JP2020161470A - Ion beam irradiation device and program therefor - Google Patents

Abstract

To allow automatic determination of an appropriate initial value for a basic operating parameter.SOLUTION: An ion beam irradiation device 100 that generates an ion beam that satisfies a processing condition by a plurality of modules M and irradiates an object W to be processed with the ion beam includes a machine learning unit 92 that generates a learning algorithm using at least the processing conditions at the time of new processing and a monitor value indicating the state of at least one module M in the processing immediately before the new processing as explanatory variables, and a basic operation parameter output unit 94 that outputs the initial value of a basic operation parameter that controls the operation of the module M using a learning algorithm.SELECTED DRAWING: Figure 2

Description

The present invention relates to an ion beam irradiator and a program for an ion beam irradiator.

As shown in Patent Document 1, the ion implanter includes a plurality of modules for generating an ion beam such as an ion source and a mass separation magnet, and these modules have various preset basic operating parameters. It is configured to work based on.

Specifically, when the control device that controls each module receives a processing condition called a recipe, first, the initial value of the basic operation parameter is input to each module. Then, the control device sets up each module by adjusting the initial values of the basic operation parameters so that the ion beam according to the recipe is generated while detecting various information about the generated ion beam.

As the initial value of this basic operation parameter, a value stored in advance according to the recipe or a value used in the past processing of the same recipe is input as the initial value.

However, if the state (consumed state or maintenance state) of each module is different between the new processing time and the previous processing time, the recipe is the same, and even if the same value is input as the initial value, the adjustment value is obtained. It may take a long time to set up, and the quality of the ion beam such as the beam current amount, beam angle, and beam current density may change.

JP-A-2007-35370

Therefore, the present invention has been made to solve the above-mentioned problems, and is suitable for the current state of the module, for example, an initial value that shortens the setup time and an initial value that can generate a desired ion beam. Its main task is to be able to appropriately determine the initial values of basic operating parameters.

That is, the ion beam irradiating device according to the present invention is an ion beam irradiating device that generates an ion beam satisfying the processing conditions by a plurality of modules and irradiates the object to be processed with new processing conditions at the time of processing and the new processing conditions. The operation of the module is controlled by using the machine learning unit that generates a learning algorithm using at least one monitor value indicating the state of the module in the process immediately before the previous process as an explanatory variable and the learning algorithm. It is characterized by including a basic operation parameter output unit that outputs initial values of basic operation parameters to be performed.

With the ion beam irradiation device configured in this way, a learning algorithm is generated using a monitor value indicating the state of the module in the process immediately before each process (hereinafter, also referred to as preprocess) as an explanatory variable. The initial value of the basic operation parameter output by using this learning algorithm takes into consideration the state of the module before starting a new process. Therefore, by using this learning algorithm, the initial values of the basic operation parameters that match the current module state, such as the initial values that shorten the setup time and the initial values that can generate the desired ion beam, are appropriate. You will be able to decide.

As a more specific embodiment, in a device configuration in which the initial value is input to the plurality of modules and each module operates based on the adjustment value obtained by adjusting the initial value, a plurality of past values are used. Learning data obtained from the processing, the processing conditions of each processing, at least one of the initial value or the adjustment value of each processing, and the monitor of at least one of the modules in the processing immediately before each processing. A machine learning unit that stores training data including a plurality of sets of data sets in which values and actual values of predetermined objective variables in each process are associated with each other is further provided, and the machine learning unit is a machine that uses the training data. Examples thereof include a configuration in which the learning algorithm is generated by learning.

The objective variables include the setup time until the adjustment value is obtained, an index value indicating whether or not the adjustment value is obtained, the beam current amount of the ion beam, the beam angle of the ion beam, or the ion beam. The beam current density of.

Of the multiple modules, many of the ion source modules have a shorter life than the extraction electrode modules and beamline electromagnetic field modules, and the ion source modules have shorter lifespans than the extraction electrode modules and beamline electromagnetic fields. More frequent maintenance and replacement than system modules.
In view of this point, it is preferable that the training data includes at least the monitor value of the ion source system module.
In this case, machine learning can be performed in consideration of the preprocessing state of the ion source module whose state is likely to change among a plurality of modules, so that the initial value at the time of new processing can be determined more appropriately. be able to.

The plasma generated by the ion source is difficult to model and control, and the dominant parameters for the plasma generation efficiency are the gas flow rate supplied to the plasma chamber and the source that generates a magnetic field in the plasma chamber. The supply current to the magnet can be mentioned.
Therefore, at least one of the gas flow rate supplied to the plasma chamber constituting the ion source or the supply current to the source magnet that generates a magnetic field in the plasma chamber is used as the basic operating parameter. Is preferable.
In this case, the initial values of the gas flow rate and the supply current to the source magnet can be appropriately determined so that the plasma can be efficiently generated.

In a configuration including a control device for setting up the module by selecting an initial value of the basic operation parameter based on the processing conditions and a predetermined setup sequence, inputting the initial value to the module, and adjusting the initial value, the control When the setup of the module by the apparatus is not completed, a recovery unit for acquiring an abnormal signal indicating that is further provided, and when the recovery unit acquires the abnormal signal, it is output from the basic operation parameter output unit. It is preferable that the initial value of the basic operation parameter is input to the module.
With such a configuration, the initial values obtained by the learning algorithm can be used without significantly changing the module setup sequence from the previous ones, further shortening the setup time and improving the setup success rate. Further improvement can be achieved.

Further, as another embodiment, a control for setting up the module by selecting an initial value of the basic operation parameter based on the processing conditions and a predetermined setup sequence, inputting the initial value to the module, and adjusting the initial value. In a configuration including an apparatus, a pre-prediction unit for predicting whether or not the setup of the module is completed when the initial values selected based on the processing conditions and the setup sequence are used is further provided. When it is predicted that the setup of the module will not be completed, the initial value of the basic operation parameter output from the basic operation parameter output unit can be input to the module.

Further, the program for an ion beam irradiator according to the present invention is a program used for an ion beam irradiator that generates an ion beam satisfying a processing condition by a plurality of modules and irradiates an object to be processed, and is used at the time of new processing. Using the machine learning unit that generates a learning algorithm using at least the processing conditions of the above and the monitor value indicating the state of at least one module in the processing immediately before the new processing as explanatory variables, and the learning algorithm. The computer is characterized in that it functions as a basic operation parameter output unit that outputs initial values of basic operation parameters that control the operation of the module.
The program for an ion beam irradiator configured in this way can exert the same action and effect as the above-mentioned ion beam irradiator.

According to the present invention configured in this way, machine learning can be performed in consideration of the state at the time of preprocessing of the module, and the initial value of the basic operation parameter at the time of new processing is appropriately determined according to the objective variable. can do.

The schematic diagram which shows the whole structure of the ion beam irradiation apparatus which concerns on this embodiment. The functional block diagram which shows the function of the control device and the machine learning device of the same embodiment. The flowchart which shows the operation of the control device and learning device of the same embodiment. The figure for demonstrating the content of the learning model of the same embodiment. The functional block diagram which shows the function of the control device and the machine learning device of another embodiment.

<First Embodiment>
The first embodiment of the ion beam irradiation device according to the present invention will be described below with reference to the drawings.

As shown in FIG. 1, the ion beam irradiation device 100 of the present embodiment is an ion implantation device for irradiating the target W with an ion beam IB and implanting ions, and a plurality of modules for generating the ion beam IB. 2 to 7 and a control device 8 for controlling the operation of these modules 2 to 7 are provided. The ion beam irradiation device 100 is not limited to the ion implantation device, and may be, for example, an ion beam etching device or the like.

Specifically, the ion beam irradiation device 100 is provided on the ion source 2 from which the ion beam IB is drawn and on the downstream side of the ion source 2, and has a mass number and a valence from the ion beam IB drawn from the ion source 2. It is provided with at least a mass separation magnet 3 which is a mass separator for selecting and deriving a desired dopant ion specified in 1. In this embodiment, a mass separation magnet 3 is provided on the downstream side of the mass separation magnet 3. An acceleration tube 4 for accelerating or decelerating the ion beam IB derived from 3 and an ion beam IB provided on the downstream side of the acceleration tube 4 and having a specific energy are selected from the ion beam IB derived from the acceleration tube 4. The energy separation magnet 5 which is the derivation energy separator and the ion beam IB which is provided on the downstream side of the energy separation magnet 5 and is derived from the energy separation magnet 5 are magnetically one-dimensionally (in FIG. 1 along the paper surface). The scanning magnet 6 for scanning and the ion beam IB provided on the downstream side of the scanning magnet 6 and being derived from the scanning magnet 6 are bent back so as to be parallel to the reference axis to form the scanning magnet 6. It is provided with a beam parallelizing magnet 7 that collaborates to perform parallel scanning of the ion beam IB. Each of these components is referred to as a module here as described above (hereinafter referred to as module M).
The above-mentioned module M does not necessarily have to be provided. For example, the ion beam irradiation device 100 provides a ribbon beam having a size larger than the size of the target W in a direction orthogonal to the mechanical scanning direction of the target W. When it is configured to irradiate, the scanning magnet 6 is unnecessary, and other modules M such as the acceleration tube 4, the energy separation magnet 5, and the parallelizing magnet 7 may be appropriately selected.

The control device 8 is a computer having a CPU, a memory, a display, an input means, and the like, and by coordinating the CPU and its peripheral devices according to a program stored in the memory, as shown in FIG. 2, a recipe receiving unit It at least exhibits the functions of 81, the basic operation parameter input unit 82 (hereinafter referred to as the basic para-input unit 82), and the basic operation parameter tuning unit 83 (hereinafter referred to as the basic para-tuning unit 83).

The recipe reception unit 81 receives processing conditions (hereinafter referred to as recipes) for the object to be processed.
The recipe is data transmitted from, for example, a host computer (not shown), and is the quality of the ion beam IB such as the ion species of the dopant ions contained in the ion beam IB, the beam energy of the ion beam IB, and the beam current of the ion beam IB. It is data including various information indicating.

The base para input unit 82 inputs the initial values of the basic operation parameters (hereinafter, referred to as base para) for controlling the operation of each module M to each module M. The basic para is a setting item necessary for controlling the operation of the module M, and is preset for each module M. One type of base para may be set for one module M, or a plurality of types of base para may be set.

Here, the module M constituting the ion beam irradiation device 100 controls the ion source system module for generating plasma, the extraction electrode system module for extracting the ion beam IB from the ion source 2, and the extracted ion beam IB. It can be roughly divided into beamline electromagnetic field system modules for.

For example, the ion source system module includes a plasma chamber constituting the ion source 2, a gas supply mechanism (flow control device, etc.) that supplies plasma generation gas to the plasma chamber, a source magnet that generates a magnetic field in the plasma chamber, and plasma. Examples include filaments that emit electrons into the chamber.
Further, examples of the basic paradox set in the ion source system module include a gas flow rate supplied to the plasma chamber, a supply current supplied to the source magnet, an arc current, and the like.

Examples of the extraction electrode system module include a suppression electrode and a ground electrode constituting an extraction electrode system for extracting an ion beam IB from the ion source 2, and an adjustment mechanism for adjusting the position and separation distance of these electrodes.
Further, examples of the basic paradox set in the extraction electrode system module include the distance between the plasma chamber and each electrode along the extraction direction, the position of each electrode, the inclination of each electrode, and the like.

Examples of the beamline electromagnetic field system module include a mass separation magnet 3, an acceleration tube 4, an energy separation magnet 5, a scanning magnet 6, a beam parallelizing magnet 7, and the like.
The basic parameters set in the beamline electromagnetic field system module include, for example, the magnetic flux density of the mass separation magnet 3, the voltage applied to the acceleration tube 4, the magnetic flux density of the energy separation magnet 5, the magnetic flux density of the parallelizing magnet 7, and the like. Can be mentioned.

It should be noted that not all of the basic paras listed above are necessarily required, and they may be appropriately selected according to the configuration of the ion beam irradiation device 100 and the like.

The base para tuning unit 83 adjusts the initial value of the base para input by the base para input unit 82, and specifically, so that the generated ion beam IB satisfies the processing conditions, that is, according to the recipe. Each module M is set up while adjusting the initial values of one or a plurality of types of basic paras as necessary so that the ion beam IB of the above is generated.

More specifically, the basic para tuning unit 83 is configured to adjust the basic para according to a predetermined setup sequence, and the adjustment value of the basic para finally obtained by this setup sequence. Based on the above, each module M will operate. The adjustment value referred to here includes a value obtained by adjusting the initial value and an initial value that has not been adjusted.

The setup sequence includes, for example, a plasma generation step of generating plasma in a plasma chamber, an ion beam IB extraction step using an extraction electrode system, a beam adjustment step of accelerating / decelerating the ion beam IB, orbit adjustment, and an ion beam IB. It includes a plurality of steps such as a beam measuring step for measuring a beam current amount, a beam angle, a beam current density, and the like. As shown in FIG. 2, the basic paratuning unit 83 makes the detection values detected by various detectors X such as a beam detector (for example, a Faraday cup) approach a predetermined target value in each step. Feedback control etc. is performed on the value of the base para.

Here, as shown in FIG. 2, the ion beam irradiation device 100 of the present embodiment has a monitor value storage unit 84 that stores a state parameter value (hereinafter, referred to as a monitor value) indicating the state of at least one module M. I have more.

Here, even if the recipe is the same and the same value is input as the initial value of the basic para, the adjustment value obtained by the basic para tuning unit 83 described above varies depending on the state of the module M at that time. Therefore, the state parameter of the present embodiment includes all or a part of the base para, and the monitor value includes all or a part of the adjustment value of the base para. Specifically, all or part of the adjustment values obtained by the basic paratuning unit 83 described above are stored in the monitor value storage unit 84 as monitor values.

Further, the state parameters other than the basic parameters may include parameters that fluctuate during the processing, that is, parameters that cause a difference before and after one or a plurality of processings. Examples of such a state parameter include filament current and filament voltage. Unlike the basic parameters, these state parameters are not input with initial values, but are variable values that fluctuate due to changes in the filament state over time (for example, a decrease in filament diameter) during processing. is there. As shown in FIG. 2, such fluctuation values are monitored by the monitor unit Z, and these fluctuation values are stored in the monitor value storage unit 84 as monitor values.

Therefore, as shown in FIG. 1, the ion beam irradiation device 100 of the present embodiment is a machine learning device for automatically and appropriately determining the initial value of the base para input by the above-mentioned base para input unit 82. 9 is further provided.

The machine learning device 9 is a computer having a CPU, a memory, a display, an input means, artificial intelligence, and the like, and is separate from the control device 8 described above, and is an ion beam irradiation device stored in the memory. By making the CPU and other peripheral devices cooperate according to the program, as shown in FIG. 2, the learning data storage unit 91, the machine learning unit 92, the algorithm storage unit 93, and the basic operation parameter output unit 94 (hereinafter, the basis). It exerts a function as a para output unit 94).
Hereinafter, the setup of each module M using the machine learning device 9 will be described with reference to the flowchart of FIG. 3, in addition to explaining the functions of the respective parts 91 to 94.

First, the learning data used for machine learning is stored in the learning data storage unit 91 (S1).
The training data is, for example, data obtained from a plurality of past processes, and includes various actual values in the plurality of past processes. Specifically, as shown in FIG. 4, this learning data includes the recipe of each past process, the initial value of the basic para input by the basic para input unit 82 in each process in each process, and the basic para tuning unit in each process. Includes a plurality of sets of data sets associated with the adjustment value of the base parameter obtained by adjusting 83, the monitor value of the state parameter of the module M at the end of each process, and the actual value of the predetermined objective variable in each process. It's a module. As described above, the monitor values here include a part or all of the adjustment values of the basic parameters, and also include fluctuation values monitored by the monitor unit Z. However, the monitor value may include at least one of the adjustment value and the fluctuation value of the base para.

Then, among these data, at least one of the processing conditions of each past processing, the initial value or the adjustment value of the basic parameter in each processing, and at least one in the processing immediately before each processing (hereinafter, also referred to as preprocessing). The monitor value of one module M and the actual value of a predetermined objective variable in each process are associated as a set of data sets. The data set here includes both the initial value and the adjustment value of the base para. Further, the data set here is associated with the degree of vacuum of the beam line in each process as needed, and a plurality of sets of this data set are included in the training data.

The objective variable is a reference for the base para output unit 94, which will be described later, to output the initial value of the base para in a new process, and here is the setup time, that is, the time required to start the process, and at least the base para. The tuning time required from when the tuning unit 83 starts adjusting the initial value of the base parameter to when the adjustment value is obtained is included.

Next, the machine learning unit 92 uses the learning data stored in the learning data storage unit 91 to obtain a recipe received at the time of a new process and a monitor value of at least one module M in the preprocessing of the new process. Is generated as an explanatory variable (S2).
The machine learning unit 92 is a function exerted by the above-mentioned artificial intelligence, and generates the above-mentioned learning algorithm by using appropriately selected machine learning such as supervised learning, unsupervised learning, reinforcement learning, and deep learning. It is configured in.

The machine learning unit 92 stores the generated learning algorithm in the algorithm storage unit 93 set in a predetermined area of the memory. The algorithm storage unit 93 may be provided in an external memory, a cloud server, or the like.

Then, when the recipe receiving unit 81 described above receives a recipe at the time of new processing, the basic para output unit 94 uses the learning algorithm generated by the machine learning unit 92 so that the objective variable satisfies a desired condition. , The initial value of the base parameter at the time of new processing is output (S3).
Specifically, the basic para output unit 94 acquires the recipe at the time of new processing received by the recipe reception unit 81 and the monitor value at the time of preprocessing of at least one module M stored in the monitor value storage unit 84. Then, estimate the setup time for the new process. More specifically, the adjustment of the basic parameters at the time of multiple past processes (for example, when the recipe is the same as the new process, or when the monitor value is the same as the preprocess or falls within a predetermined range). The setup time when the value is acquired and each adjustment value is input as the initial value at the time of new processing is classified (rating) or calculated, and the result is output. It should be noted that the setup time when the initial value of the base para in the past a plurality of times of processing is input as the initial value in the new processing may be classified or calculated, and the result may be output.

The base para output unit 94 here is a monitor value of the state parameter of the ion source system module at the end of the pretreatment, specifically, an adjustment value of the base para at the time of pretreatment of the ion source system module, and a filament state. The filament voltage and filament current, which are parameters, are acquired, and the initial values of the basic parameters are determined using these monitor values. However, the base para output unit 94 acquires the monitor values of the state parameters of the ion source system module other than the filament, the extraction electrode system module, and the beamline electromagnetic field system module, and uses these monitor values to obtain the monitor values of the base para. The initial value may be determined.

From the initial values of the basic paras output by the basic para output unit 94 in this way, the initial values determined to be optimal based on predetermined judgment conditions, such as the one with the shortest setup time, are input to the basic paras. It is input to each module M by the unit 82 (S4). Although this determination is made by the basic para-output unit 94 here, it may be made by a computer (for example, a control device 8) different from the machine learning device 9, or by an operator. good.

After that, as described above, the basic para tuning unit 83 adjusts the basic para (S5), and the setup of each module M is completed.

According to the ion beam irradiation device 100 configured in this way, since the monitor value of the state parameter of the module M in the preprocessing is included as the learning data, machine learning in consideration of the state of the module M in the preprocessing. It can be performed.
As a result, by using the learning model generated by this machine learning, it is possible to determine the initial value of the basic para that minimizes the setup time, for example, at the time of new processing.

Specifically, there are cases where the initial values of the basic paras output by the machine learning device 9 of the present embodiment are input to each module M, and cases where the initial values of the conventional existing basic paras are input to each module M. As a result of comparing the setup time in, it was confirmed that the former is about half the setup time compared to the latter, and the setup time can be shortened.

Here, among the plurality of modules M, the ion source system modules often have a shorter life than the extraction electrode system modules and the beamline electromagnetic field system modules, and the frequency of maintenance and replacement is high.
On the other hand, since the learning data of the present embodiment includes at least the monitor values of the state parameters of the ion source system module, the state at the time of preprocessing of the ion source whose state is likely to change even in the module M can be obtained. Machine learning can be performed in consideration, and the initial value at the time of new processing can be determined more appropriately.

Furthermore, the gas flow rate supplied to the plasma chamber and the supply current to the source magnet are included as the basic parameters of the ion source system module, and these basic parameters dominate the plasma generation efficiency. The initial values of the gas flow rate and the supply current to the source magnet can be appropriately determined so that plasma that is difficult to model can be efficiently generated.

<Second Embodiment>
Next, a second embodiment of the ion beam irradiation device according to the present invention will be described.

Similar to the first embodiment, the ion beam irradiation device 100 of the second embodiment is an operation of inputting and adjusting the initial value of the base para obtained based on the learning algorithm into each module M (hereinafter referred to as AI group para mode). ) And the operation of inputting the initial value of the basic para selected by the controller 8 (beam controller) into each module M and adjusting it (hereinafter referred to as BC basic para mode) without relying on this AI basic para mode. Is different from the first embodiment.

First, since the operation by the AI group paramode is as described in the first embodiment, the BC group paramode will be briefly described below.
As the BC group paramode, for example, a conventional setup sequence of the control device 8 can be mentioned. Specifically, when the recipe reception unit 81 accepts a recipe, the base para input unit 82 sets the initial value of the base para, for example, an initial value stored in advance according to the recipe, or the same recipe as the received recipe. The initial value used in the past processing is selected, and the selected initial value is input to each module M. After that, the base para tuning unit 83 adjusts the initial value of the input base para so that the ion beam IB according to the recipe is generated, and sets up each module M.

Therefore, as shown in FIG. 5, the control device 8 of the present embodiment further includes one or both functions of the recovery unit 85 and the advance prediction unit 86.

When the setup of each module M by the BC group paramode is not completed and the recovery unit 85 terminates abnormally, the recovery unit 85 acquires an abnormal signal indicating that, and changes the setup sequence of each module M from the BC group paramode to the AI group paramode. It switches to.

Specifically, when the recovery unit 85 acquires an abnormal signal, the operation of the base para input unit 82 is switched from the BC group para mode to the AI group para mode, and each module M is made to input the initial value of the base para again. That is, as described in the above-described embodiment, the base para input unit 82 acquires the initial value of the base para determined and output by the base para output unit 94 using the learning algorithm, and sets the initial value in each module M. input.

When the operation of the base para input unit 82 is switched from the BC base para mode to the AI base para mode and an abnormal signal is still detected, the recovery unit 85 may operate the base para input unit 82 again in the AI base para mode. In this case, it is preferable that the base para output unit 94 changes the initial value of the ion source system module whose state is likely to change among a plurality of modules as the initial value of the base para.

The advance prediction unit 86 predicts whether or not the setup of each module M in the BC group paramode is completed based on the recipe received by the recipe reception unit 81 before the start of the operation in the BC group paramode.

More specifically, the advance prediction unit 86 determines whether or not one or more of the following events have occurred, and if one or more of the following events have occurred, the setup by the BC group paramode is performed. Judges that there is a high probability that will not be completed.

-When the same recipe as the accepted recipe has not been accepted over the past predetermined period (for example, one month).
-In the process one or more times before the same recipe as the received recipe, the predetermined actual values (for example, the beam current amount of the ion beam IB, the beam angle of the ion beam IB, the beam current density of the ion beam IB, etc.) are predetermined. When the numerical range of is exceeded.
-When the setup by BC group paramode is not completed in the processing one or more times before the same recipe as the received recipe.
-When the processing of the same recipe as the received recipe one or more times before is performed by the first or a predetermined number of processing after opening to the atmosphere.

Then, when the advance prediction unit 86 predicts that the setup of each module M will be completed in the BC group paramode, the advance prediction unit 86 operates the group para input unit 82 in the BC group paramode.
On the other hand, when the advance prediction unit 86 predicts that the setup of each module M will not be completed by the BC group paramode, the advance prediction unit 86 operates the group para input unit 82 in the AI group paramode.

With such a configuration, for example, BC-based paramode is used for recipes that can complete the setup in a short time based on past achievements, while AI for recipes that do not complete or take a long time to complete. Basic paramode can be used.
As a result, the setup operation in the AI-based paramode can be introduced without significantly changing the setup sequence from the conventional one, and the setup time can be further shortened and the setup success rate can be further improved as compared with the conventional case.

<Other Embodiments>
The present invention is not limited to the above embodiment.

For example, in the above-described embodiment, the objective variable has been described as the setup time, but the index value indicating whether or not the adjustment value has been obtained, in other words, the index value indicating whether or not the setup has been completed, the ion beam IB. It may be the amount of beam current, the beam angle of the ion beam IB, the beam current density of the ion beam IB, or the like.

Further, as the base para output unit 94, for example, the setup time is within a predetermined time, and ions such as the beam current amount of the ion beam IB, the beam angle of the ion beam IB, or the beam current density of the ion beam IB are formed. The initial value of the base para may be output so that the quality of the beam IB satisfies a predetermined condition.

Further, the machine learning device 9 may have a function as an objective variable prediction unit that predicts an objective variable (for example, setup time) in a new process based on a learning algorithm generated using the learning data.

In addition, in the above-described embodiment, the monitor value of the state parameter at the end of the preprocessing is used as the explanatory variable, but the monitor value of the state parameter during the preprocessing may be used.

In addition, the control device 8 may include some or all of the functions included in the machine learning device 9 of the embodiment.

In addition, the present invention is not limited to the above-described embodiment, and it goes without saying that various modifications can be made without departing from the spirit of the present invention.

100 ・ ・ ・ Ion beam irradiation device 8 ・ ・ ・ Control device 81 ・ ・ ・ Recipe reception unit 82 ・ ・ ・ Basic para input unit 83 ・ ・ ・ Basic para tuning unit 9 ・ ・ ・ Machine learning device 91 ・ ・ ・ Learning data Storage unit 92 ・ ・ ・ Machine learning unit 93 ・ ・ ・ Algorithm storage unit 94 ・ ・ ・ Basic para output unit

Claims (8)

An ion beam irradiation device that generates an ion beam that satisfies the processing conditions by a plurality of modules and irradiates the object to be processed.
A machine learning unit that generates a learning algorithm using at least a processing condition at the time of a new process and a monitor value indicating the state of at least one module in the process immediately before the new process as explanatory variables.
An ion beam irradiation device including a basic operation parameter output unit that outputs initial values of basic operation parameters that control the operation of the module by using the learning algorithm. In an apparatus configuration in which the initial values are input to the plurality of modules and each module operates based on the adjustment values obtained by adjusting the initial values.
Learning data obtained from a plurality of past processes, that is, the processing conditions of each processing, at least one of the initial value or the adjustment value of each processing, and at least one of the above in the processing immediately before each processing. Further, a training data storage unit for storing training data including a plurality of data sets in which the monitor value of the module and the actual value of a predetermined objective variable in each process are associated with each other is provided.
The ion beam irradiation device according to claim 1, wherein the machine learning unit generates the learning algorithm by machine learning using the learning data. The objective variable is the setup time until the adjustment value is obtained, an index value indicating whether or not the adjustment value is obtained, the beam current amount of the ion beam, the beam angle of the ion beam, or the ion beam. The ion beam irradiation device according to claim 2, which is a beam current density. The ion beam irradiation device according to claim 2 or 3, wherein the training data includes at least the monitor value of the ion source system module. 4. Claim 4 in which at least one of the gas flow rate supplied to the plasma chamber constituting the ion source and the supply current to the source magnet that generates a magnetic field in the plasma chamber is used as the basic operating parameter. The ion beam irradiation device according to the above. In a configuration including a control device that sets up the module by selecting an initial value of the basic operation parameter based on the processing conditions and a predetermined setup sequence, inputting the initial value to the module, and adjusting the initial value.
Further, a recovery unit for acquiring an abnormal signal indicating that the setup of the module by the control device is not completed is provided.
The item according to any one of claims 1 to 5, wherein the initial value of the basic operation parameter output from the basic operation parameter output unit is input to the module when the recovery unit acquires the abnormal signal. Ion beam irradiation device. In a configuration including a control device that sets up the module by selecting an initial value of the basic operation parameter based on the processing conditions and a predetermined setup sequence, inputting the initial value to the module, and adjusting the initial value.
Further provided with a pre-prediction unit that predicts whether or not the setup of the module will be completed when the initial values selected based on the processing conditions and the setup sequence are used.
Any of claims 1 to 5 in which the initial value of the basic operation parameter output from the basic operation parameter output unit is input to the module when the advance prediction unit predicts that the setup of the module will not be completed. The ion beam irradiation device according to item 1. A program used in an ion beam irradiator that generates an ion beam that satisfies the processing conditions by a plurality of modules and irradiates the object to be processed.
A machine learning unit that generates a learning algorithm using at least a processing condition at the time of a new process and a monitor value indicating the state of at least one module in the process immediately before the new process as explanatory variables.
A program for an ion beam irradiator that uses the learning algorithm to cause a computer to function as a basic operation parameter output unit that outputs initial values of basic operation parameters that control the operation of the module.