience replay set; wherein, the experience replay set can be used to store the training tuples obtained in each training process (such as the current training tuple mentioned above). A training tuple may include a training state vector, a device adjustment vector corresponding to the corresponding training state vector, an updated state vector corresponding to the corresponding training state vector (i.e., the updated state vector corresponding to the device adjustment vector corresponding to the corresponding training state vector), the revenue information corresponding to the updated state vector (i.e., the updated state vector corresponding to the corresponding training state vector), and the process state information corresponding to the updated state vector. In other words, a training sample may include a training state vector, a device adjustment vector, an updated state vector, revenue information, and process state information.

[0070] S305, call the application rating model respectively to determine the application rating value corresponding to each training sample in at least one training sample; and calculate the adjustment model loss value of the application adjustment model based on the application rating value corresponding to each training sample.

[0071] Optionally, a scoring model can be a neural network model, meaning the applied scoring model can be a neural network model. It should be noted that this embodiment of the invention does not limit the specific model structure of the scoring model; that is, the model structure of the applied scoring model can be set according to experience or according to actual needs, and this embodiment of the invention does not limit this. Optionally, before using the applied scoring model for the first time, the electronic device can randomly initialize the applied scoring model, that is, randomly initialize the model parameters in the applied scoring model; or, the model parameters in the applied scoring model can be set according to experience or actual needs, etc.

[0072] In this embodiment of the invention, when the application scoring model is invoked to determine the application score value corresponding to each training sample in at least one training sample, for any training sample in at least one training sample, the electronic device can concatenate the training state vector and the device adjustment vector in any training sample to obtain the score estimation vector corresponding to any training sample. The electronic device can then invoke the application scoring model (i.e., the current application scoring model) to determine the application score value corresponding to any training sample based on the score estimation vector corresponding to any training sample. In other words, the score estimation vector corresponding to any training sample can be input into the application scoring model to output the application score value corresponding to any training sample through the application scoring model, thereby realizing the invocation of the application scoring model to determine the application score value corresponding to any training sample.

[0073] Optionally, the electronic device can also determine the target score value corresponding to each training sample, and determine the scoring model loss value of the application scoring model based on the target score value and the application score value corresponding to each training sample. Based on this, the model parameters in the application scoring model can be updated in the direction of reducing the scoring model loss value. In other words, the electronic device can also update the model parameters in the application scoring model through at least one training sample. In this embodiment of the invention, the electronic device can first update the model parameters in the application rating model using at least one training sample, and then update the model parameters in the application adjustment model using at least one training sample (that is, before step S305, the above-mentioned determination of the target rating value corresponding to each training sample can be triggered, and the rating model loss value of the application rating model can be determined based on the target rating value and the application rating value corresponding to each training sample, thereby updating the model parameters in the application rating model). In this case, the application rating value corresponding to each training sample used to calculate the adjustment model loss value can be determined by the application rating model after being updated by at least one training sample. That is, when determining the application rating value corresponding to each training sample used to calculate the adjustment model loss value, the current application rating model can be the application rating model updated by at least one training sample. Based on this, when determining the rating model loss value of the application rating model based on the target rating value and the application rating value corresponding to each training sample, the electronic device can also call the current application rating model (which refers to the application rating model that has not been updated by at least one current training sample) to determine the application rating value corresponding to each training sample, thereby determining the rating model loss value of the application rating model based on the target rating value and the application rating value corresponding to each training sample. It should be understood that when an electronic device calls any model, it means calling the current model. The model parameters in the current model are the latest values. If the model parameters in any model are updated, the model parameters in the current model can be the updated model parameters.

[0074] Optionally, a training tuple also includes the revenue information and process state information corresponding to the updated state vector in the corresponding training tuple, and one training sample is one training tuple. Based on this, when determining the target score value corresponding to each training sample, for any training sample in at least one training sample, the electronic device can determine whether the beam uniformity corresponding to any training sample (the beam uniformity corresponding to any training sample can refer to the beam uniformity corresponding to the beam state vector included in the updated state vector of any training sample) meets the target process requirements based on the process state information in any training sample. If the beam uniformity corresponding to any training sample meets the target process requirements, the revenue information in any training sample is used as the target score value corresponding to any training sample. If the beam uniformity corresponding to any training sample does not meet the target process requirements, a reference scoring model can be called to determine the reference score value corresponding to any training sample based on the updated state vector in any training sample, and the target score value corresponding to any training sample can be determined based on the revenue information in any training sample and the reference score value corresponding to any training sample. Optionally, the electronic device can determine a reference scoring coefficient and, using the gain information in any training sample, the reference scoring coefficient, and the reference scoring value corresponding to any training sample, determine the target scoring value corresponding to any training sample. Optionally, the reference scoring coefficient and the reference scoring value corresponding to any training sample can be multiplied to obtain the multiplication result, and the sum of the gain information in any training sample and the multiplication result can be used as the target scoring value corresponding to any training sample. Optionally, the reference scoring coefficient can be set based on experience or based on actual needs; this embodiment of the invention does not limit this.

[0075] Optionally, when calling the reference scoring model to determine the reference score value corresponding to any training sample based on the updated state vector in any training sample, the electronic device can call the reference adjustment model to determine the updated device adjustment vector corresponding to the updated state vector in any training sample (that is, input the updated state vector in any training sample into the reference adjustment model (i.e., the current reference adjustment model) so that the updated device adjustment vector corresponding to the updated state vector in any training sample can be output through the reference adjustment model), and can call the reference scoring model to determine the reference score value corresponding to any training sample based on the updated device adjustment vector (that is, input the updated device adjustment vector into the reference scoring model (i.e., the current reference scoring model) so that the reference score value corresponding to any training sample can be output through the reference scoring model).

[0076] Optionally, the model structure of the reference conditioning model can be the same as that of the applied conditioning model, and the model parameters in the reference conditioning model can be initialized with the initialization results of the model parameters in the applied conditioning model (also known as initialized model parameters). In other words, before model training, the initialized model parameters in the applied conditioning model can be used as the initialized model parameters in the reference conditioning model. Similarly, the model structure of the reference scoring model can be the same as that of the applied scoring model, and the model parameters in the reference scoring model can be initialized with the initialization results of the model parameters in the applied scoring model. In other words, before model training, the initialized model parameters in the applied scoring model can be used as the initialized model parameters in the reference scoring model.

[0077] Optionally, assuming that the number of training samples in at least one training sample is m, then when determining the scoring model loss value of the application scoring model based on the target score value and application score value corresponding to each training sample, the electronic device can use Formula 1.1 to determine the scoring model loss value of the application scoring model:

[0078] Formula 1.1

[0079] Where L can represent the scoring model loss function, used to calculate the scoring model loss value, y j Q(S) can represent the target score value corresponding to the j-th training sample in at least one training sample. j A j b) can represent the application score value corresponding to the j-th training sample, S j Let A represent the training state vector in the j-th training sample. j Let ...

[0080] Optionally, in other embodiments, the application score corresponding to each training sample used to calculate the adjustment model loss value can also be determined by an application score model that has not been updated by at least one current training sample, and the present invention does not limit this.

[0081] Optionally, when calculating the adjustment model loss value of the application adjustment model based on the application rating values ​​corresponding to each training sample, the electronic device can use Formula 1.2 to calculate the adjustment model loss value of the application adjustment model:

[0082] Equation 1.2

[0083] Here, 'a' can represent the model parameters in the applied adjustment model, and 'J' can represent the adjustment model loss function used to calculate the adjustment model loss value. Based on this, the electronic device can calculate the gradient using the backpropagation algorithm to update the model parameters in the applied adjustment model in the direction of reducing the adjustment model loss value, as described below.

[0084] S306, Update the model parameters in the applied adjustment model in the direction of reducing the loss value of the adjustment model.

[0085] In this embodiment of the invention, the adjustment model loss value can be determined by applying a scoring model. The model parameters in the applied scoring model can be updated, and the scoring model loss value of the applied scoring model can be determined based on at least one of the applied scoring model, a reference adjustment model, and a reference scoring model. Based on this, after each update of the model parameters in the applied adjustment model and the applied scoring model, the electronic device can also determine whether the reference update condition is met. If the reference update condition is met, the model parameters in the reference adjustment model can be updated based on the adjustment smoothing coefficient and the model parameters in the applied adjustment model, and the model parameters in the reference scoring model can be updated based on the scoring smoothing coefficient and the model parameters in the applied scoring model. If the reference update condition is not met, the model parameters in the reference adjustment model and the reference scoring model are not updated. Optionally, the electronic device can determine whether the reference update condition is met based on the current iteration number. For example, when the remainder between the current iteration number and the first value is equal to the second value (i.e., current iteration number % first value == second value), the reference update condition can be determined to be met; when the remainder between the current iteration number and the first value is not equal to the second value, the reference update condition can be determined not to be met, and so on. This embodiment of the invention does not limit this. Optionally, both the first and second values ​​can be set based on experience or based on actual needs, and the embodiments of the present invention do not limit this.

[0086] Optionally, the electronic device can update the model parameters in the reference conditioning model and the model parameters in the reference scoring model based on Equation 1.3:

[0087] Equation 1.3

[0088] Where b' can represent the model parameters in the reference scoring model, and a' can represent the model parameters in the reference moderated model. It can represent the score smoothing coefficient. The adjustment smoothing coefficient can be represented. Optionally, both the adjustment smoothing coefficient and the scoring smoothing coefficient can be set based on experience or actual needs, and this embodiment of the invention does not limit this; optionally, the adjustment smoothing coefficient and the scoring smoothing coefficient can be the same or different, and this embodiment of the invention does not limit this.

[0089] S307, iteratively executes to determine the current training state vector from the historical conditioning data until the target model training conditions are met, so as to complete the model training of the applied conditioning model through the historical conditioning data. The applied conditioning model supports the adjustment of the beam uniformity of the target ion implanter.

[0090] Optionally, after adding the current training tuple to the experience replay set, the electronic device can also use the current updated state vector as the current training state vector. In other words, the current training state vector can be updated to the current updated state vector. Based on this, after completing one model training process (i.e., completing one parameter update for the applied conditioning model and the applied scoring model, and / or completing the parameter update for the reference conditioning model and the reference scoring model when the reference update condition is met), the electronic device can also determine whether the current training state vector has completed the current model training process. If it is determined that the current training state vector has completed the current model training process, the above iterative execution can be triggered to determine the current training state vector from the historical conditioning data. That is, at this time, the current training state vector can be re-determined from the historical conditioning data, such as using the next training state vector in the historical conditioning data as the current training state vector. If it is determined that the current training state vector has not completed the current model training process, the applied conditioning model (i.e., the current applied conditioning model) can be called again to determine the device conditioning vector corresponding to the current training state vector, so as to continue updating the model parameters in the applied conditioning model, etc., until it is determined that the current training state vector has completed the current model training process.

[0091] Optionally, when determining whether the current training state vector has completed the current model training process, the electronic device may determine whether the beam uniformity corresponding to the beam state vector (which can be the updated beam state vector mentioned above) in the current training state vector meets the target process requirements; if it is determined that the beam uniformity corresponding to the beam state vector in the previous training state vector meets the target process requirements, then it can be determined that the current training state vector has completed the current model training process; if it is determined that the beam uniformity corresponding to the beam state vector in the previous training state vector does not meet the target process requirements, then it can be determined that the current training state vector has not completed the current model training process. Optionally, when determining whether the current training state vector has completed the current model training process, the electronic device may determine whether the beam uniformity corresponding to the beam state vector in the current training state vector meets the target process requirements. It may also determine whether the beam uniformity indicator value corresponding to the updated beam state vector is less than or equal to a preset beam uniformity threshold; or, it may determine whether the current process state information (i.e., the process state information in the current training tuple) is an indicator that meets the target process requirements; or, if the indicator that meets the target process requirements is greater than 1 and the indicator that does not meet the target process requirements is equal to 0, it may determine whether the current process state information is true, thereby achieving the determination of whether the beam uniformity corresponding to the beam state vector in the current training state vector meets the target process requirements, etc. This embodiment of the invention does not limit this aspect.

[0092] Furthermore, after meeting the target model training conditions, the electronic device can acquire the current beam state vector in the target ion implanter at preset detection intervals to determine the current beam uniformity in the target ion implanter (i.e., calculate the current beam uniformity). In other words, it can calculate the beam uniformity indicator value corresponding to the current beam state vector, which indicates the current beam uniformity. This allows detection of whether the current beam uniformity in the target ion implanter meets the target process requirements. Specifically, it detects whether the beam uniformity indicator value corresponding to the current beam state vector is less than or equal to a preset beam uniformity threshold, thus determining whether the current beam uniformity in the target ion implanter meets the target process requirements (i.e., determining whether the current beam uniformity meets the requirements). Optionally, the preset detection interval can be set based on experience or actual needs; this embodiment of the invention does not limit this. Optionally, the current beam state vector can be measured by moving a Faraday cup (i.e., it can be acquired by moving a Faraday cup) to obtain an accurate current beam state vector. Optionally, when it is detected that the current beam uniformity meets the target process requirements, the following uniformity adjustment steps may not be performed, that is, the following call to apply the adjustment model and determine the current device adjustment vector corresponding to the current state vector may not be triggered.

[0093] Optionally, when the current beam uniformity in the target ion implanter is detected to be inconsistent with the target process requirements, the electronic device can also determine the current state vector. The current state vector includes the beam state vector and device state vector of the target ion implanter at the current acquisition time. For example, the electronic device can also acquire the current device state vector to determine the current state vector based on the current beam state vector and the current device state vector. Based on this, an application adjustment model can be invoked to determine the current device adjustment vector corresponding to the current state vector. The current device adjustment vector is used to guide the adjustment of at least one device in the target ion implanter. Optionally, the electronic device can output the current device adjustment vector (e.g., display the current device adjustment vector) so that the target object (e.g., the manager of the target ion implanter) can perform device adjustment operations on the target ion implanter according to the current device adjustment vector to achieve adjustment of at least one device in the target ion implanter based on the current device adjustment vector. Alternatively, the electronic device can send the current device adjustment vector to the target ion implanter so that the target ion implanter can adjust at least one device according to the current device adjustment vector, etc. The embodiments of the present invention do not limit this.

[0094] Furthermore, the electronic device can iteratively execute the above-mentioned determination of the current state vector. That is, after the target ion implanter is adjusted, the current beam state vector and the current device state vector can be reacquired to achieve iterative determination of the current state vector, thereby obtaining the current device adjustment vector used to guide the adjustment of the target ion implanter, so that the target ion implanter can be continuously adjusted until the current beam uniformity in the target ion implanter meets the target process requirements. In the process of adjusting at least one device in the target ion implanter, the adjustment model is applied to support parameter updates when the model update conditions are met.

[0095] Optionally, before calling the application adjustment model and determining the device adjustment vector corresponding to the current state vector, the electronic device can determine whether the application adjustment model meets the model update conditions. If it is determined that the application adjustment model meets the model update conditions, the adjustment model loss value (i.e., the current adjustment model loss value) and the scoring model loss value (i.e., the current scoring model loss value) of the application adjustment model can be determined based on the current state vector and the previous state vector (i.e., the state vector at the previous acquisition time). This allows for optimization of the model parameters in the application adjustment model in the direction of reducing the adjustment model loss value, and optimization of the model parameters in the application scoring model in the direction of reducing the scoring model loss value. This enables parameter updates for both the application adjustment model and the application scoring model. Correspondingly, when it is determined that the reference update conditions are met, the model parameters in the reference adjustment model can be updated based on the adjustment smoothing coefficient and the model parameters in the application adjustment model; and the model parameters in the reference scoring model can be updated based on the scoring smoothing coefficient and the model parameters in the application scoring model, and so on. Based on this, the embodiments of the present invention can realize model training in the third model training stage to further improve the model performance. Optionally, if it is determined that the application adjustment model does not meet the model update conditions, the execution of the application adjustment model can be directly triggered to determine the device adjustment vector corresponding to the current state vector.

[0096] Optionally, when determining the adjustment model loss value of the applied adjustment model and the scoring model loss value of the applied scoring model based on the current state vector and the previous state vector, the electronic device can determine the current benefit information corresponding to the current state vector (i.e., the difference between the beam uniformity indication value corresponding to the beam state vector in the current state vector and the beam uniformity indication value corresponding to the beam state vector in the previous state vector), and determine whether the beam uniformity corresponding to the beam state vector in the current state vector meets the target process requirements; if the beam uniformity corresponding to the beam state vector in the current state vector meets the target process requirements, then the current benefit information is used as the current target scoring value; if the beam uniformity corresponding to the beam state vector in the current state vector does not meet the target process requirements, then the reference adjustment model is called to determine the current device adjustment vector corresponding to the current state vector, and the reference scoring model is called to determine the current reference scoring value based on the current state vector and the current device adjustment vector corresponding to the current state vector (the determination method of the reference scoring value is the same as the above, and will not be repeated here in this embodiment of the invention), and then the current target scoring value can be determined based on the current benefit information and the current reference scoring value. Furthermore, the difference between the current target score and the application score corresponding to the previous state vector can be used to determine the scoring model loss value of the application scoring model. Correspondingly, the adjustment model loss value of the application adjustment model can be determined based on the application score corresponding to the previous state vector. In this embodiment of the invention, the determination methods for the scoring model loss value and the adjustment model loss value are the same as those used in the first model training phase and the second model training phase, and will not be elaborated further here.

[0097] Optionally, when determining whether the applied adjustment model meets the model update conditions, it can be determined whether the previous state vector exists. If the previous state vector exists, it can be determined that the applied adjustment model meets the model update conditions; if the previous state vector does not exist, it can be determined that the applied adjustment model does not meet the model update adjustment. Alternatively, when the current beam uniformity in the target ion implanter is first detected to be inconsistent with the target process requirements at every preset detection interval, the current iteration count can be determined to be 0. After adjusting the target ion implanter once, the current iteration count can be incremented by 1 to update the current iteration count. In this case, it can be determined whether the current iteration count is equal to 0. If the current iteration count is equal to 0, it can be determined that the applied adjustment model does not meet the model update conditions; if the current iteration count is not equal to 0, it can be determined that the applied adjustment model meets the model update conditions. This achieves the determination of whether the applied adjustment model meets the model update conditions. Figure 4 As shown, etc.; the embodiments of the present invention do not limit this. Figure 4 The count in the table can represent the current iteration number.

[0098] In this embodiment of the invention, since the target ion implanter can be any ion implanter, that is, this embodiment of the invention does not limit at least one device in the target ion implanter, and it can be a multi-coil array and / or a multi-pole array; based on this, this embodiment of the invention has good generalizability and is applicable to the adjustment of multi-coil arrays and multi-pole arrays, and is also applicable to various modes such as magnetic pole symmetry adjustment and free adjustment in multi-pole arrays. For example, Figure 5 As shown, the multi-pole array and the multi-coil array are located at the front and rear ends of the parallel lens beam, respectively. In practical applications, the multi-coil array controls the magnetic field in the optical path by changing the current of each coil, while the multi-pole array controls the magnetic field by changing the distance between itself and the beam. Although the control principles are different, the idea of ​​adjusting the magnetic field is the same. The adjustment method proposed in this invention is applicable to both devices. The difference lies only in the information related to the device position or current magnitude in the device state vector and device adjustment vector.

[0099] In summary, the embodiments of the present invention may involve neural network models such as adjustment models. The model parameters in these neural network models can be updated during the actual adjustment process. That is, when it is detected that the current beam uniformity in the target ion implanter does not meet the target process requirements and the model update conditions are met during the actual application process, the model can be trained in the third model training stage. This allows the application adjustment model to give the most accurate device adjustment vector according to the actual situation of the current beam. It should be noted that this embodiment of the invention proposes a framework-based method, in which the reinforcement learning model involved can be selected autonomously during use. For example, reinforcement learning algorithms such as dual-delay deep deterministic policy gradient can also be used for model training. For example, the number of application scoring models and reference scoring models can be multiple. That is, this embodiment of the invention can train the model using an application conditioning model, multiple application scoring models, a reference conditioning model, and multiple reference scoring models. For example, the target score can be determined by the minimum value among the reference score values ​​under each reference scoring model. The loss value of the scoring model under the corresponding application scoring model can be determined based on the target score value and the application score value under each application scoring model (such as the application score value corresponding to each training sample under any application scoring model) to update the corresponding application scoring model. The conditioning model loss value of the application conditioning model can be determined by the mean among the application score values ​​under each application scoring model (such as the mean among the application score values ​​corresponding to any training sample under each application scoring model, which can be the application score value corresponding to any training sample). This embodiment of the invention does not limit this.

[0100] This invention can acquire historical conditioning data, including multiple historical state vectors of the target ion implanter. Each historical state vector includes the beam state vector and device state vector of the target ion implanter at a historical acquisition moment. The current training state vector is determined from the historical conditioning data, and an applied conditioning model is invoked to determine the device conditioning vector corresponding to the current training state vector. Then, based on the device conditioning vector, an updated beam state vector is determined, and based on the updated beam state vector, the current updated state vector corresponding to the current training state vector is determined. Based on this, a current training tuple is determined based on the current training state vector and the current updated state vector, and at least one training sample is determined based on the current training tuple. The applied scoring model is then invoked to determine the application score value corresponding to each training sample in the at least one training sample, and based on the application score value corresponding to each training sample, the conditioning model loss value of the applied conditioning model is calculated. The model parameters in the applied conditioning model are then updated in the direction of reducing the conditioning model loss value. Furthermore, the current training state vector can be iteratively determined from historical adjustment data until the target model training conditions are met, thus completing the model training of the applied adjustment model using historical adjustment data. The applied adjustment model supports the guidance of beam uniformity adjustment for the target ion implanter. It is evident that this embodiment of the invention can effectively improve model performance through historical adjustment data, thereby better fitting the relationship between the state vector and the device adjustment vector through the applied adjustment model, and accurately obtaining the device adjustment vector, so that beam uniformity meets the target process requirements after fewer device adjustments. Moreover, this embodiment of the invention proposes a three-step parameter training method for reinforcement learning models (including the aforementioned first model training stage, second model training stage, and third model training stage), which can fully utilize historical adjustment data for uniformity adjustment and effectively improve the model training speed.

[0101] Based on the description of the relevant embodiments of the beam uniformity adjustment method above, this invention also proposes a beam uniformity adjustment device, which can be a computer program (including program code) running in an electronic device; such as Figure 6 As shown, the beam uniformity adjustment device may include an acquisition unit 601 and a processing unit 602. The beam uniformity adjustment device can perform... Figure 1 or Figure 3 The beam uniformity adjustment method shown, i.e., the beam uniformity adjustment device, can operate the above-mentioned unit:

[0102] The acquisition unit 601 is used to acquire historical adjustment data, which includes multiple historical state vectors of the target ion implanter. Each historical state vector includes the beam state vector and device state vector of the target ion implanter at a historical acquisition moment.

[0103] The processing unit 602 is used to determine the current training state vector from the historical adjustment data, and call the applied adjustment model to determine the device adjustment vector corresponding to the current training state vector.

[0104] The processing unit 602 is further configured to determine the adjustment model loss value of the application adjustment model based on the device adjustment vector; and update the model parameters in the application adjustment model in the direction of reducing the adjustment model loss value.

[0105] The processing unit 602 is further configured to iteratively execute the process of determining the current training state vector from the historical adjustment data until the target model training conditions are met, so as to complete the model training of the application adjustment model through the historical adjustment data, wherein the application adjustment model supports the use of guiding the beam uniformity adjustment of the target ion implanter.

[0106] In one embodiment, when the processing unit 602 determines the adjustment model loss value of the applied adjustment model based on the device adjustment vector, it may specifically be used to:

[0107] Based on the device adjustment vector, an updated beam state vector is determined, and based on the updated beam state vector, the current updated state vector corresponding to the current training state vector is determined.

[0108] Based on the current training state vector and the current update state vector, determine the current training tuple; and based on the current training tuple, determine at least one training sample;

[0109] The application rating model is invoked to determine the application rating value corresponding to each training sample in the at least one training sample; and the adjustment model loss value of the application adjustment model is calculated based on the application rating value corresponding to each training sample.

[0110] In another embodiment, the processing unit 602 may also be used for:

[0111] The target score value corresponding to each training sample is determined respectively, and the score model loss value of the application score model is determined based on the target score value and the application score value corresponding to each training sample.

[0112] Update the model parameters in the applied scoring model in the direction of reducing the loss value of the scoring model.

[0113] In another implementation, a training tuple also includes the revenue information and process state information corresponding to the updated state vector in the corresponding training tuple, and one training sample is one training tuple; when the processing unit 602 determines the target score value corresponding to each training sample, it can be specifically used for:

[0114] For any training sample in the at least one training sample, based on the process state information in the training sample, it is determined whether the beam uniformity corresponding to the training sample meets the target process requirements.

[0115] If the beam uniformity corresponding to any training sample meets the target process requirement, then the benefit information in any training sample is used as the target score value corresponding to any training sample.

[0116] If the beam uniformity corresponding to any training sample does not meet the target process requirement, then the reference scoring model is invoked, and a reference score value corresponding to any training sample is determined based on the updated state vector in any training sample; and a target score value corresponding to any training sample is determined based on the benefit information in any training sample and the reference score value corresponding to any training sample.

[0117] In another embodiment, the target model training conditions include model training conditions under the first model training stage and / or model training conditions under the second model training stage; the updated beam state vector is determined according to the updated beam state vector determination method under the first model training stage or the updated beam state vector determination method under the second model training stage, the accuracy of the updated beam state vector determination method under the second model training stage is higher than the accuracy of the updated beam state vector determination method under the first model training stage, and the first model training stage precedes the second model training stage.

[0118] In another embodiment, the adjustment model loss value is determined by applying a scoring model, wherein the model parameters in the applied scoring model support updates, and the scoring model loss value of the applied scoring model is determined based on at least one of the applied scoring model, the reference adjustment model, and the reference scoring model. The processing unit 602 can also be used for:

[0119] After each update of the model parameters in the application adjustment model and the application scoring model, it is determined whether the reference update conditions are met.

[0120] If the reference update condition is met, the model parameters in the reference adjustment model are updated based on the adjustment smoothing coefficient and the model parameters in the application adjustment model; and the model parameters in the reference scoring model are updated based on the scoring smoothing coefficient and the model parameters in the application scoring model.

[0121] If the reference update conditions are not met, the model parameters in the reference adjustment model and the reference scoring model will not be updated.

[0122] In another embodiment, the processing unit 602 may also be used for:

[0123] When it is detected that the current beam uniformity in the target ion implanter does not meet the target process requirements, a current state vector is determined. The current state vector includes the beam state vector and the device state vector of the target ion implanter at the current acquisition time.

[0124] The application adjustment model is invoked to determine the current device adjustment vector corresponding to the current state vector. The current device adjustment vector is used to guide the adjustment of at least one device in the target ion implanter.

[0125] The determination of the current state vector is performed iteratively until the current beam uniformity in the target ion implanter meets the target process requirements; wherein, during the adjustment of at least one device in the target ion implanter, the applied adjustment model supports parameter updates when the model update conditions are met.

[0126] According to one embodiment of the present invention, Figure 6 Each unit in the beam uniformity adjustment device shown can be individually or entirely combined into one or more other units, or some of the units can be further divided into multiple functionally smaller units. This achieves the same operation without affecting the technical effect of the embodiments of the present invention. The above units are based on logical function division. In practical applications, the function of one unit can be implemented by multiple units, or the function of multiple units can be implemented by one unit. In other embodiments of the present invention, any beam uniformity adjustment device may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0127] According to another embodiment of the present invention, it is possible to perform operations such as those described above by running on a general-purpose electronic device, such as a computer, which includes processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM). Figure 1 or Figure 3 The computer program (including program code) involved in each step of the corresponding method shown, to construct such... Figure 6 The diagram illustrates a beam uniformity adjustment device and a beam uniformity adjustment method for implementing embodiments of the present invention. The computer program can be stored on, for example, a computer storage medium, loaded onto the aforementioned electronic device via the computer storage medium, and run therein.

[0128] This invention provides embodiments that can acquire historical conditioning data, including multiple historical state vectors of the target ion implanter. Each historical state vector includes the beam state vector and device state vector of the target ion implanter at a historical acquisition moment. Based on this, the current training state vector can be determined from the historical conditioning data, and the applied conditioning model can be invoked to determine the device conditioning vector corresponding to the current training state vector. Correspondingly, the conditioning model loss value of the applied conditioning model can be determined based on the device conditioning vector, and the model parameters in the applied conditioning model are updated in the direction of reducing the conditioning model loss value. Furthermore, the process of determining the current training state vector from the historical conditioning data can be iteratively executed until the target model training conditions are met, thus completing the model training of the applied conditioning model using historical conditioning data. The applied conditioning model supports guidance for beam uniformity adjustment of the target ion implanter. As can be seen, the embodiments of the present invention can train the application adjustment model using historical adjustment data to obtain a pre-trained application adjustment model, which can effectively improve the model performance of the application adjustment model. Then, the beam uniformity of the target ion implanter can be adjusted by the application adjustment model with better model performance, which can effectively improve the accuracy of adjustment. That is, the current device adjustment vector with higher accuracy can be obtained by applying the adjustment model, so as to effectively improve the adjustment efficiency of beam uniformity.

[0129] Based on the description of the method and apparatus embodiments above, an exemplary embodiment of the present invention also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, which, when executed by the at least one processor, causes the electronic device to perform the method according to an embodiment of the present invention.

[0130] An exemplary embodiment of the present invention also provides a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a computer's processor, is used to cause the computer to perform a method according to an embodiment of the present invention.

[0131] An exemplary embodiment of the present invention also provides a computer program product, including a computer program, wherein, when executed by a computer's processor, the computer program is used to cause the computer to perform a method according to an embodiment of the present invention.

[0132] refer to Figure 7The present invention will now be described in the form of a structural block diagram of an electronic device 700 that can serve as a server or client of the present invention, which is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0133] like Figure 7 As shown, the electronic device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. The RAM 703 may also store various programs and data required for the operation of the electronic device 700. The computing unit 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0134] Multiple components in electronic device 700 are connected to I / O interface 705, including: input unit 706, output unit 707, storage unit 708, and communication unit 709. Input unit 706 can be any type of device capable of inputting information to electronic device 700. Input unit 706 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of electronic device. Output unit 707 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 708 may include, but is not limited to, disk and optical disk. Communication unit 709 allows electronic device 700 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth™ devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0135] The computing unit 701 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 performs the various methods and processes described above. For example, in some embodiments, the beam uniformity adjustment method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 700 via ROM 702 and / or communication unit 709. In some embodiments, the computing unit 701 can be configured to perform the beam uniformity adjustment method by any other suitable means (e.g., by means of firmware).

[0136] Furthermore, it should be understood that the above-disclosed embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for adjusting beam uniformity, characterized in that, include: Acquire historical adjustment data, which includes multiple historical state vectors of the target ion implanter. Each historical state vector includes the beam state vector and device state vector of the target ion implanter at a historical acquisition moment. The current training state vector is determined from the historical adjustment data, and the application adjustment model is invoked to determine the device adjustment vector corresponding to the current training state vector. Based on the device adjustment vector, determine the adjustment model loss value of the application adjustment model; And update the model parameters in the applied adjustment model in the direction of reducing the loss value of the adjustment model; Wherein, determining the adjustment model loss value of the applied adjustment model based on the device adjustment vector includes: Based on the device adjustment vector, an updated beam state vector and an updated device state vector are determined, and based on the updated beam state vector and the updated device state vector, a current updated state vector corresponding to the current training state vector is determined. The updated beam state vector is the beam state vector adjusted according to the device adjustment vector. The benefit information and process status information corresponding to the current updated state vector are determined. The benefit information is the difference between the beam uniformity indicator value corresponding to the beam state vector in the current updated state vector and the beam uniformity indicator value corresponding to the beam state vector in the current training state vector. Specifically, when the beam uniformity indicator value corresponding to the updated beam state vector is less than or equal to a preset beam uniformity threshold, the process status information indicates that the target process requirement is met; when the beam uniformity indicator value corresponding to the updated beam state vector is greater than the preset beam uniformity threshold, the process status information indicates that the target process requirement is not met. Based on the current training state vector, the current update state vector, the device adjustment vector, the revenue information, and the process state information, a current training tuple is determined; and based on the current training tuple, at least one training sample is determined. The application rating model is invoked to determine the application rating value corresponding to each training sample in the at least one training sample; and the adjustment model loss value of the application adjustment model is calculated based on the application rating value corresponding to each training sample. The process iteratively executes the determination of the current training state vector from the historical adjustment data until the target model training conditions are met, thereby completing the model training of the application adjustment model using the historical adjustment data. The application adjustment model supports the guidance of beam uniformity adjustment for the target ion implanter.

2. The method according to claim 1, characterized in that, The method further includes: The target score value corresponding to each training sample is determined respectively, and the score model loss value of the application score model is determined based on the target score value and the application score value corresponding to each training sample. Update the model parameters in the applied scoring model in the direction of reducing the loss value of the scoring model.

3. The method according to claim 2, characterized in that, A training tuple also includes the revenue information and process state information corresponding to the updated state vector in the corresponding training tuple. One training sample is one training tuple. The step of determining the target score value corresponding to each training sample includes: For any training sample in the at least one training sample, based on the process state information in the training sample, it is determined whether the beam uniformity corresponding to the training sample meets the target process requirements. If the beam uniformity corresponding to any training sample meets the target process requirement, then the benefit information in any training sample is used as the target score value corresponding to any training sample. If the beam uniformity corresponding to any training sample does not meet the target process requirement, then the reference scoring model is invoked, and a reference score value corresponding to any training sample is determined based on the updated state vector in any training sample; and a target score value corresponding to any training sample is determined based on the benefit information in any training sample and the reference score value corresponding to any training sample.

4. The method according to claim 1, characterized in that, The target model training conditions include model training conditions under the first model training stage and / or model training conditions under the second model training stage; the updated beam state vector is determined according to the updated beam state vector determination method under the first model training stage or the updated beam state vector determination method under the second model training stage. The accuracy of the updated beam state vector determination method under the second model training stage is higher than that of the updated beam state vector determination method under the first model training stage. The first model training stage precedes the second model training stage.

5. The method according to any one of claims 1-4, characterized in that, The adjustment model loss value is determined by applying a scoring model, wherein the model parameters in the applied scoring model can be updated, and the scoring model loss value of the applied scoring model is determined based on at least one of the applied scoring model, the reference adjustment model, and the reference scoring model. The method further includes: After each update of the model parameters in the application adjustment model and the application scoring model, it is determined whether the reference update conditions are met. If the reference update condition is met, the model parameters in the reference adjustment model are updated based on the adjustment smoothing coefficient and the model parameters in the application adjustment model; and the model parameters in the reference scoring model are updated based on the scoring smoothing coefficient and the model parameters in the application scoring model. If the reference update conditions are not met, the model parameters in the reference adjustment model and the reference scoring model will not be updated.

6. The method according to any one of claims 1-4, characterized in that, The method further includes: When it is detected that the current beam uniformity in the target ion implanter does not meet the target process requirements, a current state vector is determined. The current state vector includes the beam state vector and the device state vector of the target ion implanter at the current acquisition time. The application adjustment model is invoked to determine the current device adjustment vector corresponding to the current state vector. The current device adjustment vector is used to guide the adjustment of at least one device in the target ion implanter. The determination of the current state vector is performed iteratively until the current beam uniformity in the target ion implanter meets the target process requirements; wherein, during the adjustment of at least one device in the target ion implanter, the applied adjustment model supports parameter updates when the model update conditions are met.

7. A beam uniformity adjustment device, characterized in that, The device includes: The acquisition unit is used to acquire historical adjustment data, which includes multiple historical state vectors of the target ion implanter. Each historical state vector includes the beam state vector and device state vector of the target ion implanter at a historical acquisition moment. The processing unit is used to determine the current training state vector from the historical adjustment data, and call the applied adjustment model to determine the device adjustment vector corresponding to the current training state vector. The processing unit is further configured to determine the adjustment model loss value of the application adjustment model based on the device adjustment vector; and update the model parameters in the application adjustment model in the direction of reducing the adjustment model loss value. When the processing unit determines the adjustment model loss value of the applied adjustment model based on the device adjustment vector, it specifically performs the following steps: Based on the device adjustment vector, it determines an updated beam state vector and an updated device state vector; and based on the updated beam state vector and the updated device state vector, it determines a current updated state vector corresponding to the current training state vector, wherein the updated beam state vector is the beam state vector adjusted according to the device adjustment vector; it determines the benefit information and process state information corresponding to the current updated state vector, wherein the benefit information is the difference between the beam uniformity indicator value corresponding to the beam state vector in the current updated state vector and the beam uniformity indicator value corresponding to the beam state vector in the current training state vector; wherein, when the updated beam state... When the beam uniformity indication value corresponding to the vector is less than or equal to a preset beam uniformity threshold, the process status information is an indication that the target process requirement is met; when the beam uniformity indication value corresponding to the updated beam state vector is greater than the preset beam uniformity threshold, the process status information is an indication that the target process requirement is not met. Based on the current training state vector, the current updated state vector, the device adjustment vector, the revenue information, and the process status information, a current training tuple is determined; and based on the current training tuple, at least one training sample is determined; the application scoring model is called respectively to determine the application score value corresponding to each training sample in the at least one training sample; and based on the application score value corresponding to each training sample, the adjustment model loss value of the application adjustment model is calculated. The processing unit is further configured to iteratively execute the process of determining the current training state vector from the historical adjustment data until the target model training conditions are met, so as to complete the model training of the application adjustment model through the historical adjustment data, wherein the application adjustment model supports the use of guiding the beam uniformity adjustment of the target ion implanter.

8. An electronic device, characterized in that, include: processor; as well as Stored program memory, The program includes instructions that, when executed by the processor, cause the processor to perform the method according to any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

Images