Design support method, design support system, and storage medium

Abstract

This invention provides a design support method, design support system, program, and storage medium capable of supporting the design of semiconductor devices. In the design method of one embodiment, a design value set including multiple design values ​​related to the semiconductor device is input to a simulator. In the design method, a characteristic value set corresponding to the input of the design value set is obtained and output from the simulator. The characteristic value set includes multiple characteristic values ​​of the semiconductor device. The multiple characteristic values ​​include a first characteristic value representing on-resistance and a second characteristic value representing withstand voltage. In the design method, an acquisition function based on Bayesian estimation is calculated based on historical data including one or more datasets, said datasets being the design value set and a score calculated by inputting a portion of the characteristic value set to a first function. A portion of the characteristic value set includes the first characteristic value and the second characteristic value. In the design method, a new design value set is generated based on the acquisition function.

Description

[0001] This application is based on Japanese Patent Application 2020-201947 (filed on December 4, 2020), and enjoys priority from that application. This application incorporates the entire contents of that application by reference. Technical Field

[0002] Embodiments of the present invention generally relate to design support methods, design support systems, and storage media. Background Technology

[0003] For semiconductor components, the goal is to develop technologies that can support the design. Summary of the Invention

[0004] Embodiments of the present invention provide a design support method, a design support system, and a storage medium capable of supporting the design of semiconductor devices.

[0005] In the design method of an embodiment of the present invention, a design value set including multiple design values ​​related to a semiconductor element is input to a simulator. In this design method, a characteristic value set is obtained and output from the simulator corresponding to the input of the design value set. The characteristic value set includes multiple characteristic values ​​of the semiconductor element. The multiple characteristic values ​​include a first characteristic value representing on-resistance and a second characteristic value representing withstand voltage. In this design method, an acquisition function based on Bayesian estimation is calculated based on historical data including one or more data sets, the data sets being the design value set and a set of scores calculated by inputting a portion of the characteristic value set to a first function. The portion of the characteristic value set includes the first characteristic value and the second characteristic value. In this design method, a new design value set is generated based on the acquisition function.

[0006] According to the above embodiments, a design support method, a design support system, and a storage medium that can support the design of semiconductor devices can be provided. Attached Figure Description

[0007] Figure 1 It is a block diagram representing the functional structure of the design support system for implementing the method.

[0008] Figure 2 This is a flowchart illustrating the design support method for implementing the method.

[0009] Figure 3 This is a schematic diagram illustrating the hardware structure.

[0010] Figure 4 This is a flowchart illustrating the design support method for implementing the method.

[0011] Figure 5It is a schematic cross-sectional view illustrating the structure of a semiconductor device.

[0012] Figure 6 This is a table that illustrates design projects.

[0013] Figure 7 It is a schematic diagram showing the correspondence between design projects and semiconductor components.

[0014] Figure 8 It is a table that lists the characteristic items.

[0015] Figure 9 This is a schematic cross-sectional view illustrating the structure of another semiconductor element.

[0016] Figure 10 This is a flowchart illustrating the design support method for the first variation of the implementation method.

[0017] Figure 11 This is a diagram illustrating the results of the design support method for the illustrative implementation and the first variation.

[0018] Figure 12 (a)~ Figure 12 (c) is the output example of the design support system of the first variation.

[0019] Figure 13 This is an output example of the design support system for the first variation.

[0020] Figure 14 (a) and Figure 14 (b) is the output example of the design support system of the first variation.

[0021] Figure 15 This is a flowchart illustrating the design support method for the second variation of the implementation method.

[0022] Figure 16 This is a flowchart illustrating the design support method for the third variation of the implementation method.

[0023] Explanation of reference numerals in the attached figures

[0024] 1: Design support system; 10: Extraction unit; 11: Scoring calculation unit; 12: Bayesian estimation unit; 13: Acquisition function calculation unit; 14: Design value calculation unit; 15: Document generation unit; 16: Command generation unit; 17: Input unit; 18: Display unit; 20: Storage unit; 30: Simulator; 90: Processing device; 91: CPU; 92: ROM; 93: RAM; 94: Storage device; 95: Input interface; 95a: Input device; 96: Output interface; 96a: Output device; 97: Communication interface; 97a: Server; 98: System bus; 100: Semiconductor element; 101: First electrode; 102: Second electrode; 102a: Contact part; 111: n + Type semiconductor layer, 112:n - 113: p-type semiconductor layer, 114: n-type semiconductor layer + Type semiconductor layer, 120: insulating layer, 120a: gate insulating layer, 121: field plate electrode, 121a: first part, 121b: second part, 122: gate electrode, 200: semiconductor element, 201: first electrode, 202: second electrode, 211: p + 212: n-type semiconductor layer, 213: n-type semiconductor layer - p-type semiconductor layer, 214: p-type semiconductor layer, 215: n-type semiconductor layer + Type semiconductor layer, 220: gate electrode, 220a: gate insulating layer, DM, DM0~DM3: design support method, ST1, ST2: construction diagram Detailed Implementation

[0025] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In this specification and the drawings, the same reference numerals are used for the same elements as those already described, and detailed descriptions are omitted where appropriate.

[0026] Figure 1 It is a block diagram representing the functional structure of the design support system for implementing the method.

[0027] The design support system 1 in this implementation is used to search for design values ​​for one or more design items related to semiconductor components. The design support system 1 uses simulator 30 to search for design values.

[0028] like Figure 1 As shown, the design support system 1 includes an extraction unit 10, a scoring calculation unit 11, a Bayesian estimation unit 12, an acquisition function calculation unit 13, a design value calculation unit 14, a document generation unit 15, a command generation unit 16, an input unit 17, an output unit 18, and a storage unit 20.

[0029] Simulator 30 accepts input of design value sets related to semiconductor components. The design value set includes multiple design values, each representing a different design item. The semiconductor components targeted by Design Support System 1 include, for example, MOSFETs (Metal-Oxide Semiconductor Field-Effect Transistors), IGBTs (Insulated Gate Bipolar Transistors), diodes, etc. Design Support System 1 is particularly suitable for power semiconductor components used for power control and supply. The multiple design items include the impurity concentration of each semiconductor layer, the thickness of each semiconductor layer, and the dimensions of other elements.

[0030] Simulator 30 outputs a set of characteristic values ​​in response to the input of a set of design values. The set of characteristic values ​​includes multiple characteristic values, each representing a different characteristic item. These multiple characteristic items include on-resistance, withstand voltage, and switching charge. Using simulator 30, it is possible to confirm what kind of set of characteristic values ​​will be obtained when the input set of design values ​​is actually applied to a semiconductor device.

[0031] When outputting a set of characteristic values ​​from the simulator 30, the extraction unit 10 extracts a portion of the characteristic values ​​from the set. The extraction unit 10 then sends the extracted portion of the characteristic values ​​to the scoring calculation unit 11.

[0032] The scoring calculation unit 11 accesses the storage unit 20 to obtain the first function. The scoring calculation unit 11 inputs the aforementioned characteristic values ​​into the first function and calculates the score. The first function is a function used to calculate the score based on the characteristic values ​​and is preset by the user. The scoring calculation unit 11 associates the calculated score with a characteristic value group containing the characteristic values ​​input to the first function and a design value group that forms the basis of that characteristic value group, and saves it in the storage unit 20.

[0033] Storage unit 20 stores historical data. The historical data includes one or more datasets. Each dataset includes a combination of design value groups and scores. When storage unit 20 stores a new dataset, Bayesian estimation unit 12 accesses storage unit 20 to retrieve the historical data obtained up to that point. Bayesian estimation unit 12 estimates a surrogate model for the scores based on the historical data and sends it to acquisition function calculation unit 13. Acquisition function calculation unit 13 calculates the acquisition function based on the surrogate model for the scores. Acquisition function calculation unit 13 sends the acquisition function to design value calculation unit 14.

[0034] The design value calculation unit 14 calculates new design values ​​based on the acquisition function. The design value calculation unit 14 saves the calculated design values ​​to the storage unit 20 and sends the calculated design values ​​to the file generation unit 15.

[0035] The file generation unit 15 generates a file for input to the simulator 30. The file includes data for a design value group. For example, the design value calculation unit 14 may calculate the same number of new design values ​​as the design value group contains. In this case, the design value calculation unit 14 generates a new design value group. Alternatively, the design value calculation unit 14 may calculate fewer new design values ​​than the design value group contains. If the number of new design values ​​to be calculated is less than the design value group contains, the file generation unit 15 appropriately adds the insufficient design values. In this case, the file generation unit 15 generates a new design value group. The file generation unit 15 inputs the design value group to the simulator 30.

[0036] The command generation unit 16 generates execution commands for causing the simulator 30 to perform the simulation. The command generation unit 16 sends the execution commands to the simulator 30 at a predetermined time.

[0037] Simulator 30 performs a simulation using the set of design values ​​input from file generation unit 15. Simulator 30 outputs the new set of characteristic values ​​obtained from the simulation results to extraction unit 10.

[0038] The input unit 17 is used for user input of data. The user uses the input unit 17 to save the data required for the processing of the design support system 1 to the storage unit 20.

[0039] Output unit 18 outputs data to the user representing at least one selected from a group including design values, characteristic values, and scores. For example, output unit 18 displays the relationship between the number of repetitions of a processing set and the characteristic values, where the processing set includes inputting a design value group, acquiring a characteristic value group, calculating a score, and generating a new design value group. Hereinafter, the processing set will also be referred to as an "experiment." Output unit 18 may also display a structural diagram of a semiconductor device reflecting the design value group after repeated execution of the experiment.

[0040] Figure 2 This is a flowchart illustrating the design support method for implementing the method.

[0041] exist Figure 2 In the design support method DM shown, the document generation unit 15 inputs a set of design values ​​into the simulator 30 (step S1). The extraction unit 10 acquires a set of characteristic values ​​(step S2). The scoring calculation unit 11 inputs a portion of the set of characteristic values ​​into the first function and calculates the score. The Bayesian estimation unit 12 estimates a surrogate model of the Bayesian-estimated score based on historical data, where the historical data includes a dataset of one or more sets of design values ​​and scores. The acquisition function calculation unit 13 calculates an acquisition function based on the surrogate model of the score (step S3). The design value calculation unit 14 generates a new set of design values ​​based on the acquisition function (step S4).

[0042] By repeatedly performing experiments including steps S1 to S4, the design values ​​of semiconductor components are searched.

[0043] Figure 3 This is a schematic diagram illustrating the hardware structure.

[0044] The design support system 1 for the implementation method can utilize Figure 3 The hardware structure shown is implemented. Figure 3 The processing device 90 shown includes a CPU 91, a ROM 92, a RAM 93, a storage device 94, an input interface 95, an output interface 96, and a communication interface 97.

[0045] ROM 92 stores programs used to control the computer's operations. ROM 92 stores the programs necessary for the computer to perform the aforementioned processes. RAM 93 functions as a storage area for the programs stored in ROM 92.

[0046] CPU 91 includes processing circuitry. CPU 91 uses RAM 93 as its working memory and executes programs stored in at least one of ROM 92 or storage device 94. During program execution, CPU 91 controls various structures via system bus 98 to perform various processes.

[0047] Storage device 94 stores the data required for executing the program and the data obtained by executing the program.

[0048] The input interface (I / F) 95 connects the processing device 90 and the input device 95a. The input I / F 95 is, for example, a serial bus interface such as USB. The CPU 91 can read various data from the input device 95a via the input I / F 95.

[0049] Output interface (I / F) 96 connects processing device 90 and output device 96a. Output I / F 96 is, for example, a video output interface such as DVI (Digital Visual Interface) or HDMI (High-Definition Multimedia Interface: registered trademark). CPU 91 sends data to output device 96a via output I / F 96. Output device 96a outputs data.

[0050] The communication interface (I / F) 97 connects the server 97a external to the processing device 90 and the processing device 90. The communication I / F 97 is, for example, a network card such as a LAN card. The CPU 91 can read various data from the server 97a via the communication I / F 97.

[0051] Storage device 94 includes one or more selected from HDD (Hard Disk Drive) and SSD (Solid State Drive). Input device 95a includes one or more selected from mouse, keyboard, microphone (voice input), and touchpad. Output device 96a includes one or more selected from monitor, printer, speaker, and projector. A device that combines the functions of both input device 95a and output device 96a, such as a touchpad, may also be used.

[0052] For example, the processing device 90 as Figure 1 The extraction unit 10, scoring calculation unit 11, Bayesian estimation unit 12, acquisition function calculation unit 13, design value calculation unit 14, document generation unit 15, and command generation unit 16 shown in the diagram function as storage unit 20. The storage device 94 functions as storage unit 20. The input device 95a functions as input unit 17. The output device 96a functions as output unit 18.

[0053] As simulator 30, it can use TCAD (Technology Computer-Aided Design) from Synopsys or TCAD from SILVACO. For example, with Synopsys TCAD, it can combine Sentaurus Process software and Sentaurus Device software. With SILVACO TCAD, it can combine Athena software and Atlas software.

[0054] Explain the advantages of the implementation method.

[0055] For example, the design values ​​of semiconductor components are determined by humans through repeated trial and error using simulators. However, semiconductor components involve many design projects, placing a significant burden on humans. Moreover, the characteristics of semiconductor components, especially on-resistance and withstand voltage, vary depending on the influence of various design projects. Therefore, evaluating the characteristic values ​​for the input design values ​​and determining the next design value requires specialized knowledge and experience. Furthermore, the evaluation of the characteristic values ​​for the design values ​​can also be biased depending on the individual.

[0056] According to the implementation method, the process of inputting design value sets into the simulator 30, obtaining characteristic value sets from the simulator 30, evaluating, and generating new design value sets is automatically and repeatedly performed. Therefore, it reduces the human burden. No specialized knowledge or experience is required, and biases in the evaluation of characteristic values ​​can be suppressed.

[0057] According to the implementation method, the new set of design values ​​is generated using an acquisition function based on Bayesian estimation. When applying Bayesian estimation in the design of semiconductor components that involve many design values, the initial search requires significantly more time compared to response surface methodology or human-based design. However, the inventors have found that Bayesian estimation is particularly well-suited for semiconductor component design. Ultimately, Bayesian estimation can set the design values ​​of semiconductor components to more desirable values ​​in a shorter time.

[0058] The design support system 1 for the implementation method will be described below with reference to specific examples.

[0059] Figure 4 This is a flowchart illustrating the design support method for implementing the method.

[0060] exist Figure 4 In the design support method DM0 shown, the user sets the design items to be searched and the search range for each design item (step S10). Initial sampling is performed (step S11). In the initial sampling, the design value calculation unit 14 randomly sets a set of settings, and the file generation unit 15 inputs the set of settings to the simulator 30. The extraction unit 10 extracts a portion of the characteristic values ​​from the characteristic value set output by the simulator 30. The scoring calculation unit 11 calculates a score based on the extracted portion of the characteristic values. In the initial sampling, experiments including inputting design value sets, acquiring characteristic value sets, calculating scores, and generating new design value sets are repeatedly performed. For example, the initial sampling is repeated 10 to 30 times. By repeatedly performing the initial sampling, the data set of design value sets and scores is repeatedly saved to the storage unit 20.

[0061] Bayesian estimation unit 12 estimates a surrogate model for the score based on multiple data sets stored in storage unit 20 (step S12). Acquisition function calculation unit 13 calculates an acquisition function using the surrogate model for the score (step S13). Design value calculation unit 14 calculates multiple design values ​​based on the acquisition function and generates a new set of design values ​​(step S14). File generation unit 15 generates a file including the set of design values ​​and inputs it into simulator 30 (step S15). Extraction unit 10 acquires the set of characteristic values ​​output from simulator 30 (step S16) and extracts a portion of the characteristic values ​​(step S17). Scoring calculation unit 11 inputs the extracted portion of the characteristic values ​​into the first function and calculates the score (step S18). Output unit 18 displays the results of the experiment (step S19). Bayesian estimation unit 12 determines whether the termination condition is met (step S20). The experiment in steps S12 to S19 is repeated until the termination condition is met. As an example of the termination condition, steps S12 to S19 are set to be repeated a predetermined number of times, or the score reaches a preset target value.

[0062] The experiment is repeated, including generating the design value group in step S14, inputting the design value group in step S15, obtaining the characteristic value group in step S16, and calculating the score in step S18. Through repeated experiments, the optimal design values ​​related to the semiconductor device are searched.

[0063] Figure 5 It is a schematic cross-sectional view illustrating the structure of a semiconductor device.

[0064] Using the design support system 1 of the implementation method, search and Figure 5 The design values ​​related to the construction of the semiconductor element 100 are shown. Semiconductor element 100 is a MOSFET.

[0065] Semiconductor element 100 includes a first electrode 101, a second electrode 102, and an n-type electrode. + Type semiconductor layer 111, n - p-type semiconductor layer 112, p-type semiconductor layer 113, n + The semiconductor layer 114, insulating layer 120, field plate electrode (hereinafter referred to as FP electrode) 121, and gate electrode 122 are included. The p-type and n-type of each semiconductor layer can also be reversed.

[0066] The first electrode 101 and the second electrode 102 are separated from each other. The direction from the first electrode 101 towards the second electrode 102 is defined as the Z direction. + A semiconductor layer 111 is disposed between the first electrode 101 and the second electrode 102. + The semiconductor layer 111 is electrically connected to the first electrode 101. - Type semiconductor layer 112 is disposed on n + Between the semiconductor layer 111 and the second electrode 102. - The n-type impurity concentration in the n-type semiconductor layer 112 is higher than that in the n-type semiconductor layer 112. + The n-type impurity concentration in the n-type semiconductor layer 111 is low. The p-type semiconductor layer 113 is disposed on the n-type semiconductor layer. - Between the semiconductor layer 112 and the second electrode 102. + The p-type semiconductor layer 114 is disposed between the p-type semiconductor layer 113 and the second electrode 102.

[0067] Insulating layer 120 is provided at n in the Z direction + Between the semiconductor layer 111 and the second electrode 102. The insulating layer 120 is perpendicular to the Z direction and n in the X direction. - A portion of the p-type semiconductor layer 112, the p-type semiconductor layer 113, and the n-type semiconductor layer 112 + A portion of the semiconductor layer 114 is arranged side-by-side. A portion of the insulating layer 120 is provided around the FP electrode 121. From n -A portion of the p-type semiconductor layer 112 faces the FP electrode 121 along the X direction. A gate electrode 122 is disposed in the Z direction between the FP electrode 121 and the second electrode 102. Another portion of an insulating layer 120 is disposed around the gate electrode 122. From the gate electrode 122, a portion of the insulating layer 120 extends towards the p-type semiconductor layer 113 and the n-type semiconductor layer 102. + The orientation of the p-type semiconductor layer 114 is along the X direction. The second electrode 102 is connected to the p-type semiconductor layer 113, n... + The semiconductor layer 114 and the FP electrode 121 are electrically connected, and electrically separated from the gate electrode 122.

[0068] In semiconductor device 100, FP electrode 121 includes a first portion 121a and a second portion 121b. The second portion 121b is disposed between the first portion 121a and the gate electrode 122. The width (length in the X direction) of the second portion 121b is larger than the width of the first portion 121a. The gate electrode 122 is separated from the FP electrode 121 and electrically isolated from it. The second electrode 102 includes a contact portion 102a. The contact portion 102a protrudes toward the p-type semiconductor layer 113.

[0069] With a positive voltage applied to the first electrode 101 relative to the second electrode 102, a voltage above a threshold is applied to the gate electrode 122. This forms a channel (inversion layer) in the p-type semiconductor layer 113, and the semiconductor element 100 becomes conductive. Electrons flow from the second electrode 102 to the first electrode 101 via the channel. If the voltage applied to the gate electrode 122 is lower than the threshold, the channel in the p-type semiconductor layer 113 disappears, and the semiconductor element 100 becomes cut off.

[0070] If semiconductor element 100 is switched to the off state, the positive voltage applied to the first electrode 101 relative to the second electrode 102 increases. At this time, the depletion layer separates from the insulating layer 120 and n... - The interface of the semiconductor layer 112 is oriented towards n. - The depletion layer 112 is extended. By extending this depletion layer, the breakdown voltage of the semiconductor device 100 can be improved. Alternatively, it is possible to increase the n-type voltage while maintaining the breakdown voltage of the semiconductor device 100. - The concentration of n-type impurities in the semiconductor layer 112 is reduced, thereby lowering the on-resistance of the semiconductor element 100.

[0071] Figure 6 This is a table that illustrates design projects. Figure 7 This is a schematic diagram showing the correspondence between design projects and semiconductor components.

[0072] Figure 6 The cell spacing in the table is related to the spacing CP in the X direction of the gate electrode 122. Figure 7(As shown) Corresponds to the thickness of the gate oxide film in the X direction, T_G, of the gate insulating layer 120a. The source depth corresponds to n + The depth D_S in the Z direction of the p-type semiconductor layer 114 corresponds to the depth of the base in the n-type semiconductor layer 113. + The boundary between the n-type semiconductor layer 114 and the p-type semiconductor layer 113 and n-type semiconductor layer 114 - The depth D_B in the Z direction of the boundary of the semiconductor layer 112 corresponds to the gate length L_G in the Z direction. The gate depth corresponds to the depth from n + The depth D_G in the Z direction from the upper surface of the semiconductor layer 114 to the lower end of the gate electrode 122 corresponds to the source concentration. + The concentration of n-type impurities C_S in the n-type semiconductor layer 114 corresponds to this concentration. The substrate thickness is related to the n-type impurity concentration. + The thickness T_Sub in the Z direction of the semiconductor layer 111 corresponds to this. The substrate concentration is related to n. + The n-type impurity concentration C_Sub in the n-type semiconductor layer 111 corresponds to the trench bottom curvature R_TB at the lower end of the insulating layer 120. The contact depth corresponds to the distance from the n-type impurity concentration in the semiconductor layer 111. + The depth D_TC in the Z direction from the upper surface of the semiconductor layer 114 to the lower end of the contact portion 102a corresponds to the contact half-width, which corresponds to half the width W_TC of the contact portion 102a. The drift layer thickness corresponds to n - The thickness T_D in the Z direction of the semiconductor layer 112 corresponds to this. The drift layer concentration is related to n. - The n-type impurity concentration C_D in the p-type semiconductor layer 112 corresponds to the base concentration C_B in the p-type semiconductor layer 113. The trench tilt angle corresponds to the angle Taper between the side of the insulating layer 120 and the X direction. The trench half-width corresponds to half the width W_T of the insulating layer 120. The gate / FP distance and the Z-direction distance between the FP electrode 121 and the gate electrode 122 correspond to D_GFP. The FP1 half-width corresponds to half the width W_FP1 of the first part 121a. The FP1 length corresponds to the Z-direction length L_FP1 of the first part 121a. The FP2 half-width corresponds to half the width W_FP2 of the second part 121b. The FP2 length corresponds to the Z-direction length L_FP2 of the second part 121b. The trench bottom FP thickness corresponds to the Z-direction thickness T_FP of the lower end of the insulating layer 120.

[0073] Figure 8 It is a table that lists the characteristic items.

[0074] exist Figure 8In the table, the on-resistance RonA is the on-resistance per unit area when a voltage greater than the threshold is applied to the gate electrode 122 and the semiconductor element 100 is in the on-state. The withstand voltage Vdss is the withstand voltage of the semiconductor element 100 when the voltage of the gate electrode 122 relative to the second electrode 102 is set to 0V. The withstand voltage Vdsx is the withstand voltage of the semiconductor element 100 when the voltage of the gate electrode 122 relative to the second electrode 102 is set to -20V. The switching charge Qsw is the total amount of charge stored in the gate electrode 122 when the gate voltage is above the threshold Vth and below the mirror voltage in the switching state of the semiconductor element 100. The gate stored charge Qg is the amount of charge stored in the gate electrode 122 when the semiconductor element 100 is in the on-state. The gate-source stored charge Qgs is a portion of the amount of charge stored between the gate electrode 122 and the source electrode 102 when the semiconductor element 100 is in the on-state. The gate-drain stored charge Qgd is a portion of the charge stored between the gate electrode 122 and the drain electrode 101 when the semiconductor element 100 is in the on state. The output charge Qoss is a portion of the charge stored at the drain electrode when the semiconductor element 100 is in the off state. The threshold voltage Vth is the applied voltage to the gate electrode 122 required to form a channel (inversion layer) in the p-type semiconductor layer 113. The channel length L_Ch is the distance from the p-type semiconductor layer 113 to the n-type semiconductor layer 101. + The boundary between the n-type semiconductor layer 114 and the p-type semiconductor layer 113 and n-type semiconductor layer 114 - The length along the Z direction of the boundary of the semiconductor layer 112.

[0075] exist Figure 6 In the example, there are 23 design projects. A design support method that can apply implementation methods to these 23 design projects is provided. Figure 8 In the example, there are 12 characteristic items. A subset of these characteristic items is extracted. The extracted subset of characteristic items is then used to search for design items that can yield better scores. For example, this subset of characteristic items includes on-resistance and withstand voltage.

[0076] The first function outputs a score based on the input characteristic values ​​of the aforementioned characteristic items. As an example, the first function f(x) is represented by the following Equation 1. The inputs to the first function f(x) are the characteristic values ​​of the on-resistance RonA (the first characteristic value), the characteristic value of the withstand voltage Vdss (an example of the second characteristic value), and the characteristic value of the withstand voltage Vdsx (another example of the second characteristic value).

[0077] (Equation 1)

[0078] f(x)=(1 / 30)×RonA+10×(ReLU(110-Vdss)+ReLU(90-Vdsx))

[0079] In Equation 1, "1 / 30" and "10" are values ​​appropriately set by the user. ReLU is the ramp function. "110" is the target value of Vdss. "90" is the target value of Vdsx. By inputting RonA, Vdss, and Vdsx output from simulator 30 into Equation 1, scores for these characteristic values ​​are obtained.

[0080] For example, using equation (1) as the first function, for Figure 6 The design projects represented all underwent 1000 trials. This search can find design values ​​that exhibit characteristics equivalent to those found by skilled personnel. By performing even more trials, design values ​​that exhibit characteristics superior to those found by skilled personnel can be found.

[0081] As mentioned above, the characteristics of semiconductor devices vary depending on the influence of various design projects. For example, among the characteristics of semiconductor devices, on-resistance and breakdown voltage are generally of paramount importance. The concentration and thickness of each semiconductor layer affect on-resistance. In particular, n - The impurity concentration (drift layer concentration) in the semiconductor layer 112 affects the on-resistance. - The higher the impurity concentration in the n-type semiconductor layer 112, the lower the on-resistance. On the other hand, n - The lower the impurity concentration in the semiconductor layer 112, the higher the breakdown voltage. In part of the design project, on-resistance and breakdown voltage are traded off.

[0082] When aiming to improve performance characteristics, there can be situations where design elements influence each other. For example, when the unit spacing becomes narrower, the n-cell spacing between insulating layers 120 increases. - The semiconductor layer 112 is easily depleted. Therefore, it is possible to improve the withstand voltage. Alternatively, by increasing the n-type dielectric layer between the insulating layer 120 and the dielectric layer 120... - The semiconductor layer 112 is easily depleted, which correspondingly increases n. - The impurity concentration in the semiconductor layer 112 can also reduce the on-resistance while maintaining the withstand voltage. That is, the design value of the cell pitch can affect n... - The impurity concentration in the semiconductor layer 112.

[0083] Properly setting multiple design parameters that affect both on-resistance and withstand voltage while focusing on the on-resistance and withstand voltage obtained from simulator 30 requires significant effort, even for skilled personnel. According to the implementation method, optimal design values ​​for both on-resistance and withstand voltage can be found without requiring detailed research.

[0084] In this example, for Figure 6 The multiple design items shown can be searched for their respective design values. You can also search... Figure 6The diagram shows a subset of design values ​​for multiple design items, while setting other design items to fixed values. - The impurity concentration in the n-type semiconductor layer 112 has a significant impact on the on-resistance and breakdown voltage. Therefore, n-type impurity concentration is preferred. - The impurity concentration in the semiconductor layer 112 is the target of the search.

[0085] Figure 9 This is a schematic cross-sectional view illustrating the structure of another semiconductor element.

[0086] Using the design support system 1 of the implementation method, it is also possible to search for and... Figure 9 The design values ​​related to the construction of the semiconductor element 200 are shown. Semiconductor element 200 is an IGBT.

[0087] Semiconductor element 200 includes a first electrode 201, a second electrode 202, and a p-type electrode. + n-type semiconductor layer 211, n-type semiconductor layer 212, n - p-type semiconductor layer 213, p-type semiconductor layer 214, n + The semiconductor layer 215 and the gate electrode 220 are also present. The p-type and n-type of each semiconductor layer can also be reversed.

[0088] The first electrode 201 and the second electrode 202 are separated from each other. The direction from the first electrode 201 towards the second electrode 202 is defined as the Z direction. + A semiconductor layer 211 is disposed between the first electrode 201 and the second electrode 202. + The n-type semiconductor layer 211 is electrically connected to the first electrode 201. The n-type semiconductor layer 212 is disposed on the p-type semiconductor layer 211. + Between the semiconductor layer 211 and the second electrode 202. - The n-type semiconductor layer 213 is disposed between the n-type semiconductor layer 212 and the second electrode 202. - The n-type impurity concentration in the n-type semiconductor layer 213 is lower than that in the n-type semiconductor layer 212. The p-type semiconductor layer 214 is disposed on the n-type semiconductor layer. - Between the semiconductor layer 213 and the second electrode 202. + The p-type semiconductor layer 215 is disposed between the p-type semiconductor layer 214 and the second electrode 202.

[0089] The gate electrode 220 is disposed in the Z direction between the n-type semiconductor layer 212 and the second electrode 202. From n... - A portion of the p-type semiconductor layer 213, the p-type semiconductor layer 214, and the n-type semiconductor layer 213 + A portion of the type semiconductor layer 215 faces the gate electrode 220 along the X direction. In n -Between the p-type semiconductor layer 213 and the gate electrode 220, between the p-type semiconductor layer 214 and the gate electrode 220, and between the n-type semiconductor layer 213 and the gate electrode 220. + A gate insulating layer 220a is provided between the p-type semiconductor layer 215 and the gate electrode 220. The second electrode 202 is connected to the p-type semiconductor layer 214 and the n-type semiconductor layer 220. + The semiconductor layer 215 is electrically connected and electrically separated from the gate electrode 220.

[0090] With a positive voltage applied to the first electrode 201 relative to the second electrode 202, a voltage exceeding a threshold value is applied to the gate electrode 220. This forms a channel (inversion layer) in the p-type semiconductor layer 214. Electrons travel from the second electrode 202 to the n-type semiconductor layer via the channel. - The semiconductor layer 213 flows. Holes flow via p + The semiconductor layer 211 extends from the first electrode 201 to the n-type electrode. - The semiconductor layer 213 flows. In n... - Conductivity modulation is generated in semiconductor layer 213, n - The resistance of the p-type semiconductor layer 213 decreases. As a result, the semiconductor element 200 becomes conductive. If the voltage applied to the gate electrode 220 is lower than the threshold voltage, the channel in the p-type semiconductor layer 214 disappears, and the semiconductor element 200 becomes cut off.

[0091] Using the design support system 1 of the implementation method, it is possible to perform the same operations as the semiconductor device 100. Figure 9 The semiconductor device 200 shown searches for the impurity concentration, thickness, and dimensions of each semiconductor layer. For example, it can search for n... - n-type impurity concentration (drift layer concentration) in semiconductor layer 213 - The thickness of the semiconductor layer 213 in the Z direction (drift layer thickness), the spacing of the gate electrode 220 in the X direction (cell spacing), the thickness of the gate insulating layer 220a in the X direction (gate oxide thickness), n + The depth (source depth) in the Z direction of the p-type semiconductor layer 215, the length (channel length) in the Z direction of the p-type semiconductor layer 214, the length (gate length) in the Z direction of the gate electrode 220, and the distance from n + The depth (gate electrode depth) in the Z direction from the upper surface of the semiconductor layer 215 to the lower end of the gate electrode 220, n + The n-type impurity concentration (source concentration) in the n-type semiconductor layer 215, the curvature of the lower end of the gate insulating layer 220a (trench bottom curvature), the angle between the side of the gate insulating layer 220a and the X direction (trench tilt angle), and half the width of the gate insulating layer 220a (trench half width), etc.

[0092] (First variation)

[0093] Figure 10 This is a flowchart illustrating the design support method for the first variation of the implementation method.

[0094] exist Figure 4 In the design support method DM0 shown, the design values ​​for all design items can be changed in a single trial. Figure 10 In the design support method DM1 of the first variant shown, only a portion of the design values ​​are changed in one test.

[0095] exist Figure 10 In the design support method DM1 shown, and Figure 4 The design support method DM0 shown similarly performs steps S10 and S11. The design value calculation unit 14 extracts a portion of the design items from multiple design items (step S31). Initial sampling is performed on the extracted design items (step S32). In the initial sampling, the design value calculation unit 14 randomly sets the design values ​​for the extracted design items. For the remaining unextracted design items, the design values ​​that yielded the best score in the initial sampling of step S11 are set. The design value calculation unit 14 generates a design value set. The file generation unit 15 inputs the design value set to the simulator 30. The extraction unit 10 extracts a portion of the characteristic values ​​from the characteristic value set output by the simulator 30. The scoring calculation unit 11 calculates a score based on the extracted portion of the characteristic values. In the initial sampling of step S32, experiments including inputting the design value set, acquiring the characteristic value set, calculating the score, and generating a new design value set are repeatedly performed. For example, the initial sampling of step S32 is repeated 3 to 10 times. The data set of design value sets and scores obtained through repeated initial sampling is stored in the storage unit 20.

[0096] After step S32, with Figure 4 Similarly, steps S12 to S19 are executed in the flowchart shown. The Bayesian estimation unit 12 determines whether the number of trials in steps S12 to S18 has reached the predetermined number (step S33). If the number of trials has reached the predetermined number, the design value calculation unit 14 determines whether the termination condition is met (step S34). Steps S31, S32, S12 to S19, and S33 are repeated until the termination condition is met.

[0097] If the termination condition is not met, step S31 is executed again to extract a portion of multiple design items again. At least a portion of the extracted design items may be different from or the same as at least a portion of the design items extracted in the previous step S31.

[0098] As an example of a termination condition, steps S31, S32, S12-S19, and S33 are set to be repeated a predetermined number of times, or the score is set to a predetermined target value. For example, if the predetermined number of repetitions in step S34 is 1000, the predetermined number of repetitions in step S33 is set to be within the range of 10 to 30.

[0099] For example, in design support method DM1, a first processing set for a subset of design values ​​and a second processing set for another subset of design values ​​are executed. In the first processing set, the subset of design values ​​is calculated based on the acquisition function. The remaining design values ​​are fixed to the values ​​that yield the best score. A new set of design values ​​is generated, including the calculated subset of design values ​​and the fixed remaining design values. Multiple first processing sets are executed to search for the subset of design values. In the second processing set, the other subset of design values ​​is calculated based on the acquisition function. The remaining design values ​​are fixed to the values ​​that yield the best score. A new set of design values ​​is generated, including the calculated other subset of design values ​​and the fixed remaining design values. Multiple second processing sets are executed to search for the other subset of design values.

[0100] According to the design support method DM1 of the first variation, since only a portion of the design items are changed in a single experiment, it is possible to obtain a locally optimal solution. For example, the characteristics of the design value obtained by design support method DM1 may be worse than those obtained by performing a large number of experiments with design support method DM0. However, the inventors have found that, according to design support method DM1, better characteristic values ​​are more easily obtained than those obtained by design support method DM0, within the range of the number of experiments that can actually be performed. In particular, it is preferable to extract only one design item in step S31 and perform experiments on one design value sequentially. As a result, better characteristic values ​​can be obtained in a shorter time within a limited number of experiments.

[0101] Figure 11 This is a diagram illustrating the results of the design support method for the illustrative implementation and the first variation.

[0102] Figure 11 This indicates the result when searching for multiple design items using the design support method of the implementation method and the first variation. Regarding the design support method of the first variation, the number of design items extracted in step S31 is set to "1". Figure 11 In the diagram, the horizontal axis represents the number of trials. The vertical axis represents the minimum score (best score) output from the first function in the previous trials.

[0103] Solid lines indicate Figure 4 The results of the design support method DM0 are shown. The dashed lines indicate... Figure 10The results for Design Support Method DM1 are shown. The function f(x) in the graph is the first-order function. With fewer trials, Design Support Method DM0 obtains a smaller score faster than Design Support Method DM1. However, if the number of trials exceeds 200, Design Support Method DM1 obtains a smaller score than Design Support Method DM0.

[0104] exist Figure 11 In the diagram, the dashed line represents the score obtained based on the design values ​​obtained by skilled personnel over a long period of time. This score is set to "1". If the number of trials reaches 400, the score of Design Support Method DM1 is essentially "1". In Design Support Method DM1, it is also possible to lower the score below "1" by further increasing the number of trials.

[0105] Output unit 18 can also display Figure 11 The chart shown illustrates the relationship between the number of trials and the score. Users can easily understand how to improve their scores based on the number of trials.

[0106] Figure 12 (a)~ Figure 12 (c) Figure 13 , Figure 14 (a) and Figure 14 (b) is the output example of the design support system of the first variation.

[0107] Design support system 1 pair Figure 13 The design support method of the first variation is executed on the multiple design projects shown. The number of design projects extracted in step S31 is set to "1". Figure 12 (a)~ Figure 12 (c) indicates the chart displayed using the output section 18.

[0108] exist Figure 12 (a)~ Figure 12 In (c), the horizontal axis represents the number of trials. Figure 12 The vertical axis of graph (a) represents the minimum score (best score) when each number of trials shown on the horizontal axis is performed. Figure 12 The vertical axis of chart (b) represents Figure 13 The table shown contains the identification numbers of the extracted design projects from multiple design projects. Figure 12 The vertical axis of chart (c) represents the design values ​​of the extracted design items. Multiple design values ​​are standardized so that they can be displayed with the same number of decimal places.

[0109] By display Figure 12 As shown in chart (a), users can easily understand how to improve scores based on the number of trials. By removing... Figure 12 In addition to the chart in (a), it also shows from Figure 12The chart shown in (b) and Figure 12 By selecting one or both of the charts shown in (c), users can easily determine the conditions under which experiments are conducted for each number of trials.

[0110] It can also be removed Figure 12 In addition to the chart in (b), it also shows Figure 13 This table allows users to easily identify which design project is being tested at each iteration.

[0111] Output unit 18 can also display Figure 14 (a) or Figure 14 A structural diagram of a semiconductor device as shown in (b). Output section 18 displays... Figure 12 The diagram in (a) is shown, along with the construction diagram. The construction shown reflects the design values ​​that yielded the best score prior to this. For example, Figure 14 (a) is a construction diagram ST1 of a semiconductor device when the number of tests is 200. Figure 14 (b) is the construction diagram ST2 of the semiconductor device after 600 tests. By displaying the construction diagram, users can easily grasp the construction that yields the best score.

[0112] exist Figure 12 (a)~ Figure 12 In the example shown in (c), the number of tests is the same for each design item. The number of tests for each design item can also be different. It is preferable to set the number of tests for important design items to be greater than the number of tests for other design items. For example, the number of tests for drift layer concentration is set to be greater than the number of tests for other design items. This increases the likelihood of obtaining better characteristic values ​​with fewer tests.

[0113] During repeated trials, the search range for design values ​​can be changed based on the obtained scores. For example, a baseline value for the score can be preset by the user, different from the target value of the termination condition. Alternatively, the design value calculation unit 14 can also set a baseline value based on the target value of the score. During repeated trials, if the optimal score reaches the baseline value, the design value calculation unit 14 narrows the search range for design values. For example, the design value calculation unit 14 narrows the search range for the design values ​​of all design items that are the search objects.

[0114] The proportion by which the search range is narrowed is preset by the user. For example, if the score reaches a benchmark value, at least one of the upper and lower limits of the search range changes by a predetermined proportion. Alternatively, the search range can be narrowed based on the acquisition function. For example, the design value calculation unit 14 removes the range with smaller output values ​​from the search range. This increases the likelihood of obtaining a better characteristic value with fewer trials.

[0115] In this example, a lower score indicates better performance. For instance, a first score based on a first withstand voltage value is lower than a second score based on a second withstand voltage value. A first score based on a first on-resistance value is lower than a second score based on a second on-resistance value. A first withstand voltage value is higher than a second withstand voltage value. A first on-resistance value is lower than a second on-resistance value.

[0116] Alternatively, better characteristics result in a higher score. For example, a first score based on the first withstand voltage value is higher than a second score based on the second withstand voltage value. A first score based on the first on-resistance is higher than a second score based on the second on-resistance. The first withstand voltage value is higher than the second withstand voltage value. The first on-resistance is lower than the second on-resistance.

[0117] (Second Modification)

[0118] Figure 15 This is a flowchart illustrating the design support method for the second variation of the implementation method.

[0119] exist Figure 15 In the design support method DM2 shown, weighting is performed on the design items extracted in step S31.

[0120] In the design support method DM2, after steps S10 and S11, the design value calculation unit 14 weights each design item (step S41). For example, a larger weight is assigned to more important design items than previously set. The design value calculation unit 14 extracts the design items to be searched according to probability. The probability is set based on the weight. A larger weight results in a higher probability of extraction. Afterwards, with... Figure 10 Similarly, the design support method DM2 shown executes steps S31, S32, S12 to S19, S33 and S34.

[0121] This design support method facilitates the extraction of key design items. For example, the weight of impurity concentration (especially drift layer concentration) in semiconductor layers that significantly impact characteristic values ​​is set higher than that of other design items. This increases the likelihood of obtaining better characteristic values.

[0122] For example, in design support method DM2, a subset of design items is extracted. A first processing set is performed on a subset of design values ​​representing the values ​​of the extracted subset of design items. After the first processing set, another subset of design items is extracted. A second processing set is performed on another subset of design values ​​representing the values ​​of the extracted second subset of design items. The probability of extracting the subset of design items is different from the probability of extracting the second subset of design items. For example, one design value of the subset of design items (the first design value) represents the impurity concentration in the semiconductor layer. The probability of extracting the subset of design items is higher than the probability of extracting the second subset of design items.

[0123] Alternatively, the probability of extracting each design item can be the same, and the prescribed number of trials in step S33 can vary according to the weights. The larger the weight, the more trials are set. According to this design support method, the number of trials for important design items increases, thereby increasing the likelihood of obtaining better characteristic values.

[0124] For example, in design support method DM2, the number of times the first processing set is repeated differs from the number of times the second processing set is repeated. When a portion of the design item includes the impurity concentration in the semiconductor layer, it is preferable that the first processing set is repeated more times than the second processing set.

[0125] (Third Modification)

[0126] Figure 16 This is a flowchart illustrating the design support method for the third variation of the implementation method.

[0127] In the design support method DM3 of the third variation, with Figure 4 The design support method DM0 shown also performs steps S10 and S11. The number of design items to be extracted is set to "1" (step S51). With Figure 10 Similarly, the design support method DM1 shown executes steps S31, S32, S12-S19, S33, and S34. Immediately following step S51, one design item is extracted in step S31. In step S34, if it is determined that the termination condition is not met, the design value calculation unit 14 determines whether the score has reached the benchmark value (step S52). For example, in step S52, the best score obtained previously is referenced. The benchmark value is preset by the user. Alternatively, the design value calculation unit 14 may also set the benchmark value based on the target value of the score.

[0128] If the score reaches the benchmark value, the number of design items extracted is set to "n" (step S53). n is an integer greater than 1. Then, in step S31, n design items are extracted. In step S31, one design item is extracted, until the score reaches the benchmark value.

[0129] According to the design support method DM3, the initial search can be completed faster than that of the design support method DM2. Furthermore, the likelihood of encountering local optima in subsequent searches is lower compared to DM2.

[0130] For example, in the design support method DM3, a first processing set for a subset of design values ​​and a second processing set for another subset of design values ​​are executed. The second processing set is executed after the first processing set. In the first processing set, the subset of design values ​​is calculated based on the acquisition function. In the second processing set, the other subset of design values ​​is calculated based on the acquisition function. The subset of design values ​​contains more design values ​​than the other subset. The first processing set is executed before the score reaches the benchmark value. The second processing set is executed after the score reaches the benchmark value.

[0131] In this application specification, "perpendicular" and "parallel" do not mean strictly perpendicular and strictly parallel. For example, deviations in manufacturing processes are included. It is sufficient that they are substantially perpendicular and substantially parallel.

[0132] The implementation methods may also include the following technical solutions.

[0133] (Technical Solution 1)

[0134] A design support method, wherein,

[0135] Input a set of design values ​​into the simulator, which includes multiple design values ​​related to semiconductor components.

[0136] The simulator outputs a set of characteristic values ​​corresponding to the input of the design value set. The set of characteristic values ​​includes multiple characteristic values ​​of the semiconductor element, including a first characteristic value representing on-resistance and a second characteristic value representing withstand voltage.

[0137] A Bayesian estimation-based acquisition function is calculated based on historical data comprising one or more datasets, wherein the datasets are the design value set and the datasets for scores calculated by inputting a portion of the characteristic value set into a first function, wherein the portion of the characteristic value set includes the first characteristic value and the second characteristic value.

[0138] A new set of design values ​​is generated based on the acquisition function.

[0139] (Technical Solution 2)

[0140] According to the design support method described in technical solution 1, wherein,

[0141] The process set is repeatedly executed, which includes inputting the design value set into the simulator, acquiring the characteristic value set, calculating the acquisition function, and generating a new design value set.

[0142] (Technical Solution 3)

[0143] According to the design support method described in technical solution 2, wherein,

[0144] The plurality of processing sets include a first processing set and a second processing set.

[0145] In the first processing set:

[0146] A portion of the multiple design values ​​is calculated based on the acquisition function.

[0147] Generate a new set of design values ​​that includes a portion of the plurality of design values ​​and the remaining design values ​​that have been fixed.

[0148] In the second processing set:

[0149] Another portion of the plurality of design values ​​is calculated based on the acquisition function.

[0150] Generate a new set of design values ​​that includes the other part of the plurality of design values ​​and the remaining design values ​​that have been fixed.

[0151] (Technical Solution 4)

[0152] According to the design support method described in technical solution 3, wherein...

[0153] Execute multiple sets of the first processing and multiple sets of the second processing.

[0154] (Technical Solution 5)

[0155] According to the design support method described in technical solution 4, wherein,

[0156] The first processing set is executed more times than the second processing set.

[0157] (Technical Solution 6)

[0158] According to the design support method described in technical solution 5, wherein,

[0159] A portion of the plurality of design values ​​includes a first design value representing the impurity concentration in the semiconductor layer of the semiconductor element.

[0160] (Technical Solution 7)

[0161] According to any one of technical solutions 3 to 6, the design support method wherein,

[0162] If the score reaches the benchmark value, then the range of at least a portion of the plurality of design values ​​that can be used is narrowed.

[0163] (Technical Solution 8)

[0164] According to any one of technical solutions 3 to 7, the design support method wherein...

[0165] The second processing set is executed after the first processing set.

[0166] The number of design values ​​contained in the other portion of the plurality of design values ​​is greater than the number of design values ​​contained in the first portion of the plurality of design values.

[0167] (Technical Solution 9)

[0168] According to the design support method described in technical solution 8, wherein,

[0169] The first processing set is executed before the score reaches the benchmark value.

[0170] The second processing set is executed after the score reaches the benchmark value.

[0171] (Technical Solution 10)

[0172] According to the design support method described in technical solution 3, wherein...

[0173] The portion of the plurality of design values ​​respectively represents a portion of the values ​​of the plurality of design projects.

[0174] The other portion of the plurality of design values ​​respectively represents the value of another portion of the plurality of design items.

[0175] After extracting the portion of the plurality of design projects, the first processing set is executed.

[0176] After extracting the other part of the plurality of design projects, the second processing set is executed.

[0177] The probability of extracting one part of the plurality of design projects is higher than the probability of extracting another part of the plurality of design projects.

[0178] (Technical Solution 11)

[0179] According to the design support method described in technical solution 10, wherein,

[0180] A portion of the plurality of design values ​​includes a first design value representing the impurity concentration in the semiconductor layer of the semiconductor element.

[0181] (Technical Solution 12)

[0182] According to the design support method described in technical solution 3, wherein...

[0183] In the first processing set, the portion of the plurality of design values ​​includes a total of 1 design value.

[0184] In the second processing set, the number of design values ​​included in the other portion of the plurality of design values ​​is 1.

[0185] (Technical Solution 13)

[0186] According to any one of technical solutions 2 to 12, the design support method wherein,

[0187] Output the number of times the processing set is executed and the score for each processing set.

[0188] (Technical Solution 14)

[0189] According to any one of technical solutions 1 to 13, the design support method wherein,

[0190] The design value set includes at least one design value selected from the group consisting of the impurity concentration in the semiconductor layer of the semiconductor element, the thickness of the semiconductor layer, and the spacing of the gate electrodes of the semiconductor element.

[0191] (Technical Solution 15)

[0192] According to any one of technical solutions 1 to 14, the design support method wherein,

[0193] The characteristic value group also includes at least one of the characteristic values ​​selected from the group consisting of switching charge, gate stored charge, output charge, gate-drain stored charge, gate-source stored charge, channel quantity, and threshold voltage.

[0194] (Technical Solution 16)

[0195] A design support system, wherein the design support system includes a processing device.

[0196] The processing device inputs a set of design values, including multiple design values ​​related to semiconductor components, into the simulator.

[0197] The simulator outputs a set of characteristic values ​​corresponding to the input of the design value set. The set of characteristic values ​​includes multiple characteristic values ​​of the semiconductor element, including a first characteristic value representing on-resistance and a second characteristic value representing withstand voltage.

[0198] A Bayesian estimation-based acquisition function is calculated based on historical data comprising one or more datasets, wherein the datasets are the design value set and the datasets for scores calculated by inputting a portion of the characteristic value set into a first function, wherein the portion of the characteristic value set includes the first characteristic value and the second characteristic value.

[0199] A new set of design values ​​is generated based on the acquisition function.

[0200] (Technical Solution 17)

[0201] According to the design support system described in technical solution 16, wherein,

[0202] The processing device repeatedly executes a set of processes, which includes inputting the design value set into the simulator, acquiring the characteristic value set, calculating the acquisition function, and generating a new design value set.

[0203] (Technical Solution 18)

[0204] A program, wherein the program causes a processing device to:

[0205] Input a set of design values ​​into the simulator, which includes multiple design values ​​related to semiconductor components.

[0206] The simulator outputs a set of characteristic values ​​corresponding to the input of the design value set. The set of characteristic values ​​includes multiple characteristic values ​​of the semiconductor element, including a first characteristic value representing on-resistance and a second characteristic value representing withstand voltage.

[0207] A Bayesian estimation-based acquisition function is calculated based on historical data comprising one or more datasets, wherein the datasets are the design value set and the datasets for scores calculated by inputting a portion of the characteristic value set into a first function, wherein the portion of the characteristic value set includes the first characteristic value and the second characteristic value.

[0208] A new set of design values ​​is generated based on the acquisition function.

[0209] (Technical Solution 19)

[0210] According to the procedure described in technical solution 18, wherein,

[0211] The processing device is made to repeatedly execute a set of processes, which includes inputting the design value set to the simulator, acquiring the characteristic value set, calculating the acquisition function, and generating a new design value set.

[0212] (Technical Solution 20)

[0213] A storage medium storing the program described in technical solution 18 or 19.

[0214] The embodiments of the present invention have been described above with reference to specific examples. However, the embodiments of the present invention are not limited to these specific examples. For example, the specific structure of each element included in the design support system, such as the extraction unit, scoring calculation unit, Bayesian estimation unit, acquisition function calculation unit, design value calculation unit, document generation unit, command generation unit, input unit, output unit, and storage unit, is included within the scope of the present invention, as long as those skilled in the art can similarly implement the present invention and obtain the same effects by appropriately selecting from the known range.

[0215] Furthermore, any solution obtained by combining any two or more elements of each specific example within the scope of technical possibility, as long as it contains the essence of the present invention, is also included within the scope of the present invention.

[0216] Furthermore, all design support methods, design support systems, programs, and storage media implemented by those skilled in the art based on appropriate design modifications possible by those skilled in the art based on the design support methods, design support systems, programs, and storage media described above as embodiments of the present invention, as long as they contain the essence of the present invention, are also within the scope of the present invention.

[0217] Furthermore, within the scope of the present invention, any person skilled in the art will be able to conceive of various modifications and alterations, and it is clear that these modifications and alterations also fall within the scope of the present invention.

[0218] Some embodiments of the invention have been described, but these embodiments are given by way of example and are not intended to limit the scope of the invention. These novel embodiments can be implemented in a wide variety of other ways, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, and are included within the scope of the invention as set forth in the claims and its equivalents.

Claims

1. A design support method, wherein, Input a set of design values ​​into the simulator, which includes multiple design values ​​related to semiconductor components. The simulator outputs a set of characteristic values ​​corresponding to the input of the design value set. The set of characteristic values ​​includes multiple characteristic values ​​of the semiconductor element, including a first characteristic value representing on-resistance and a second characteristic value representing withstand voltage. A Bayesian estimation-based acquisition function is calculated based on historical data comprising one or more datasets, wherein the datasets are the design value set and the datasets for scores calculated by inputting a portion of the characteristic value set into a first function, wherein the portion of the characteristic value set includes the first characteristic value and the second characteristic value. A new set of design values ​​is generated based on the acquisition function. The process set is repeatedly executed, which includes inputting the design value set into the simulator, obtaining the characteristic value set, calculating the acquisition function, and generating a new design value set. The plurality of processing sets include a first processing set and a second processing set. In the first processing set: A portion of the multiple design values ​​is calculated based on the acquisition function. Generate a new set of design values ​​that includes a portion of the plurality of design values ​​and the remaining design values ​​that have been fixed. In the second processing set: Another portion of the plurality of design values ​​is calculated based on the acquisition function. Generate a new set of design values ​​that includes the other portion of the plurality of design values ​​and the remaining design values ​​that have been fixed.

2. The design support method according to claim 1, wherein, Execute multiple sets of the first processing and multiple sets of the second processing.

3. The design support method according to claim 2, wherein, The first processing set is executed more times than the second processing set.

4. The design support method according to claim 1, wherein, In the first processing set, the portion of the plurality of design values ​​includes a total of 1 design value. In the second processing set, the number of design values ​​included in the other portion of the plurality of design values ​​is 1.

5. A design support system comprising a processing device, wherein, The processing device Input a set of design values ​​into the simulator, which includes multiple design values ​​related to semiconductor components. The simulator outputs a set of characteristic values ​​corresponding to the input of the design value set. The set of characteristic values ​​includes multiple characteristic values ​​of the semiconductor element, including a first characteristic value representing on-resistance and a second characteristic value representing withstand voltage. A Bayesian estimation-based acquisition function is calculated based on historical data comprising one or more datasets, wherein the datasets are the design value set and the datasets for scores calculated by inputting a portion of the characteristic value set into a first function, wherein the portion of the characteristic value set includes the first characteristic value and the second characteristic value. A new set of design values ​​is generated based on the acquisition function. The process set is repeatedly executed, which includes inputting the design value set into the simulator, obtaining the characteristic value set, calculating the acquisition function, and generating a new design value set. The plurality of processing sets include a first processing set and a second processing set. In the first processing set: A portion of the multiple design values ​​is calculated based on the acquisition function. Generate a new set of design values ​​that includes a portion of the plurality of design values ​​and the remaining design values ​​that have been fixed. In the second processing set: Another portion of the plurality of design values ​​is calculated based on the acquisition function. Generate a new set of design values ​​that includes the other portion of the plurality of design values ​​and the remaining design values ​​that have been fixed.

6. A storage medium storing a program, wherein, The program is used to enable the processing device to: Input a set of design values ​​into the simulator, which includes multiple design values ​​related to semiconductor components. The simulator outputs a set of characteristic values ​​corresponding to the input of the design value set. The set of characteristic values ​​includes multiple characteristic values ​​of the semiconductor element, including a first characteristic value representing on-resistance and a second characteristic value representing withstand voltage. A Bayesian estimation-based acquisition function is calculated based on historical data comprising one or more datasets, wherein the datasets are the design value set and the datasets for scores calculated by inputting a portion of the characteristic value set into a first function, wherein the portion of the characteristic value set includes the first characteristic value and the second characteristic value. A new set of design values ​​is generated based on the acquisition function. The process set is repeatedly executed, which includes inputting the design value set into the simulator, obtaining the characteristic value set, calculating the acquisition function, and generating a new design value set. The plurality of processing sets include a first processing set and a second processing set. In the first processing set: A portion of the multiple design values ​​is calculated based on the acquisition function. Generate a new set of design values ​​that includes a portion of the plurality of design values ​​and the remaining design values ​​that have been fixed. In the second processing set: Another portion of the plurality of design values ​​is calculated based on the acquisition function. Generate a new set of design values ​​that includes the other portion of the plurality of design values ​​and the remaining design values ​​that have been fixed.

Images