Device and method for predicting semiconductor etching profile by using artificial intelligence model

Abstract

The present invention relates to a technique for predicting a profile deviation of an etching process for forming a microstructure in a semiconductor manufacturing process. A semiconductor etching profile prediction device according to an embodiment of the present invention comprises: a memory in which an artificial intelligence model is stored; and a processor, wherein the memory stores program instructions executable by the processor. The instructions may cause a target process condition to be input into the artificial intelligence model so as to generate a pillar profile prediction image corresponding to the condition, wherein the artificial intelligence model is pre-trained on the basis of various preset training process conditions and pillar profile training images corresponding thereto. According to the present invention, a profile deviation that may occur in an etching process is predicted in advance through an artificial intelligence algorithm-based prediction model, whereby process condition optimization can be induced, yield can be improved, and development time and costs can be effectively reduced by decreasing the number of repetitive experiments.

Description

Device and method for predicting semiconductor etching shape using an artificial intelligence model

[0001] The present invention relates to a technology for predicting shape deviations in an etching process that forms microstructures in a semiconductor manufacturing process, and in particular, to an apparatus and method for predicting semiconductor etching shapes of various structures by utilizing an artificial intelligence-based prediction model.

[0002] In the semiconductor industry, the importance of micro-process technology is increasing day by day due to the trend toward high integration and high performance. In particular, the etching process is a core process for forming desired circuit patterns on a wafer, and dry etching technology using plasma is widely used. Dry etching is a method that removes the wafer surface through chemical reactions and physical ion collisions, and it has the advantages of high etching selectivity and suitability for realizing fine patterns.

[0003] Representative dry etching technologies include Reactive Ion Etching (RIE) and Deep Reactive Ion Etching (DRIE). While RIE can be controlled under relatively simple process conditions, the etching rate decreases as the etching depth increases, and lateral etching occurs due to isotropic etching characteristics, making it difficult to form fine patterns. On the other hand, DRIE can achieve deep etching with a high aspect ratio by repeating the etching and protective film formation steps; however, due to complex process conditions and difficult control, shape deviations are prone to occur.

[0004] Since shape deviations occurring during the etching process significantly impact the performance and reliability of semiconductor devices, various studies have been conducted to minimize them. Conventional research on etching processes has primarily focused on optimizing process conditions through experimental methods or analyzing etching mechanisms to identify the influence of process variables on etching results. Additionally, research utilizing etching simulations to predict etching outcomes and improve process conditions is also actively underway.

[0005] However, conventional experimental methods are time-consuming and costly, and it is practically impossible to experimentally verify all process conditions. Furthermore, existing etching process simulations have limitations, such as the difficulty in accurately modeling complex etching mechanisms and the potential for discrepancies between simulation results and actual etching results. In particular, accurately predicting shape deviations is even more challenging in actual etching processes, where various variables such as pattern density, wafer position, and process conditions interact in a complex manner.

[0006] Therefore, existing technologies focus on quality evaluation or anomaly detection throughout the etching process and have limitations in directly predicting the final shape of specific microstructures. There is currently a lack of technology capable of predicting shape deviations by considering complex variables such as pattern density, wafer position, and process conditions.

[0007] The present invention aims to utilize a deep learning-based prediction model to accurately predict shape deviations caused by various factors occurring in the etching process in advance, thereby enabling process optimization and yield improvement.

[0008] According to one embodiment of the present invention, a semiconductor etching shape prediction device is disclosed, comprising: a memory for storing an artificial intelligence model; and a processor; wherein the memory stores program instructions that are executable by the processor, input target process conditions into the artificial intelligence model to generate a pillar shape prediction image corresponding to the target process conditions, and wherein the artificial intelligence model is pre-learned through a pre-set learning process condition and a pillar shape learning image matched to the learning process condition.

[0009] According to an embodiment, the learning process conditions may include information on diameter, pitch, plasma output, process time, gas composition, and gas pressure.

[0010] According to an embodiment, the pillar shape learning image may correspond to an etched pillar image generated by the matched learning process conditions.

[0011] According to an embodiment, the column shape learning image may be an electron microscope image normalized by a preset method.

[0012] According to an embodiment, the column shape learning image may correspond to the separation of only the column image from the electron microscope image.

[0013] According to an embodiment, the column shape learning image may correspond to the separated column image converted to a preset size.

[0014] According to an embodiment, the column shape learning image may correspond to the separated column image being binarized and processed in black and white.

[0015] According to an embodiment, the artificial intelligence model may include a first artificial intelligence model trained to extract features under the learning process conditions and a second artificial intelligence model trained to generate a pillar shape prediction image corresponding to the features extracted from the first artificial intelligence model.

[0016] According to an embodiment, the first artificial intelligence model and the second artificial intelligence model may be trained together simultaneously to generate the column shape prediction image corresponding to the target process conditions.

[0017]

[0018] According to another embodiment of the present invention, a semiconductor etching shape prediction method performed in a semiconductor etching shape prediction device in which an artificial intelligence model is stored in a storage space may be provided, comprising: a step of inputting target process conditions into the artificial intelligence model; and a step of generating a pillar shape prediction image corresponding to the target process conditions through the artificial intelligence model, wherein the artificial intelligence model is pre-learned through a pre-set learning process condition and a pillar shape learning image matched to the learning process condition.

[0019] According to an embodiment, the learning process conditions may include information on diameter, pitch, plasma output, process time, gas composition, and gas pressure.

[0020] According to an embodiment, the pillar shape learning image may correspond to an etched pillar image generated by the matched learning process conditions.

[0021] According to an embodiment, the column shape learning image may be an electron microscope image normalized by a preset method.

[0022] According to an embodiment, the column shape learning image may correspond to the separation of only the column image from the electron microscope image.

[0023] According to an embodiment, the column shape learning image may correspond to the separated column image converted to a preset size.

[0024] According to an embodiment, the column shape learning image may correspond to the separated column image being binarized and processed in black and white.

[0025] According to an embodiment, the artificial intelligence model may include a first artificial intelligence model trained to extract features under the learning process conditions and a second artificial intelligence model trained to generate a pillar shape prediction image corresponding to the features extracted from the first artificial intelligence model.

[0026] According to an embodiment, the first artificial intelligence model and the second artificial intelligence model may be trained together simultaneously to generate the column shape prediction image corresponding to the target process conditions.

[0027] According to the present invention, by predicting shape deviations of the etching process in advance through a deep learning-based prediction model, process conditions can be optimized and yield improved, and development time and costs can be reduced by reducing the number of experiments.

[0028] The present invention may be better understood in conjunction with the following drawings, wherein the components are not necessarily depicted to scale and the focus is on illustrating various preferred embodiments of the invention.

[0029] FIG. 1 is a block diagram showing the schematic configuration of a semiconductor etching shape prediction device according to one embodiment of the present invention.

[0030] FIG. 2 is a flowchart illustrating a semiconductor etching shape prediction operation according to an embodiment of the present invention.

[0031] FIG. 3 is a flowchart illustrating an artificial intelligence model learning method according to one embodiment of the present invention.

[0032] FIG. 4 is a table showing specific examples of learning process conditions according to one embodiment of the present invention.

[0033] FIG. 5 is a diagram showing the acquisition and preprocessing process of a column shape learning image according to one embodiment of the present invention.

[0034] FIG. 6 is a diagram showing the structure of an artificial intelligence model according to one embodiment of the present invention.

[0035] FIGS. 7 to 9 are examples of predicted images generated when 24 process parameters are input into an artificial intelligence model according to an embodiment of the present invention.

[0036] The technical concept of the present disclosure is subject to various modifications and may have various embodiments, and specific embodiments are illustrated in the drawings and described in detail. However, this is not intended to limit the technical concept of the present disclosure to specific embodiments, and it should be understood that it includes all modifications, equivalents, and substitutions that fall within the scope of the technical concept of the present disclosure.

[0037] In describing the technical concept of the present disclosure, detailed descriptions of related prior art are omitted if it is determined that such descriptions would unnecessarily obscure the essence of the technical concept of the present disclosure. Furthermore, numbers used in the description of the present invention (e.g., First, Second, etc.) are merely identification symbols to distinguish one component from another.

[0038] In addition, when a component is described in this document as being "connected" or "joined" with another component, it should be understood that the component may be directly connected or joined to the other component, but unless otherwise specifically stated, it may also be connected or joined through another component in between.

[0039] Additionally, terms such as “~part,” “~device,” and “~part” described herein refer to a unit that processes at least one function or operation, and this may be implemented as hardware such as a processor, microprocessor, microcontroller, CPU (Central Processing Unit), GPU (Graphics Processing Unit), APU (Accelerate Processor Unit), DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit), FPGA (Field Programmable Gate Array), software, or a combination of hardware and software.

[0040] Furthermore, it is intended to clarify that the classification of the components in this document is merely based on the primary function each component is responsible for. That is, two or more components described below may be combined into a single component, or a single component may be divided into two or more components based on more subdivided functions. Additionally, each component described below may additionally perform some or all of the functions of other components in addition to its own primary function, and it goes without saying that some of the primary functions of each component may be exclusively performed by other components.

[0041] Hereinafter, various embodiments according to the technical concept of the present disclosure will be described in detail in turn.

[0042]

[0043] FIG. 1 shows a schematic configuration of a semiconductor etching shape prediction device (100) according to one embodiment of the present invention. As shown in FIG. 1, the semiconductor etching shape prediction device (100) may include a memory (110) and a processor (120).

[0044] The memory (110) may be a data storage medium such as a magnetic disk (HDD) or an SSD (Solid State Drive). In particular, the memory (110) may serve the function of storing program instructions and an artificial intelligence model that can be executed by the processor (120). These program instructions may be executed under the control of the processor (120) to input target process conditions into the artificial intelligence model and perform the function of generating a column shape prediction image corresponding to the process conditions. The artificial intelligence model may be a model trained through pre-set learning process conditions and a column shape learning image matched thereto.

[0045] The processor (120) is a central processing unit (CPU), a graphics processing unit (GPU), and / or a corresponding computing unit capable of controlling the overall operation of the semiconductor etching shape prediction device (100), and in particular, can control the overall operation of the semiconductor etching shape prediction device (100) by executing program instructions stored in memory (110). That is, the processor (120) can perform the function of executing program instructions to input target process conditions into an artificial intelligence model and outputting a pillar shape prediction image generated from the artificial intelligence model. Through this prediction image, the user can check the results in advance before the actual etching process and obtain the desired results by optimizing the process conditions.

[0046] Embodiments related to the specific operation process of the processor (120) will be described in detail below with reference to FIG. 2.

[0047]

[0048] FIG. 2 is a flowchart illustrating a semiconductor etching shape prediction method according to an embodiment of the present invention. Each step illustrated in FIG. 2 is performed under the control of a processor (120) and is implemented by program instructions stored in memory (110).

[0049] In step S210, the processor (120) can determine whether to input target process conditions. The target process conditions are information about various variables to be used in the etching process, and may include, for example, information about diameter, pitch, plasma output, process time, gas composition, pressure, etc.

[0050] In step S220, the processor (120) can input target process conditions into the artificial intelligence model. When target process conditions are input in step S210, the processor (120) can transmit the conditions to the artificial intelligence model stored in memory (110). At this time, if necessary, the processor (120) may perform a preprocessing process to convert the input target process conditions into a form that the artificial intelligence model can recognize. For example, conditions expressed as strings, such as gas composition, may be converted into numerical data, or the range of the data may be adjusted through processes such as normalization or standardization. Through this preprocessing process, the artificial intelligence model can efficiently process the input target process conditions and derive accurate prediction results.

[0051] In step S230, the processor (120) can generate a column shape prediction image through an artificial intelligence model. The processor (120) can input target process conditions into the artificial intelligence model to generate a column shape prediction image expected under the corresponding process conditions. The artificial intelligence model can calculate an output for the input using program instructions and parameter values ​​stored in memory (110), and various deep learning algorithms such as a Deep Neural Network (DNN) and a Convolutional Neural Network (CNN) can be utilized in this process.

[0052] For example, an artificial intelligence model according to one embodiment of the present invention may include two models, wherein the first model is a Deep Neural Network (DNN) model and the second model is a Convolutional Neural Network (CNN) model. In this case, the first model, the DNN, can extract features from input target process conditions, and the second model, the CNN, can generate a high-resolution pillar shape prediction image based on the features extracted from the first model. The generated image is visually provided to a user through a display (not shown), so that the user can check the results in advance and optimize process conditions before the actual etching process.

[0053] Hereinafter, the learning process of an artificial intelligence model according to an embodiment of the present invention will be described in detail with reference to FIGS. 3 to 6.

[0054]

[0055] FIG. 3 is a flowchart illustrating an artificial intelligence model learning method according to one embodiment of the present invention.

[0056] The training of the artificial intelligence model is based on actual etching process data, and various process conditions are utilized as training data.

[0057] In step S310, the processor (120) receives learning process conditions. The learning process conditions are data used when a deep learning model learns, and may include values ​​of various variables used in an actual etching process. The learning process conditions will be described in detail below with reference to FIG. 4.

[0058]

[0059] FIG. 4 is a table showing specific examples of learning process conditions according to an embodiment of the present invention. The table exemplified in FIG. 4 lists examples of various process variables and their values ​​used to train an artificial intelligence model according to an embodiment of the present invention. The process conditions and values ​​exemplified in this table may be those utilized in an actual etching process.

[0060] 1. Pressure: Value: 82 [%]

[0061] Pressure is a value representing the gas pressure inside the etching chamber and can directly affect plasma density and ion energy. Pressure is an important variable that affects the etching rate, etching profile, and etching uniformity. If the pressure is too high, the etching rate decreases, and if the pressure is too low, the etching uniformity may deteriorate. A pressure of 82% may mean that the opening degree of the valve controlling the amount of gas entering the chamber is fixed at 82%.

[0062] 2. Gas Composition:

[0063] (1) Etch step: SF6 (sulfur hexafluoride) 260 [sccm], O₂ (oxygen): 26 [sccm]

[0064] (2) Passivation step: C₄F8 (octafluorocyclobutane) 120 [sccm]

[0065] Gas composition can indicate the type and flow rate of gases used in the etching process. During the etching step, silicon can be etched primarily using SF6 gas, and etching byproducts can be removed using O₂ gas. In the passivation step, C₄F₅ gas is used to form a protective film on the etched surface. Gas composition has a significant impact on the etching rate, etching selectivity, and etching profile. Increasing the flow rate of SF6 gas accelerates the etching rate, while increasing the flow rate of O₂ gas allows for the effective removal of etching byproducts. C₄F₅ gas can form a protective film on the etched surface to prevent lateral etching and facilitate vertical etching.

[0066] 3. Plasma Power:

[0067] (1) Etch step: ICP (Inductively Coupled Plasma) power 1800 [W], Platen power: 16 [W]

[0068] (2) Passivation Step: ICP Power: 1200 [W]

[0069] Plasma power can represent the amount of power used for plasma generation. ICP power is the power applied to the inductively coupled plasma (ICP) source and can affect plasma density and ion flux. Platen power is the power applied to the bottom electrode where the wafer is placed and can affect ion energy and etching directionality.

[0070] Plasma power is an important variable that affects etching rate, etching uniformity, and ion energy. Increasing ICP power raises plasma density, leading to faster etching rates, while increasing Platen power raises ion energy, which can improve etching directionality.

[0071] 4. Process Time:

[0072] (1) Etch step: 7 [sec]

[0073] (2) Passivation phase: 6 [sec]

[0074] Process time refers to the time for plasma irradiation. In the etching step, plasma is irradiated for 7 seconds to etch silicon, and in the passivation step, plasma is irradiated for 6 seconds to form a protective film on the etched surface. Process time can have a direct effect on the etch depth and etch profile. If the etching time is extended, the etch depth increases, and if the passivation time is extended, lateral etching is suppressed.

[0075] 5. Column Pattern Design:

[0076] Diameter: 50 [um]

[0077] Pitch: 70 [um]

[0078] Pillar pattern design can represent the geometric information of the pillars to be created. Diameter refers to the cross-sectional size of the pillar, while pitch refers to the spacing between pillars. Diameter and pitch determine the density and aspect ratio of the pillars and can significantly influence the etching process results. While smaller diameters and narrower pitches allow for the realization of highly integrated patterns, they may present difficulties during the etching process.

[0079] The information presented in Figure 4 may be an example of one of several variables constituting the learning process conditions. The learning process conditions may include various conditions used in the actual etching process, and it is evident that such diversity of process conditions can contribute to improving the performance of the artificial intelligence model.

[0080]

[0081] Referring again to FIG. 3, in step S320, the processor (120) may receive a pillar shape learning image matched to the learning process conditions. The pillar shape learning image may refer to an image of a pillar obtained when an actual etching process is performed under the corresponding learning process conditions. Below, the pillar shape learning image will be described in detail with reference to FIG. 5.

[0082]

[0083] FIG. 5 is a diagram showing the acquisition and preprocessing process of a column shape learning image according to one embodiment of the present invention.

[0084] 510. SEM Image Acquisition

[0085] After the actual etching process is performed, images using a Scanning Electron Microscope (SEM) can be acquired to observe the shape of the generated pillars. The SEM images may contain various information such as the height, width, angle of inclination, and profile (cross-sectional shape) of the pillars.

[0086] 520. Image Cleaning (Noise Reduction & Specific Region Extraction)

[0087] The SEM image (510) may undergo a preset image refinement process. This process aims to improve the quality of the image and remove unnecessary information to increase the learning efficiency of the artificial intelligence model.

[0088] For example, the SEM image (510) may undergo refinement processes such as noise reduction, specific region extraction, and threshold processing. Each operation will be explained in more detail.

[0089] Noise Reduction: SEM images may contain noise due to factors such as electron beam scattering and contamination of the sample surface. Various image processing techniques can be used to remove noise contained in SEM images. For example, noise in SEM images can be removed by applying smoothing filters and / or median filters.

[0090] Specific Region Extraction: Since only information regarding pillar regions can be used for training AI models, only the pillar regions must be accurately extracted from SEM images. This process can be automated through thresholding (a technique that separates pillar regions from background regions by comparing image pixel values ​​with a threshold), region growing (a technique that groups similar pixels into a single region by exploring surrounding pixels starting from a specific pixel), and / or object detection (a technique that detects pillar regions in images using a pre-trained object detection model).

[0091] 530: Separating pillar images

[0092] In the SEM image (520) where noise is removed and the pillar region is specified, only the pillar image can be accurately separated.

[0093] 540: Size Normalization

[0094] The separated pillar images (530) can be enlarged or reduced to a specific size and normalized to a predetermined size. For example, all pillar images (530) can be adjusted to a size of 256x256 pixels. At this time, they can be normalized so that the aspect ratio of the separated pillar images (530) is maintained.

[0095] 550: Binarization & Grayscale Conversion

[0096] After separation, the size-rectified column image (540) can be binarized and processed into black and white. Binarization can be performed by assigning pixel values ​​to 0 or 255 based on a threshold.

[0097] Each normalized column shape learning image obtained through the above-described operation can be matched with any one of the learning process conditions. That is, an SEM image generated by any one of the learning process conditions can be normalized by the above-described operation.

[0098]

[0099] Referring again to FIG. 3, in step S330, the artificial intelligence model can be trained using labeled training process conditions and a column shape training image (hereinafter referred to as 'training data'). The artificial intelligence model may include a first model and a second model. The first model may be a DNN model, and the second model may be a CNN model. Furthermore, the first model and the second model may be trained simultaneously (End-to-End learning method) rather than being combined after being trained separately. The structure of an artificial intelligence model according to an embodiment of the present invention will be described in detail below with reference to FIG. 6.

[0100]

[0101] FIG. 6 is a diagram showing the structure of an artificial intelligence model (600) according to one embodiment of the present invention.

[0102] An artificial intelligence model (600) according to one embodiment of the present invention learns the complex relationship between etching process conditions and the resulting pillar shape, and aims to accurately predict the pillar shape when new process conditions are given.

[0103] The artificial intelligence model (600) can be largely composed of an input layer (610), a DNN (Deep Neural Network, 620), and a CNN (Convolutional Neural Network, 630), and each component can perform the following roles.

[0104] 610: Input Layer

[0105] The input layer (610) is a part that defines the data input to the artificial intelligence model (600), and the training data required for performing training can be transmitted to the artificial intelligence model (600) through the input layer (610).

[0106] 620: Model 1, DNN (Deep Neural Network): Feature Extractor

[0107] The first model, DNN (620), can perform the role of extracting key features that affect the column shape from process condition data transmitted from the input layer (610). The DNN (620) can extract key features that affect the column shape from the input data. The extracted features can be expressed in the form of a feature vector and used as input to the second model, CNN (630).

[0108] 630: Model 2, CNN (Convolutional Neural Network): Image Generator

[0109] The CNN (630) can receive feature vectors extracted from the DNN (620) as input and ultimately generate a high-resolution pillar shape prediction image. Since the CNN (630) can have a structure specialized for image processing and generation, it can receive feature vectors extracted from the DNN (620) as input and generate a high-resolution pillar shape prediction image.

[0110] The first model (DNN, 620) and the second model (CNN, 630) of the artificial intelligence model (600) can be trained simultaneously (End-to-End method) in a way that minimizes the error between the column shape prediction image and the column shape learning image.

[0111] Through this structure, an artificial intelligence model (600) capable of precisely predicting changes in column shape according to changes in process conditions can be trained.

[0112]

[0113] FIGS. 7 to 9 are examples of predicted images generated when 24 process parameters are input into an artificial intelligence model according to an embodiment of the present invention.

[0114] In FIGS. 7 to 9, predicted images and actual images (Ground Truth images) for different diameter and pitch combinations are shown side by side so that the prediction performance of the artificial intelligence model (600) can be visually verified.

[0115] The bottom of each image pair represents a pillar shape image (Predicted image) predicted by an artificial intelligence model (600), and the top represents a SEM image (Ground Truth image) obtained by performing an actual etching process under the same process conditions. The diameter and pitch values ​​used to generate the image pair are specified at the top of each image pair.

[0116] Through FIGS. 7 to 9, it can be confirmed that an artificial intelligence model (600) according to one embodiment of the present invention can effectively learn various column shape changes generated by changing diameter and pitch values, and based on this, generate a prediction image for new process conditions. In addition, since the prediction image generated by the artificial intelligence model (600) according to one embodiment of the present invention shows a shape very similar to an actual SEM image, it can be confirmed that the artificial intelligence model (600) presented in the present invention can effectively learn the relationship between process variables and shape and accurately predict actual process results.

[0117]

[0118] As described above, according to the present invention, by predicting shape deviations of the etching process in advance through an artificial intelligence-based prediction model, process conditions can be optimized and yield improved, and development time and costs can be reduced by reducing the number of experiments.

[0119]

[0120] Meanwhile, the embodiments described above may be implemented as hardware components, software components, and / or combinations of hardware and software components. For example, the devices, methods, and components described in the embodiments may be implemented using a general-purpose computer or a special-purpose computer, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. In addition, other processing configurations, such as parallel processors, are also possible.

[0121] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium, or device so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on computer-readable recording media.

[0122] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may store program instructions, data files, data structures, etc., either individually or in combination, and the program instructions recorded on the medium may be those specifically designed and configured for the embodiment or those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc.

[0123] The hardware device described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

Claims

1. Memory for storing an artificial intelligence model; and processor; Includes, The above memory is executable by the above processor, It stores program commands that input target process conditions into the artificial intelligence model to generate a column shape prediction image corresponding to the target process conditions, and A semiconductor etching shape prediction device in which the above artificial intelligence model is pre-trained through preset learning process conditions and a pillar shape learning image matched to the above learning process conditions.

2. In Paragraph 1, A semiconductor etching shape prediction device, wherein the above learning process conditions include information on diameter, pitch, plasma output, process time, gas composition, and gas pressure.

3. In Paragraph 1, A semiconductor etching shape prediction device in which the above-mentioned pillar shape learning image corresponds to an etching pillar image generated by the matched learning process conditions.

4. In Paragraph 3, A semiconductor etching shape prediction device in which the above-mentioned pillar shape learning image is an electron microscope image normalized by a preset method.

5. In Paragraph 4, A semiconductor etching shape prediction device in which the above-mentioned pillar shape learning image corresponds to only the pillar image separated from the above-mentioned electron microscope image.

6. In Paragraph 5, A semiconductor etching shape prediction device in which the above-mentioned pillar shape learning image corresponds to the separated pillar image converted to a preset size.

7. In Paragraph 5, A semiconductor etching shape prediction device in which the above-mentioned pillar shape learning image corresponds to the separated pillar image that has been binarized and processed into black and white.

8. In Paragraph 1, The above artificial intelligence model A semiconductor etching shape prediction device comprising a first artificial intelligence model trained to extract features under the above-mentioned learning process conditions and a second artificial intelligence model trained to generate a pillar shape prediction image corresponding to the features extracted from the first artificial intelligence model.

9. In Paragraph 8, A semiconductor etching shape prediction device in which the first artificial intelligence model and the second artificial intelligence model are simultaneously trained to generate the pillar shape prediction image corresponding to the target process conditions.

10. A semiconductor etching shape prediction method performed in a semiconductor etching shape prediction device in which an artificial intelligence model is stored in a provided storage space, wherein A step of inputting target process conditions into the artificial intelligence model; and A step of generating a column shape prediction image corresponding to the target process conditions through the artificial intelligence model; Includes, A semiconductor etching shape prediction method in which the above artificial intelligence model is pre-trained through pre-set learning process conditions and a pillar shape learning image matched to the above learning process conditions.

11. In Paragraph 10, A method for predicting semiconductor etching shape, wherein the above learning process conditions include information on diameter, pitch, plasma output, process time, gas composition, and gas pressure.

12. In Paragraph 10, A semiconductor etching shape prediction method in which the above-mentioned pillar shape learning image corresponds to an etching pillar image generated by the matched learning process conditions.

13. In Paragraph 12, A semiconductor etching shape prediction method in which the above-mentioned pillar shape learning image is an electron microscope image normalized by a preset method.

14. In Paragraph 13, A semiconductor etching shape prediction method in which the above-mentioned pillar shape learning image corresponds to only the pillar image separated from the above-mentioned electron microscope image.

15. In Paragraph 14, A semiconductor etching shape prediction method in which the above-mentioned pillar shape learning image corresponds to the separated pillar image converted to a preset size.

16. In Paragraph 14, A semiconductor etching shape prediction method in which the above-mentioned pillar shape learning image corresponds to the separated pillar image that has been binarized and processed into black and white.

17. In Paragraph 10, The above artificial intelligence model A semiconductor etching shape prediction method comprising a first artificial intelligence model trained to extract features under the above-mentioned learning process conditions and a second artificial intelligence model trained to generate a pillar shape prediction image corresponding to the features extracted from the first artificial intelligence model.

18. In Paragraph 17, A semiconductor etching shape prediction method in which the first artificial intelligence model and the second artificial intelligence model are simultaneously trained to generate the pillar shape prediction image corresponding to the target process conditions.

Images