Model training method and device and storage medium

Abstract

The invention discloses a model training method and device and a storage medium, and the method comprises the steps: obtaining a photoacid concentration distribution sample, calling a photoresist post-baking prediction model to carry out the post-baking prediction of the photoacid concentration distribution sample, and obtaining a residual photoacid concentration distribution prediction result, an alkali concentration distribution prediction result and a polymer concentration distribution prediction result; respectively carrying out post-baking chemical diffusion correction on the prediction results to obtain a residual photoacid concentration distribution correction result, an alkali concentration distribution correction result and a polymer concentration distribution correction result; and correcting model parameters of the photoresist post-baking prediction model according to the residual photoacid concentration distribution correction result, the residual photoacid concentration distribution label, the alkali concentration distribution correction result, the alkali concentration distribution label, the polymer concentration distribution correction result and the polymer concentration distribution label. According to the method, the prediction precision of the trained model during photoresist modeling simulation can be improved, so that the prediction accuracy of the trained model can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of quantum computing, and particularly relates to a model training method and device applied to quantum computing and a storage medium. BACKGROUND

[0002] In the manufacturing of ultra-large scale integrated circuits, post-baking and developing processes are crucial for feature size reduction and yield improvement. The post-baking process can reduce standing wave effects and make chemical reactions more complete by heating the photoresist on the wafer after exposure. The subsequent developing process dissolves the photoresist in the exposed area or the photoresist in the unexposed area with a developing solution, and finally realizes the transfer of the chip pattern to the photoresist.

[0003] In order to ensure chip performance, capacity and yield, photoresist modeling simulation of the post-baking process is generally performed in advance. In the related art, a method of using a deep learning model to perform photoresist modeling simulation is proposed, which can realize simulation prediction of the photoresist with less time consumption. However, the deep learning model used for photoresist modeling simulation still has the problem of poor prediction accuracy. Therefore, how to improve the prediction accuracy of the deep learning model used for photoresist modeling simulation is a technical problem to be solved. SUMMARY

[0004] The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims.

[0005] The embodiments of the present application provide a model training method, device and storage medium, which can improve the prediction accuracy of the trained model in photoresist modeling simulation, thereby improving the prediction accuracy of the trained model.

[0006] In one aspect, the embodiments of the present application provide a model training method, comprising the following steps:

[0007] obtaining a photoacid concentration distribution sample and a sample label corresponding to the photoacid concentration distribution sample, the sample label comprising a residual photoacid concentration distribution label, an alkali concentration distribution label and a polymer concentration distribution label;

[0008] calling a photoresist post-baking prediction model to perform post-baking prediction on the photoacid concentration distribution sample to obtain a residual photoacid concentration distribution prediction result, an alkali concentration distribution prediction result and a polymer concentration distribution prediction result;

[0009] performing post-baking chemical diffusion correction on the residual photoacid concentration distribution prediction result, the alkali concentration distribution prediction result and the polymer concentration distribution prediction result, respectively, to obtain a residual photoacid concentration distribution correction result, an alkali concentration distribution correction result and a polymer concentration distribution correction result;

[0010] According to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label, the base concentration distribution correction result and the base concentration distribution label, and the polymer concentration distribution correction result and the polymer concentration distribution label, the model parameters of the photoresist post-baking prediction model are corrected.

[0011] In another aspect, the embodiments of the present application also provide a model training device, comprising:

[0012] A sample acquisition unit is configured to acquire a photoacid concentration distribution sample and a sample label corresponding to the photoacid concentration distribution sample, wherein the sample label comprises a remaining photoacid concentration distribution label, a base concentration distribution label, and a polymer concentration distribution label.

[0013] A post-baking prediction unit is configured to call a photoresist post-baking prediction model to perform post-baking prediction on the photoacid concentration distribution sample, to obtain a remaining photoacid concentration distribution prediction result, a base concentration distribution prediction result, and a polymer concentration distribution prediction result.

[0014] A result correction unit is configured to perform post-baking chemical diffusion correction on the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, and the polymer concentration distribution prediction result, respectively, to obtain a remaining photoacid concentration distribution correction result, a base concentration distribution correction result, and a polymer concentration distribution correction result.

[0015] A parameter correction unit is configured to correct model parameters of the photoresist post-baking prediction model according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label, the base concentration distribution correction result and the base concentration distribution label, and the polymer concentration distribution correction result and the polymer concentration distribution label.

[0016] Optionally, the sample acquisition unit is specifically configured to:

[0017] acquire an initial light intensity distribution and an initial photoacid concentration distribution in a wafer photoresist;

[0018] perform exposure simulation processing according to the initial light intensity distribution and the initial photoacid concentration distribution, to obtain a post-exposure photoacid concentration distribution;

[0019] calculate a difference between the post-exposure photoacid concentration distribution and the initial photoacid concentration distribution, to obtain the photoacid concentration distribution sample.

[0020] Optionally, the sample acquisition unit is specifically configured to:

[0021] performing an exposure simulation calculation process with the initial light intensity distribution and the initial photoacid concentration distribution as input parameters, wherein the exposure simulation calculation process comprises: calculating a light intensity change rate with depth according to the initial light intensity distribution and the initial photoacid concentration distribution; calculating a next-time light intensity distribution according to the light intensity change rate with depth and the initial light intensity distribution; calculating a photoacid concentration change rate with time according to the next-time light intensity distribution and the initial photoacid concentration distribution; and calculating a next-time photoacid concentration distribution according to the photoacid concentration change rate with time and the initial photoacid concentration distribution;

[0022] repeating the exposure simulation calculation process with the next-time light intensity distribution and the next-time photoacid concentration distribution as input parameters until a photoacid concentration distribution at an exposure time is obtained;

[0023] taking the photoacid concentration distribution at the exposure time as a post-exposure photoacid concentration distribution.

[0024] Optionally, the sample obtaining unit is specifically configured to:

[0025] obtain an initial base concentration distribution and an initial polymer concentration distribution;

[0026] performing a post-baking chemical diffusion simulation process according to the photoacid concentration distribution sample, the initial base concentration distribution and the initial polymer concentration distribution to obtain a residual photoacid concentration distribution simulation result, a base concentration distribution simulation result and a polymer concentration distribution simulation result;

[0027] taking the residual photoacid concentration distribution simulation result as the residual photoacid concentration distribution label;

[0028] taking the base concentration distribution simulation result as the base concentration distribution label;

[0029] taking the polymer concentration distribution simulation result as the polymer concentration distribution label.

[0030] Optionally, the sample obtaining unit is specifically configured to:

[0031] performing a post-baking chemical diffusion simulation calculation process with the photoacid concentration distribution sample, the initial base concentration distribution and the initial polymer concentration distribution as input parameters, wherein the post-baking chemical diffusion simulation calculation process comprises: calculating a photoacid concentration distribution at a next time according to the photoacid concentration distribution sample, the initial base concentration distribution, a preset photoacid diffusion rate and a preset acid-base neutralization rate; calculating a base concentration distribution at the next time according to the photoacid concentration distribution sample, the initial base concentration distribution, a preset base diffusion rate and the acid-base neutralization rate; and calculating a polymer concentration distribution at the next time according to the photoacid concentration distribution sample, the initial polymer concentration distribution and a preset polymer deprotection reaction rate;

[0032] repeating the post-baking chemical diffusion simulation calculation process with the photoacid concentration distribution at the next time, the base concentration distribution at the next time and the polymer concentration distribution at the next time as input parameters until a photoacid concentration distribution at a post-baking time, a base concentration distribution at the post-baking time and a polymer concentration distribution at the post-baking time are obtained;

[0033] corresponding the photoacid concentration distribution at the post-baking time, the base concentration distribution at the post-baking time and the polymer concentration distribution at the post-baking time as a remaining photoacid concentration distribution simulation result, a base concentration distribution simulation result and a polymer concentration distribution simulation result.

[0034] Optionally, the result correction unit is specifically configured to:

[0035] performing a post-baking chemical diffusion correction calculation process with the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result and the polymer concentration distribution prediction result as input parameters, wherein the post-baking chemical diffusion correction calculation process comprises: calculating a photoacid concentration distribution at a first iteration according to the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, a preset photoacid diffusion rate and a preset acid-base neutralization rate; calculating a base concentration distribution at the first iteration according to the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, a preset base diffusion rate and the acid-base neutralization rate; and calculating a polymer concentration distribution at the first iteration according to the remaining photoacid concentration distribution prediction result, the polymer concentration distribution prediction result and a preset polymer deprotection reaction rate;

[0036] repeating the post-baking chemical diffusion correction calculation process with the photoacid concentration distribution at the first iteration, the base concentration distribution at the first iteration and the polymer concentration distribution at the first iteration as input parameters until a photoacid concentration distribution at an iteration number, a base concentration distribution at the iteration number and a polymer concentration distribution at the iteration number are obtained;

[0037] The light acid concentration distribution reached the iteration number, the base concentration distribution reached the iteration number, and the polymer concentration distribution reached the iteration number are corresponded as a remaining light acid concentration distribution correction result, a base concentration distribution correction result, and a polymer concentration distribution correction result.

[0038] Optionally, the parameter correction unit is specifically used for:

[0039] According to the remaining light acid concentration distribution correction result and the remaining light acid concentration distribution label, a light acid concentration distribution loss value is calculated.

[0040] According to the base concentration distribution correction result and the base concentration distribution label, a base concentration distribution loss value is calculated.

[0041] According to the polymer concentration distribution correction result and the polymer concentration distribution label, a polymer concentration distribution loss value is calculated.

[0042] According to the light acid concentration distribution loss value, the base concentration distribution loss value, and the polymer concentration distribution loss value, the model parameters of the photoresist post-baking prediction model are corrected.

[0043] Optionally, the parameter correction unit is specifically used for:

[0044] The square of the difference between the remaining light acid concentration distribution correction result and the remaining light acid concentration distribution label is calculated to obtain a first light acid concentration distribution loss term.

[0045] According to the remaining light acid concentration distribution correction result and the remaining light acid concentration distribution prediction result, a second light acid concentration distribution loss term is calculated.

[0046] According to the first light acid concentration distribution loss term and the second light acid concentration distribution loss term, a light acid concentration distribution loss value is calculated.

[0047] Optionally, the parameter correction unit is specifically used for:

[0048] The square of the difference between the base concentration distribution correction result and the base concentration distribution label is calculated to obtain a first base concentration distribution loss term.

[0049] According to the base concentration distribution correction result and the base concentration distribution prediction result, a second base concentration distribution loss term is calculated.

[0050] According to the first base concentration distribution loss term and the second base concentration distribution loss term, a base concentration distribution loss value is calculated.

[0051] Optionally, the parameter correction unit is specifically used for:

[0052] calculating a square of a difference between the polymer concentration distribution correction result and the polymer concentration distribution label to obtain a first polymer concentration distribution loss term;

[0053] calculating a second polymer concentration distribution loss term according to the polymer concentration distribution correction result and the polymer concentration distribution prediction result;

[0054] calculating a polymer concentration distribution loss value according to the first polymer concentration distribution loss term and the second polymer concentration distribution loss term.

[0055] Optionally, the post-baking prediction unit is specifically configured to:

[0056] performing local feature extraction on the photoacid concentration distribution sample to obtain a sample local feature;

[0057] performing dimension expansion processing on the photoacid concentration distribution sample to obtain a first feature, a dimension number of the first feature being greater than a dimension number of the sample local feature;

[0058] performing Fourier transform processing on the first feature to obtain a second feature;

[0059] performing complex convolution processing on the second feature and a model weight parameter of the photoresist post-baking prediction model to obtain a third feature;

[0060] performing inverse Fourier transform processing on the third feature to obtain a fourth feature;

[0061] performing dimension reduction processing on the fourth feature to obtain a sample global feature, a dimension number of the sample global feature being consistent with the dimension number of the sample local feature;

[0062] performing post-baking prediction based on the sample global feature and the sample local feature to obtain a post-baking prediction result, wherein the post-baking prediction result comprises one of the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result or the polymer concentration distribution prediction result.

[0063] In another aspect, an electronic device is provided, comprising:

[0064] at least one processor;

[0065] at least one memory configured to store at least one program;

[0066] when at least one of the programs is executed by at least one of the processors, the model training method as described above is implemented.

[0067] In another aspect, an embodiment of the present application further provides a computer readable storage medium, wherein a computer program executable by a processor is stored, and the computer program executable by the processor is used to implement the model training method as above when executed by the processor.

[0068] In another aspect, an embodiment of the present application further provides a computer program product, comprising a computer program or computer instructions, the computer program or computer instructions being stored in a computer readable storage medium, and a processor of an electronic device reading the computer program or computer instructions from the computer readable storage medium, and the processor executing the computer program or computer instructions to enable the electronic device to perform the model training method as above.

[0069] In the technical scheme provided by the embodiment of the present application, after the photoresist post-baking prediction model is used to perform post-baking prediction on the photoacid concentration distribution sample to obtain the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result and the polymer concentration distribution prediction result, the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result and the polymer concentration distribution prediction result are respectively corrected by post-baking chemical diffusion to obtain the remaining photoacid concentration distribution correction result, the base concentration distribution correction result and the polymer concentration distribution correction result. By correcting the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result and the polymer concentration distribution prediction result by post-baking chemical diffusion, the remaining photoacid concentration distribution correction result, the base concentration distribution correction result and the polymer concentration distribution correction result after correction can be more accurate, which is conducive to improving the imaging fidelity. Then, the model parameters of the photoresist post-baking prediction model are corrected according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label, the base concentration distribution correction result and the base concentration distribution label, and the polymer concentration distribution correction result and the polymer concentration distribution label. Since the model parameters of the photoresist post-baking prediction model are corrected in combination with the results obtained by post-baking chemical diffusion correction, the model parameters of the corrected photoresist post-baking prediction model can incorporate the physical information of the post-baking process, thereby effectively improving the prediction accuracy of the trained photoresist post-baking prediction model when performing photoresist modeling simulation, and further effectively improving the prediction accuracy of the trained photoresist post-baking prediction model.

[0070] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be realized and attained by the structure particularly pointed out in the description and the appended drawings. BRIEF DESCRIPTION OF DRAWINGS

[0071] The accompanying drawings are used to provide a further understanding of the technical solutions of the present application, constitute a part of the specification, and are used together with the embodiments of the present application to explain the technical solutions of the present application, and do not constitute a limitation on the technical solutions of the present application.

[0072] Figure 1 is a schematic diagram of an implementation environment provided by an embodiment of the present application;

[0073] Figure 2 is a flowchart of a model training method provided by an embodiment of the present application;

[0074] Figure 3 is a schematic diagram of the overall structure of a model training method provided by an embodiment of the present application;

[0075] Figure 4 is a schematic diagram of a pre-trained photoresist post-baking prediction model provided by an embodiment of the present application;

[0076] Figure 5 is a flowchart of a process of obtaining a photoacid concentration distribution sample provided by an embodiment of the present application;

[0077] Figure 6 is a flowchart of a process of obtaining a sample label provided by an embodiment of the present application;

[0078] Figure 7 is a flowchart of step 230 in Figure 2 ;

[0079] Figure 8 is a flowchart of step 240 in Figure 2 ;

[0080] Figure 9 is a schematic diagram of prediction errors of multiple algorithms provided by an embodiment of the present application;

[0081] Figure 10 is a schematic diagram of training time consumption of multiple algorithms provided by an embodiment of the present application;

[0082] Figure 11 is a schematic diagram of inference time consumption of multiple algorithms provided by an embodiment of the present application;

[0083] Figure 12 is a schematic diagram of a model training device provided by an embodiment of the present application;

[0084] Figure 13 is a schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0085] The present application will be further described below with reference to the drawings and specific embodiments. The described embodiments should not be considered limiting to the present application, all other embodiments obtained by those of ordinary skill in the art without creative work under the premise of not conflicting with the described embodiments belong to the scope of protection of the present application.

[0086] In the following description, "some embodiments" are described, which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0087] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0088] Before the embodiments of the present application are further described in detail, the terms and phrases involved in the embodiments of the present application are explained, and the terms and phrases involved in the embodiments of the present application are applicable to the following explanations.

[0089] Artificial intelligence (AI) is the use of digital computers or digital computer controlled machines to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology of computer science, which tries to understand the essence of intelligence and produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence is the design principle and implementation method of various intelligent machines, so that machines have the functions of perception, reasoning and decision making. Artificial intelligence technology is a comprehensive discipline, involving a wide range of fields, both hardware and software technologies. Artificial intelligence basic technologies generally include technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, pre-training model technology, operation / interaction system, mechatronics, etc. Among them, the pre-training model is also called large model, basic model, which can be widely applied to downstream tasks in various directions of artificial intelligence after fine tuning. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology and machine learning / deep learning, etc.

[0090] Wafer photoresist: A key material used in semiconductor manufacturing, also known as photoresist or photoresist. It is a special photosensitive polymer material used to make micro patterns and structures in semiconductor processes. Wafer photoresist is usually coated on the surface of a silicon wafer (wafer), then exposed to light through a photolithography machine, developed and other process steps, finally forming the desired pattern. The main role of wafer photoresist is as a photoresist layer, used to transfer the pattern in the photolithography machine to the surface of the silicon wafer. During the photoresist process, the wafer photoresist will undergo chemical changes after exposure to light, forming exposed and unexposed areas, then removing the unexposed areas through development and other steps, leaving the desired pattern. Wafer photoresist usually has the following characteristics: photosensitivity: can undergo chemical reaction after exposure to light through photolithography machine. Etch resistance: can protect the surface of the wafer in the environment of chemical corrosion and etching. Resolution: can make micro-sized patterns and structures. Film forming: can be uniformly coated on the surface of the wafer and form a flat photoresist layer.

[0091] Photoacid: A molecule that turns into a strong acid under light irradiation. After absorbing light, photoacid generates acid, which acts as a catalyst during post-exposure bake (PEB) process, causing the unstable groups hanging on the resin to fall off, thereby changing the base solubility of the resin. When enough hanging groups fall off, the resin can be dissolved in the developer. Mainly used to improve the purity and quality of the product. In semiconductor processes, especially in the processing of photosensitive materials, the generation and diffusion of photoacid is an important chemical process. First, photoacid is generated under light irradiation, and its generation rate may be related to the intensity of light irradiation, material properties and reaction conditions. The generated photoacid will diffuse in the material, and its diffusion process can be described by Fick's second law.

[0092] Post exposure bake: A heating process that can accelerate the diffusion and reaction of photoacid. During post exposure bake, the diffusion coefficient of photoacid may change with temperature, which can be described by the Arrhenius equation. Combining Fick's second law and the Arrhenius equation, we can get a set of equations describing the concentration distribution of photoacid and chemical diffusion during post exposure bake. Generally, in the process of photoresist exposure, light irradiates the silicon substrate through the photoresist, which will produce reflection at the interface between the photoresist and the substrate. These reflected light and incident light will form interference, making the distribution of light intensity along the depth direction of the glue uneven, forming standing wave effect. This effect will affect the imaging resolution of photoresist and the verticality of pattern sidewall. In order to eliminate the standing wave effect, we can use the method of thermal baking (i.e. post exposure bake), use photoresist with dye or add antireflection layer.

[0093] The photoresist post-baking prediction model is a deep learning model provided by the embodiment of the present application, which is used for post-baking prediction of photoacid concentration distribution to obtain relevant prediction results. The relevant prediction results include photoacid concentration distribution prediction results, base concentration distribution prediction results and polymer concentration distribution prediction results.

[0094] In the manufacture of ultra-large scale integrated circuits, post-baking and developing processes are crucial for feature size reduction and yield improvement. In order to ensure chip performance, capacity and yield, photoresist modeling simulation of post-baking process is generally performed in advance. In the related art, a method of using a deep learning model to perform photoresist modeling simulation is proposed, which can realize simulation prediction of photoresist with less time consumption, but has the problem of poor prediction accuracy.

[0095] Based on this, the embodiment of the present application provides a model training method, a model training device, an electronic device, a computer readable storage medium and a computer program product. First, the photoresist post-baking prediction model is called to perform post-baking prediction on the photoacid concentration distribution sample to obtain the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result and the polymer concentration distribution prediction result. Then, the above three prediction results are respectively corrected by post-baking chemical diffusion to obtain the remaining photoacid concentration distribution correction result, the base concentration distribution correction result and the polymer concentration distribution correction result, wherein the correction results contain the physical information of the post-baking process. The model parameters of the photoresist post-baking prediction model are corrected using the above three correction results and the corresponding labels. Since the model parameters of the photoresist post-baking prediction model are corrected in combination with the results obtained by post-baking chemical diffusion correction, the model parameters of the corrected photoresist post-baking prediction model can be fused with the physical information of the post-baking process, thereby effectively improving the prediction accuracy of the trained photoresist post-baking prediction model in photoresist modeling simulation, and further effectively improving the prediction accuracy of the trained photoresist post-baking prediction model.

[0096] Reference is made to Figure 1 , Figure 1 A schematic diagram of an implementation environment provided by the embodiment of the present application is provided, which includes a terminal 101 and a server 102. The terminal 101 and the server 102 can be directly or indirectly connected through wired or wireless communication. Among them, the terminal 101 and the server 102 can be nodes in the blockchain, and the embodiment does not make specific limitations thereto.

[0097] In some embodiments, the terminal 101 can be a smartphone, a tablet computer, a notebook computer, a desktop computer, a smart voice interaction device, a smart home appliance, a vehicle-mounted terminal, an aircraft, a VR (Virtual Reality) device, an AR (Augmented Reality) device, etc., but is not limited thereto. Optionally, the terminal 101 can obtain a photoacid concentration distribution sample and a sample label corresponding to the photoacid concentration distribution sample, and send a model training request to the server 102.

[0098] In some embodiments, the server 102 can be a standalone server, a server cluster composed of multiple servers, or a distributed system, and can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs (Content Delivery Networks), and big data and artificial intelligence platforms. In some embodiments, the server 102 mainly undertakes computing work, and the terminal 101 undertakes secondary computing work; or the server 102 undertakes secondary computing work, and the terminal 101 undertakes primary computing work; or the server 102 and the terminal 101 cooperatively compute in a distributed computing architecture. Optionally, the server 102 can provide a model training service according to the received model training request, i.e., correct the model parameters of the photoresist post-bake prediction model based on the photoacid concentration distribution sample and the sample label corresponding to the photoacid concentration distribution sample.

[0099] Referring to FIG. 1, Figure 1 In an application scenario, it is assumed that the terminal 101 is a computer, and the terminal 101 provides a model training service. In the process of model training by the target object using the terminal 101, the terminal 101 obtains a photoacid concentration distribution sample and a sample label corresponding to the photoacid concentration distribution sample, and the sample label includes a residual photoacid concentration distribution label, an alkali concentration distribution label, and a polymer concentration distribution label. The terminal 101 calls the photoresist post-bake prediction model to perform post-bake prediction on the photoacid concentration distribution sample, and obtains a residual photoacid concentration distribution prediction result, an alkali concentration distribution prediction result, and a polymer concentration distribution prediction result. The terminal 101 performs post-bake chemical diffusion correction on the residual photoacid concentration distribution prediction result, the alkali concentration distribution prediction result, and the polymer concentration distribution prediction result, respectively, and obtains a residual photoacid concentration distribution correction result, an alkali concentration distribution correction result, and a polymer concentration distribution correction result. The terminal 101 corrects the model parameters of the photoresist post-bake prediction model according to the residual photoacid concentration distribution correction result and the residual photoacid concentration distribution label, the alkali concentration distribution correction result and the alkali concentration distribution label, and the polymer concentration distribution correction result and the polymer concentration distribution label.

[0100] The model training service can be divided into a model prediction sub-service, a diffusion correction sub-service, and a model correction sub-service. In an embodiment, the model prediction sub-service and the model correction sub-service can be provided by the terminal 101, and the diffusion correction sub-service can be provided by the server 102. Specifically, in the process of providing the model prediction sub-service by the terminal 101, the photoresist post-baking prediction model is called to perform post-baking prediction on the photoacid concentration distribution sample to obtain the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, and the polymer concentration distribution prediction result. After the terminal 101 calls the photoresist post-baking prediction model to perform post-baking prediction on the photoacid concentration distribution sample to obtain the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, and the polymer concentration distribution prediction result, the terminal 101 sends the prediction results and a diffusion correction request to the server 102. In the process of providing the diffusion correction sub-service by the server 102, the diffusion correction sub-service corrects the prediction results according to the received diffusion correction request to obtain the remaining photoacid concentration distribution correction result, the base concentration distribution correction result, and the polymer concentration distribution correction result, and sends the remaining photoacid concentration distribution correction result, the base concentration distribution correction result, and the polymer concentration distribution correction result to the terminal 101. In the process of providing the model correction sub-service by the terminal 101, the terminal 101 receives the remaining photoacid concentration distribution correction result, the base concentration distribution correction result, and the polymer concentration distribution correction result sent by the server 102, and then corrects the model parameters of the photoresist post-baking prediction model based on the correction results and the label.

[0101] In an embodiment, the server 102 can also provide the model training service, including the model prediction sub-service, the diffusion correction sub-service, and the model correction sub-service. The process of providing the model training service by the server 102 is similar to the process of providing the model training service by the terminal 101 described above, and will not be described here.

[0102] Figure 2 FIG. 1 is a flowchart of a model training method provided by an embodiment of the present application. The model training method is executed by a terminal or a server alone or jointly by a terminal and a server. In this embodiment, the model training method is executed by a terminal as an example. Referring to FIG. 1, the model training method includes but is not limited to steps 210 to 240. Figure 2 Figure 2

[0103] Step 210: Obtain a photoacid concentration distribution sample and a sample label corresponding to the photoacid concentration distribution sample. The sample label includes a remaining photoacid concentration distribution label, a base concentration distribution label, and a polymer concentration distribution label.

[0104] Step 220: Call a photoresist post-baking prediction model to perform post-baking prediction on the photoacid concentration distribution sample to obtain a remaining photoacid concentration distribution prediction result, a base concentration distribution prediction result, and a polymer concentration distribution prediction result.

[0105] Step 230: performing post-baking chemical diffusion correction on the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result and the polymer concentration distribution prediction result respectively to obtain a remaining photoacid concentration distribution correction result, a base concentration distribution correction result and a polymer concentration distribution correction result;

[0106] Step 240: correcting the model parameters of the photoresist post-baking prediction model according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label, the base concentration distribution correction result and the base concentration distribution label, and the polymer concentration distribution correction result and the polymer concentration distribution label.

[0107] It should be noted that the model training method provided in the embodiments of the present application is a method for correcting the model parameters of the photoresist post-baking prediction model.

[0108] Figure 3 is a general structure diagram of the model training method according to the embodiments of the present application.

[0109] Figure 3 The leftmost in is the photoacid concentration distribution sample in step 210. The photoacid concentration distribution sample refers to the data sample collected about the photoacid concentration distribution in a certain experiment or research. The photoacid concentration distribution sample can be the photoacid concentration distribution collected through laboratory testing, sampling investigation or other means, used to describe the concentration distribution of photoacid under certain conditions. For example, it can be obtained from the historical photoacid concentration distribution in the historical post-baking process. In the present embodiment, the sample label corresponding to the photoacid concentration distribution sample can be the remaining photoacid concentration distribution obtained after post-baking treatment of the historical photoacid concentration distribution in the historical post-baking process. When the photoresist post-baking prediction model is trained using the historical photoacid concentration distribution as the training sample, the remaining photoacid concentration distribution obtained after post-baking treatment of the historical photoacid concentration distribution can be used as the sample label to realize the training of the photoresist post-baking prediction model.

[0110] Based on the photoacid concentration distribution sample, step 220 is performed. In an embodiment, referring to Figure 3 , the photoresist post-baking prediction model in step 220 can include three independent models, namely a first photoresist post-baking prediction model RFNO_A, a second photoresist post-baking prediction model RFNO_Q and a third photoresist post-baking prediction model RFNO_P. The first photoresist post-baking prediction model RFNO_A can be used to perform post-baking prediction on the photoacid concentration distribution sample to obtain a remaining photoacid concentration distribution prediction result A ML . The second photoresist post-baking prediction model RFNO_Q can be used to perform post-baking prediction on the photoacid concentration distribution sample to obtain a base concentration distribution prediction result Q MLThe third photoresist post-exposure prediction model RFNO P can be used to perform post-exposure prediction on the photoacid concentration distribution sample to obtain a polymer concentration distribution prediction result P ML Correspondingly, when the model parameters are corrected, the three models described above need to be corrected respectively.

[0111] Referring to Figure 4 In an embodiment, before the photoresist post-exposure prediction model is called to perform post-exposure prediction, the photoresist post-exposure prediction model can be pre-trained in the following manner: first, a photoacid concentration distribution sample A0 and a sample label corresponding to the photoacid concentration distribution sample are obtained, and the sample label includes a residual photoacid concentration distribution label A T , a base concentration distribution label Q T , and a polymer concentration distribution label P T ; second, the photoresist post-exposure prediction model is called to perform post-exposure prediction on the photoacid concentration distribution sample A0 to obtain a residual photoacid concentration distribution prediction result, a base concentration distribution prediction result, and a polymer concentration distribution prediction result. Specifically, the first photoresist post-exposure prediction model RFNO A is called to perform post-exposure prediction on the photoacid concentration distribution sample A0 to obtain the residual photoacid concentration distribution prediction result, the second photoresist post-exposure prediction model RFNO Q is called to perform post-exposure prediction on the photoacid concentration distribution sample A0 to obtain the base concentration distribution prediction result, and the third photoresist post-exposure prediction model RFNO P is called to perform post-exposure prediction on the photoacid concentration distribution sample A0 to obtain the polymer concentration distribution prediction result; then, the model parameters of the first photoresist post-exposure prediction model RFNO A are corrected according to the residual photoacid concentration distribution prediction result and the residual photoacid concentration distribution label A T , the model parameters of the second photoresist post-exposure prediction model RFNO Q are corrected according to the base concentration distribution prediction result and the base concentration distribution label Q T , and the model parameters of the third photoresist post-exposure prediction model RFNO P are corrected according to the polymer concentration distribution prediction result and the polymer concentration distribution label P T .

[0112] In another embodiment, the photoresist post-exposure prediction model can include three sub-models, namely a photoresist post-exposure first prediction sub-model, a photoresist post-exposure second prediction sub-model, and a photoresist post-exposure third prediction sub-model. The photoresist post-exposure first prediction sub-model can be used to perform post-exposure prediction on the photoacid concentration distribution sample to obtain a residual photoacid concentration distribution prediction result A ML . The photoresist post-exposure second prediction sub-model can be used to perform post-exposure prediction on the photoacid concentration distribution sample to obtain a base concentration distribution prediction result Q ML . The photoresist post-exposure third prediction sub-model can be used to perform post-exposure prediction on the photoacid concentration distribution sample to obtain a polymer concentration distribution prediction result P MLCorrespondingly, when the model parameters are corrected, the loss value can be fed back to the large photoresist post-baking prediction model, and then the three sub-models are corrected in the large photoresist post-baking prediction model.

[0113] After step 220 is performed, based on the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result and the polymer concentration distribution prediction result, step 230 is performed. The post-baking chemical diffusion correction in step 230 is a correction method of post-baking chemical diffusion of data / results. The post-baking chemical diffusion equation is a diffusion equation, and the post-baking chemical diffusion correction can be performed by using the post-baking chemical diffusion equation. Referring to Figure 3 The remaining photoacid concentration distribution prediction result A ML is corrected by post-baking chemical diffusion to obtain a remaining photoacid concentration distribution correction result. The base concentration distribution prediction result Q ML is corrected by post-baking chemical diffusion to obtain a base concentration distribution correction result. The polymer concentration distribution prediction result P ML is corrected by post-baking chemical diffusion to obtain a polymer concentration distribution correction result.

[0114] After step 230, step 240 is performed. Referring to Figure 3 In an embodiment, the prediction loss value FitLoss of the three concentration distributions can be calculated according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label, the base concentration distribution correction result and the base concentration distribution label, and the polymer concentration distribution correction result and the polymer concentration distribution label, and then the model parameters of the photoresist post-baking prediction model are corrected based on the prediction loss value FitLoss. In another embodiment, the correction loss value PEB Loss of the three concentration distributions can also be calculated according to the remaining photoacid concentration distribution prediction result and the remaining photoacid concentration distribution correction result, the base concentration distribution prediction result and the base concentration distribution correction result, and the polymer concentration distribution prediction result and the polymer concentration distribution correction result, and then the model parameters of the photoresist post-baking prediction model are corrected based on the prediction loss value FitLoss and the correction loss value PEB Loss. The specific calculation process of the prediction loss value FitLoss and the specific calculation process of the correction loss value PEB Loss will be described in detail below, which is omitted here.

[0115] The benefits of the above steps 210 to 240 are that the prediction accuracy of the trained model when performing photoresist modeling simulation can be improved, thereby the prediction accuracy of the trained model can be improved.

[0116] In an embodiment, referring to Figure 5 , the photoacid concentration distribution sample in step 210 can be obtained according to the following steps:

[0117] Step 510: Obtain an initial light intensity distribution and an initial photoacid concentration distribution in the photoresist of the wafer.

[0118] Step 520: Perform exposure simulation processing according to the initial light intensity distribution and the initial photoacid concentration distribution to obtain a photoacid concentration distribution after exposure.

[0119] Step 530: Calculate the difference between the photoacid concentration distribution after exposure and the initial photoacid concentration distribution to obtain a photoacid concentration distribution sample.

[0120] The embodiment has the benefit that, on the basis of being able to obtain a photoacid concentration distribution sample, the accuracy of the photoacid concentration distribution sample can be improved by using exposure simulation processing, and the applicability is relatively high.

[0121] For step 510, the photoresist of the wafer has been introduced above, and will not be described again here. The initial light intensity distribution refers to the light intensity distribution before exposure. The initial photoacid concentration distribution refers to the concentration distribution of the photoacid generated under the initial light intensity distribution.

[0122] The way to obtain the initial light intensity distribution can include: (1) using an optical microscope, a laser scanning microscope or other optical instruments to observe and obtain the light intensity distribution on the surface or in the body of the photoresist. (2) By measuring the spatial distribution and intensity of light intensity, the illumination of light in the photoresist (i.e. the light intensity distribution) can be obtained.

[0123] The way to obtain the initial photoacid concentration distribution can include: (1) using a fluorescent probe or a fluorescently labeled molecule to detect the diffusion process of the photoacid in the photoresist, and directly measuring the initial photoacid concentration distribution. (2) By observing and obtaining the intensity and distribution of the fluorescent signal through a fluorescence microscope or other optical techniques, the initial photoacid concentration distribution is indirectly measured.

[0124] For step 520, exposure simulation processing generally refers to using computer simulation and numerical methods to model and simulate the exposure step in the lithography process to predict and optimize the formation of the lithography pattern. The exposure model is a deep learning algorithm model capable of exposure simulation of input data. In an embodiment, the initial light intensity distribution and the initial photoacid concentration distribution can be input to the exposure model for exposure simulation processing to obtain the photoacid concentration distribution after exposure.

[0125] In an embodiment, step 520 can include:

[0126] The initial light intensity distribution and the initial photoacid concentration distribution are taken as input parameters to perform an exposure simulation calculation process, wherein the exposure simulation calculation process comprises: calculating a light intensity change rate with depth according to the initial light intensity distribution and the initial photoacid concentration distribution; calculating a next time light intensity distribution according to the light intensity change rate with depth and the initial light intensity distribution; calculating a photoacid concentration change rate with time according to the next time light intensity distribution and the initial photoacid concentration distribution; and calculating a next time photoacid concentration distribution according to the photoacid concentration change rate with time and the initial photoacid concentration distribution.

[0127] The next time light intensity distribution and the next time photoacid concentration distribution are taken as input parameters to repeat the exposure simulation calculation process until a photoacid concentration distribution reaching an exposure time is obtained.

[0128] The photoacid concentration distribution reaching the exposure time is taken as a post-exposure photoacid concentration distribution.

[0129] The above exposure simulation calculation process can be shown in the following formula:

[0130]

[0131]

[0132] G(t=0)=G0 Formula (3).

[0133] In the formula (1) to (3), represents the light intensity change rate with depth, x, y, z are three-dimensional coordinates, I(x, y, z, t) represents the light intensity distribution in the wafer photoresist at t time, and G(x, y, z, t) represents the photoacid concentration distribution at t time, represents the photoacid concentration change rate with time, and G0 represents the initial photoacid concentration distribution.C Dill is the Dill (Dill parameter) coefficient (cm 2 / mJ), and a is the light absorption coefficient, and the two coefficients need to be determined according to experiments.

[0134] In one example, the initial light intensity distribution and the initial photoacid concentration distribution G0 are input into equation (1) to calculate the rate of change of light intensity with depth at time t = 0. For time t, the rate of change of light intensity with depth at time t - 1 and the light intensity distribution at time t - 1 are used to calculate the light intensity distribution at time t. The light intensity distribution at time t and the photoacid concentration distribution at time t - 1 are input into equation (2) to calculate the rate of change of photoacid concentration with time at time t. The rate of change of photoacid concentration with time at time t and the photoacid concentration distribution at time t - 1 are used to calculate the photoacid concentration distribution at time t. The photoacid concentration distribution at the final time t_final (the time at which exposure ends) is taken as the post-exposure photoacid concentration distribution.

[0135] In one example, the process of performing exposure simulation (using finite difference method as an example) can include the following steps:

[0136] (1) Discretization: Discretize the space (x, y, z) and time (t) into a series of grid points and approximate the solution of partial differential equations on these grid points.

[0137] (2) Initialization: Set the initial condition G(t = 0) = G0 and set the initial light intensity distribution I(x, y, z, t = 0).

[0138] (3) Time iteration: For each time step Δt, perform the following steps:

[0139] Using the current light intensity distribution I(x, y, z, t) and the photoacid concentration distribution G(x, y, z, t), calculate the rate of change of light intensity with depth dI / dz by finite difference approximation.

[0140] Update the light intensity distribution I(x, y, z, t + Δt) using the rate of change of light intensity with depth dI / dz.

[0141] Using the current light intensity distribution I(x, y, z, t + Δt) and the photoacid concentration distribution G(x, y, z, t), calculate the rate of change of photoacid concentration with time dG / dt by finite difference approximation.

[0142] Update the photoacid concentration distribution G(x, y, z, t + Δt) using the rate of change of photoacid concentration with time dG / dt.

[0143] (5) Iteration to end: Repeat the time iteration step until the desired exposure time is reached.

[0144] (6) Result analysis: Analyze the final photoacid concentration distribution G(x, y, z, t_final), where t_final is the time at which exposure ends.

[0145] The embodiment has the benefits that the post-exposure photoacid concentration distribution is calculated by repeatedly performing the exposure simulation calculation process, and the rate of change of light intensity with depth and the rate of change of photoacid concentration with time are considered simultaneously in the calculation process, thereby improving the calculation accuracy.

[0146] In an embodiment, referring to Figure 6 The sample label in step 210 is obtained according to the following steps:

[0147] Step 610: Obtain an initial base concentration distribution and an initial polymer concentration distribution;

[0148] Step 620: Perform a post-baking chemical diffusion simulation process according to the photoacid concentration distribution sample, the initial base concentration distribution, and the initial polymer concentration distribution to obtain a residual photoacid concentration distribution simulation result, a base concentration distribution simulation result, and a polymer concentration distribution simulation result;

[0149] Step 630: Take the residual photoacid concentration distribution simulation result as the residual photoacid concentration distribution label;

[0150] Step 640: Take the base concentration distribution simulation result as the base concentration distribution label;

[0151] Step 650: Take the polymer concentration distribution simulation result as the polymer concentration distribution label.

[0152] The embodiment has the benefits that the post-baking chemical diffusion simulation process is used to realize automatic labeling, without manual labeling or other complex labeling methods, thereby improving the accuracy and acquisition efficiency of the sample label.

[0153] For step 610, the initial base concentration distribution refers to the base concentration distribution before the exposure simulation process. The base concentration distribution refers to the variation of the concentration of the base substance (usually the photoacid in the photoresist) in the photoresist with the spatial position. In the photoetching process, the photoacid in the photoresist will undergo a chemical reaction after exposure, causing local acid-base neutralization or acid-base reaction, thereby changing the properties of the photoresist, and finally forming the required pattern. The initial polymer concentration distribution refers to the polymer concentration distribution before the exposure simulation process. The polymer concentration distribution refers to the variation of the concentration of the polymer (one of the main components of the photoresist) in the photoresist with the spatial position. In the photoetching process, the polymer in the photoresist will undergo a chemical reaction after exposure, causing local polymerization or depolymerization, thereby changing the properties of the photoresist, and finally forming the required pattern.

[0154] For step 620, the post-bake chemical diffusion process is an important process step in the microelectronic device manufacturing process. During the post-bake chemical diffusion process, the exposed photoresist needs to be baked to remove the solvent and to solidify the photoresist. Meanwhile, the baking can also promote the diffusion of the photo-base, further change the properties of the photoresist, and finally achieve the desired pattern transfer. The post-bake chemical diffusion simulation process in step 620 is a simulation process for the post-bake chemical diffusion process. In an embodiment, the photo-acid concentration distribution sample, the initial base concentration distribution, and the initial polymer concentration distribution can be input to the post-bake chemical diffusion equation model for post-bake chemical diffusion simulation processing to obtain the remaining photo-acid concentration distribution simulation result, the base concentration distribution simulation result, and the polymer concentration distribution simulation result.

[0155] In an embodiment, step 620 can include:

[0156] inputting the photo-acid concentration distribution sample, the initial base concentration distribution, and the initial polymer concentration distribution as input parameters, and performing a post-bake chemical diffusion simulation calculation process, wherein the post-bake chemical diffusion simulation calculation process includes: calculating the photo-acid concentration distribution at the next time according to the photo-acid concentration distribution sample, the initial base concentration distribution, a preset photo-acid diffusion rate, and a preset acid-base neutralization rate; calculating the base concentration distribution at the next time according to the photo-acid concentration distribution sample, the initial base concentration distribution, a preset base diffusion rate, and the acid-base neutralization rate; and calculating the polymer concentration distribution at the next time according to the photo-acid concentration distribution sample, the initial polymer concentration distribution, and a preset polymer deprotection reaction rate;

[0157] repeating the post-bake chemical diffusion simulation calculation process with the photo-acid concentration distribution at the next time, the base concentration distribution at the next time, and the polymer concentration distribution at the next time as input parameters until the photo-acid concentration distribution at the post-bake time, the base concentration distribution at the post-bake time, and the polymer concentration distribution at the post-bake time are obtained;

[0158] the photo-acid concentration distribution at the post-bake time, the base concentration distribution at the post-bake time, and the polymer concentration distribution at the post-bake time correspond to the remaining photo-acid concentration distribution simulation result, the base concentration distribution simulation result, and the polymer concentration distribution simulation result.

[0159] In an embodiment, the post-bake chemical diffusion simulation calculation process described above can include the following three sub-processes:

[0160] (1) A sub-process of calculating the photoacid concentration distribution at the next time according to the photoacid concentration distribution sample, the initial base concentration distribution, the preset photoacid diffusion rate and the preset acid-base neutralization rate. The sub-process can include: first calculating the change rate of the photoacid concentration distribution with time according to the photoacid concentration distribution sample, the initial base concentration distribution, the preset photoacid diffusion rate and the preset acid-base neutralization rate; and then calculating the photoacid concentration distribution at the next time according to the change rate of the photoacid concentration distribution with time and the photoacid concentration distribution sample. (2) A sub-process of calculating the base concentration distribution at the next time according to the photoacid concentration distribution sample, the initial base concentration distribution, the preset base diffusion rate and the acid-base neutralization rate. The sub-process can include: first calculating the change rate of the base concentration distribution with time according to the photoacid concentration distribution sample, the initial base concentration distribution, the preset base diffusion rate and the acid-base neutralization rate; and then calculating the base concentration distribution at the next time according to the change rate of the base concentration distribution with time and the initial base concentration distribution. (3) A sub-process of calculating the polymer concentration distribution at the next time according to the photoacid concentration distribution sample, the initial polymer concentration distribution and the preset polymer deprotection reaction rate. The sub-process can include: calculating the change rate of the polymer concentration distribution with time according to the photoacid concentration distribution sample, the initial polymer concentration distribution and the preset polymer deprotection reaction rate; and then calculating the polymer concentration distribution at the next time according to the change rate of the polymer concentration distribution with time and the initial polymer concentration distribution.

[0161] In an embodiment, the above post-baking chemical diffusion simulation calculation process can be shown in the following formula:

[0162]

[0163]

[0164]

[0165] A(t=0)=A0=G0-G(t=t exposure ) Formula (7).

[0166] Q(t=0)=Q0 Formula (8).

[0167] P(t=0)=P0 Formula (9).

[0168] In formula (4) to formula (9), represents the change rate of the photoacid concentration distribution with time; represents the change rate of the base concentration distribution with time; represents the rate of change of the polymer concentration profile with time; A represents the photoacid concentration profile at time t; Q represents the base concentration profile at time t; P represents the polymer concentration profile at time t; A0represents a photoacid concentration profile sample, obtained by taking the difference between the photoacid concentration G(t = t exposure ) after exposure and the initial photoacid concentration profile G0; Q0and P0represent the initial base concentration profile and the initial polymer concentration profile, respectively; D A and D Q represent the photoacid diffusion rate and the base diffusion rate, respectively (both in nm 2 / s); k n and k P represent the acid-base neutralization rate and the polymer deprotection reaction rate, respectively. The four parameters D A , D Q , k n , and k P need to be determined experimentally. represents the spatial second derivative, which can be calculated by numerical differentiation. By inputting the photoacid concentration profile sample, the initial base concentration profile, and the initial polymer concentration profile into the post-exposure chemical diffusion equation model, the remaining photoacid concentration profile simulation result, the base concentration profile simulation result, and the polymer concentration profile simulation result can be calculated. In an embodiment, the post-exposure chemical diffusion equation can be solved using a finite difference algorithm, thereby obtaining the photoacid concentration profile at the post-exposure time, the base concentration profile at the post-exposure time, and the polymer concentration profile at the post-exposure time.

[0169] In an example, for t = 1, the process of calculating the photoacid concentration profile can include: inputting the photoacid concentration profile sample A0, the initial base concentration profile Q0, the photoacid diffusion rate D A , and the acid-base neutralization rate k n into equation (4) to calculate the rate of change of the photoacid concentration profile with time at t = 1, and calculating the photoacid concentration profile at t = 1 according to the rate of change of the photoacid concentration profile with time and the photoacid concentration profile sample A0. For t = 1, the process of calculating the base concentration profile can include: inputting the photoacid concentration profile sample A0, the initial base concentration profile Q0, the base diffusion rate D Q , and the acid-base neutralization rate k n into equation (5) to calculate the rate of change of the base concentration profile with time at t = 1, and calculating the base concentration profile at t = 1 according to the rate of change of the base concentration profile with time at t = 1 and the initial base concentration profile Q0. For t = 1, the process of calculating the polymer concentration profile can include: inputting the photoacid concentration profile sample A0, the initial polymer concentration profile P0, and the polymer deprotection reaction rate k PThe rate of change of the polymer concentration distribution at time t = 1 is calculated by inputting equation (6); and the polymer concentration distribution at time t = 1 is calculated based on the rate of change of the polymer concentration distribution at time t = 1 and the initial polymer concentration distribution P0. For time t (t > 1), the process of calculating the photoacid concentration distribution can include: based on the photoacid concentration distribution at time t - 1, the base concentration distribution at time t - 1, the photoacid diffusion rate D A and the acid-base neutralization rate k n The rate of change of the photoacid concentration distribution at time t is calculated by inputting equation (4); and the photoacid concentration distribution at time t is calculated based on the rate of change of the photoacid concentration distribution at time t and the photoacid concentration distribution at time t - 1. For time t (t > 1), the process of calculating the base concentration distribution can include: based on the photoacid concentration distribution at time t - 1, the base concentration distribution at time t - 1, the base diffusion rate D Q and the acid-base neutralization rate k n The rate of change of the base concentration distribution at time t is calculated by inputting equation (5); and the base concentration distribution at time t is calculated based on the rate of change of the base concentration distribution at time t and the base concentration distribution at time t - 1. For time t (t > 1), the process of calculating the polymer concentration distribution can include: based on the photoacid concentration distribution sample at time t - 1, the polymer concentration distribution at time t - 1, and the polymer deprotection reaction rate k P The rate of change of the polymer concentration distribution at time t is calculated by inputting equation (6); and the polymer concentration distribution at time t is calculated based on the rate of change of the polymer concentration distribution at time t and the polymer concentration distribution at time t - 1. If t reaches the post-exposure baking time, the photoacid concentration distribution at the post-exposure baking time is taken as the remaining photoacid concentration distribution simulation result, the base concentration distribution at the post-exposure baking time is taken as the base concentration distribution simulation result, and the polymer concentration distribution at the post-exposure baking time is taken as the polymer concentration distribution simulation result.

[0170] In an embodiment, the post-exposure baking chemical diffusion equation can be solved by using a finite difference algorithm, and specifically, the above equation (4), equation (5), and equation (6) are converted into the following finite difference form:

[0171]

[0172]

[0173] P t+1 = P t - Δtk p A t P t Equation (12).

[0174] In the formula (10) to the formula (12), At represents a time step, A t , Q t and P t respectively represent the photoacid concentration distribution, the base concentration distribution and the polymer concentration distribution at t moment, A t+1 , Q t+1 and P t+1 respectively represent the photoacid concentration distribution, the base concentration distribution and the polymer concentration distribution at t+1 moment, and then the concentration distribution at the corresponding moment can be obtained by iterative calculation.

[0175] The above iterative solution process can be equivalent to the gradient descent algorithm, which is derived as follows:

[0176]

[0177]

[0178]

[0179] In the formula (13) to the formula (15), and represent the effective gradient of the three post-baking chemical diffusion equations of the formula (10) to the formula (12). As can be seen, by using the formula (13) to the formula (15), the photoacid concentration distribution, the base concentration distribution and the polymer concentration distribution at the next moment can be solved, so that the remaining photoacid concentration distribution simulation result, the base concentration distribution simulation result and the polymer concentration distribution simulation result can be obtained after the post-baking time is reached.

[0180] After step 620, by performing steps 630 to 650, the remaining photoacid concentration distribution simulation result can be taken as the remaining photoacid concentration distribution label, the base concentration distribution simulation result can be taken as the base concentration distribution label, and the polymer concentration distribution simulation result can be taken as the polymer concentration distribution label.

[0181] In step 220, the photoresist post-baking prediction model needs to be called to perform post-baking prediction on the photoacid concentration distribution sample. The photoresist post-baking prediction model can be realized by using a reduced Fourier operator neural network framework, which is mainly composed of two parts. The first part is a local feature extractor, and the second part is a global feature extractor. The local feature extractor is a kind of neural network structure. For example, the local feature extractor includes multiple cascaded convolutional neural network layers. The global feature extractor is also a kind of neural network structure. For example, the global feature extractor includes three linear transformation layers and one reduced Fourier operator. In an example, the sample local feature and the sample global feature of the photoacid concentration distribution sample are extracted first, and then post-baking prediction is performed based on the sample local feature and the sample global feature, which improves the accuracy of post-baking prediction.

[0182] In an embodiment, the step 220 comprises: performing local feature extraction on the photoacid concentration distribution sample to obtain a sample local feature; performing dimension expansion on the photoacid concentration distribution sample to obtain a first feature; performing Fourier transform on the first feature to obtain a second feature; performing complex convolution on the second feature and the model weight parameters of the photoresist post-bake prediction model to obtain a third feature; performing inverse Fourier transform on the third feature to obtain a fourth feature; performing dimension reduction on the fourth feature to obtain a sample global feature; and performing post-bake prediction based on the sample global feature and the sample local feature to obtain a post-bake prediction result, wherein the post-bake prediction result comprises one of a residual photoacid concentration distribution prediction result, an alkali concentration distribution prediction result, or a polymer concentration distribution prediction result.

[0183] In an embodiment, the local feature extractor can be used to perform local feature extraction on the photoacid concentration distribution sample. For example, the local feature extractor comprises four convolutional neural network layers, which are respectively: a three-dimensional convolutional layer comprising 25 1x1 filters, a three-dimensional convolutional layer comprising 1 3x3 filter, an inflation three-dimensional convolutional layer comprising 1 3x3 filter, and a two-dimensional convolutional layer comprising 1 5x5 filter. The activation function after each convolutional layer can use Leaky-ReLU. For the photoacid concentration distribution sample with a dimension number of 70x800x800, the local feature extractor can output a sample local feature with a dimension number of 70x800x800.

[0184] The dimension expansion refers to an operation of expanding the feature dimension of the photoacid concentration distribution sample. The photoacid concentration distribution sample can be input to a linear transformation layer for dimension expansion to obtain a first feature. The dimension number of the first feature is greater than the dimension number of the sample local feature. The Fourier transform is a method of converting a signal or an image from a time domain or a space domain to a frequency domain. The Fourier transform operator can be used to perform Fourier transform on the first feature to obtain a second feature. The complex convolution refers to a convolution operation in the complex domain. The complex convolution operator can be used to perform complex convolution on the second feature and the model weight parameters of the photoresist post-bake prediction model to obtain a third feature. The inverse Fourier transform is a method of converting a signal or an image from a frequency domain to a time domain or a space domain. The inverse Fourier transform operator can be used to perform inverse Fourier transform on the third feature to obtain a fourth feature. The dimension reduction refers to an operation of reducing the dimension of the fourth feature. The fourth feature can be input to a linear transformation layer for dimension reduction to obtain a sample global feature. The dimension number of the sample global feature is consistent with the dimension number of the sample local feature.

[0185] In an example, the fourth feature can be obtained by the following formula:

[0186]

[0187] In the formula (16), V in represents the first feature, F(·) represents a Fourier transform operation, represents a complex convolution operation, F -1 (·) represents an inverse Fourier transform operation. W P represents a model parameter of the local feature extractor, W R represents a model parameter of the global feature extractor, W P ∈ C CI×CI×1×1 , where C represents a complex number, CI represents a channel number, for example, CI takes 25; W R ∈ C CI ×CI×k×k , k represents a dimension truncation, for example, k takes 50. σ represents an activation function, for example, Leaky-ReLU is used as the activation function. Assuming that the dimension number of the fourth feature is 25×70×800×800, it is input to another two linear transformation layers to reduce the channel number, and a sample global feature with a dimension number of 70×800×800 can be output. Finally, the sample local feature and the sample global feature are added to obtain a post-baking prediction result, wherein the post-baking prediction result includes one of a residual photoacid concentration distribution prediction result, a base concentration distribution prediction result, or a polymer concentration distribution prediction result.

[0188] The above embodiment has the benefit that the post-baking prediction is realized by the fusion of multiple operations such as local feature extraction, Fourier transform processing, complex convolution processing, inverse Fourier transform processing, and dimension reduction processing, thereby improving the accuracy of the post-baking prediction.

[0189] In an embodiment, with reference to Figure 7 , step 230 includes:

[0190] Step 710: taking the residual photoacid concentration distribution prediction result, the base concentration distribution prediction result, and the polymer concentration distribution prediction result as input parameters, performing a post-baking chemical diffusion correction calculation process, wherein the post-baking chemical diffusion correction calculation process includes: calculating a photoacid concentration distribution of a first iteration according to the residual photoacid concentration distribution prediction result, the base concentration distribution prediction result, a preset photoacid diffusion rate, and a preset acid-base neutralization rate; calculating a base concentration distribution of the first iteration according to the residual photoacid concentration distribution prediction result, the base concentration distribution prediction result, a preset base diffusion rate, and the acid-base neutralization rate; and calculating a polymer concentration distribution of the first iteration according to the residual photoacid concentration distribution prediction result, the polymer concentration distribution prediction result, and a preset polymer deprotection reaction rate.

[0191] Step 720: Repeating the post-baking chemical diffusion correction calculation process with the first iteration of the photoacid concentration distribution, the first iteration of the base concentration distribution, and the first iteration of the polymer concentration distribution as input parameters until the photoacid concentration distribution at the iteration number, the base concentration distribution at the iteration number, and the polymer concentration distribution at the iteration number are obtained;

[0192] Step 730: Corresponding the photoacid concentration distribution at the iteration number, the base concentration distribution at the iteration number, and the polymer concentration distribution at the iteration number as the remaining photoacid concentration distribution correction result, the base concentration distribution correction result, and the polymer concentration distribution correction result.

[0193] For step 710, in an embodiment, the post-baking chemical diffusion correction calculation process can include the following three sub-processes:

[0194] (1) A sub-process for calculating the first iteration of the photoacid concentration distribution according to the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, the preset photoacid diffusion rate, and the preset acid-base neutralization rate. The sub-process can include: first calculating the rate of change of the photoacid concentration distribution over time according to the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, the preset photoacid diffusion rate, and the preset acid-base neutralization rate; and then calculating the first iteration of the photoacid concentration distribution according to the rate of change of the photoacid concentration distribution over time and the remaining photoacid concentration distribution prediction result.(2) A sub-process for calculating the first iteration of the base concentration distribution according to the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, the preset base diffusion rate, and the acid-base neutralization rate. The sub-process can include: first calculating the rate of change of the base concentration distribution over time according to the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, the preset base diffusion rate, and the acid-base neutralization rate; and then calculating the first iteration of the base concentration distribution according to the rate of change of the base concentration distribution over time and the base concentration distribution prediction result.(3) A sub-process for calculating the first iteration of the polymer concentration distribution according to the remaining photoacid concentration distribution prediction result, the polymer concentration distribution prediction result, and the preset polymer deprotection reaction rate. The sub-process can include: calculating the rate of change of the polymer concentration distribution over time according to the remaining photoacid concentration distribution prediction result, the polymer concentration distribution prediction result, and the preset polymer deprotection reaction rate; and then calculating the first iteration of the polymer concentration distribution according to the rate of change of the polymer concentration distribution over time and the polymer concentration distribution prediction result.

[0195] For step 720, the post-bake chemical diffusion correction calculation process in step 720 is basically the same as step 710, but the input parameters are different. The post-bake chemical diffusion correction calculation process can be repeatedly executed until the photoacid concentration distribution reaches the iteration number, the base concentration distribution reaches the iteration number, and the polymer concentration distribution reaches the iteration number.

[0196] It can be understood that the post-bake chemical diffusion correction calculation process is basically the same as the post-bake chemical diffusion simulation calculation process in the above, so the post-bake chemical diffusion correction calculation process can also be represented by formula (4) to formula (6).

[0197] In an example, for d = 1 iteration, the process of calculating the photoacid concentration distribution can include: inputting the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, the photoacid diffusion rate D A and the acid-base neutralization rate k n into formula (4) to calculate the rate of change of the photoacid concentration distribution with time for d = 1 iteration, and calculating the photoacid concentration distribution for d = 1 iteration according to the rate of change of the photoacid concentration distribution with time for d = 1 iteration and the remaining photoacid concentration distribution prediction result. For d = 1 iteration, the process of calculating the base concentration distribution can include: inputting the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, the base diffusion rate D Q and the acid-base neutralization rate k n into formula (5) to calculate the rate of change of the base concentration distribution with time for d = 1 iteration, and calculating the base concentration distribution for d = 1 iteration according to the rate of change of the base concentration distribution with time for d = 1 iteration and the base concentration distribution prediction result. For d = 1 iteration, the process of calculating the polymer concentration can include: inputting the remaining photoacid concentration distribution prediction result, the polymer concentration distribution prediction result, and the polymer deprotection reaction rate k P into formula (6) to calculate the rate of change of the polymer concentration distribution with time for d = 1 iteration, and calculating the polymer concentration distribution for d = 1 iteration according to the rate of change of the polymer concentration distribution with time for d = 1 iteration and the polymer concentration distribution prediction result. For d (d > 1) iteration, the process of calculating the remaining photoacid concentration distribution can include: inputting the photoacid concentration distribution for d-1 iteration, the base concentration distribution for d-1 iteration, the photoacid diffusion rate D A and the acid-base neutralization rate k nThe input formula (4) is calculated to obtain the rate of change of the photoacid concentration distribution with time for the dth iteration, and the photoacid concentration distribution for the dth iteration is calculated according to the rate of change of the photoacid concentration distribution with time for the dth iteration and the photoacid concentration distribution for the (d-1)th iteration. For the d (d>1)th iteration, the process of calculating the base concentration distribution can include: inputting the photoacid concentration distribution for the (d-1)th iteration, the base concentration distribution for the (d-1)th iteration, the base diffusion rate D Q and the acid-base neutralization rate k n The input formula (5) is calculated to obtain the rate of change of the base concentration distribution with time for the dth iteration, and the base concentration distribution for the dth iteration is calculated according to the rate of change of the base concentration distribution with time for the dth iteration and the base concentration distribution for the (d-1)th iteration. For the d (d>1)th iteration, the process of calculating the polymer concentration can include: inputting the photoacid concentration distribution sample for the (d-1)th iteration, the polymer concentration distribution for the (d-1)th iteration, and the polymer deprotection reaction rate k P The input formula (6) is calculated to obtain the rate of change of the polymer concentration distribution with time for the dth iteration, and the polymer concentration distribution for the dth iteration is calculated according to the rate of change of the polymer concentration distribution with time for the dth iteration and the polymer concentration distribution for the dth iteration. If d reaches the number of iterations, the photoacid concentration distribution for the number of iterations is taken as the remaining photoacid concentration distribution correction result, the base concentration distribution for the number of iterations is taken as the base concentration distribution correction result, and the polymer concentration distribution for the number of iterations is taken as the polymer concentration distribution correction result.

[0198] The solution method of the post-baking chemical diffusion correction calculation process is similar to the solution method of the post-baking chemical diffusion simulation calculation process described above, and will not be described again here.

[0199] The benefits of the above steps 710 to 730 are that by performing post-baking chemical diffusion correction on the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, and the polymer concentration distribution prediction result, the corrected remaining photoacid concentration distribution correction result, the base concentration distribution correction result, and the polymer concentration distribution correction result are more accurate, which is beneficial to improve the imaging fidelity.

[0200] In step 240, the model parameters of the post-baking photoresist prediction model need to be corrected according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label, the base concentration distribution correction result and the base concentration distribution label, and the polymer concentration distribution correction result and the polymer concentration distribution label.

[0201] In an embodiment, referring to Figure 8 , step 240 includes:

[0202] Step 810: Calculate the photoacid concentration distribution loss value according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label;

[0203] Step 820: Calculate the base concentration distribution loss value according to the base concentration distribution correction result and the base concentration distribution label;

[0204] Step 830: Calculate the polymer concentration distribution loss value according to the polymer concentration distribution correction result and the polymer concentration distribution label;

[0205] Step 840: Correct the model parameters of the photoresist post-exposure prediction model according to the photoacid concentration distribution loss value, the base concentration distribution loss value, and the polymer concentration distribution loss value.

[0206] For step 810, in an embodiment, the process of calculating the photoacid concentration distribution loss value can include: calculating the square of the difference between the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label to obtain a first photoacid concentration distribution loss term; and taking the first photoacid concentration distribution loss term as the photoacid concentration distribution loss value. For example, the first photoacid concentration distribution loss term L fit-A ML T 2 , where PEB(A ML ) represents the remaining photoacid concentration distribution correction result, and A T represents the remaining photoacid concentration distribution label.

[0207] In another embodiment, the process of calculating the photoacid concentration distribution loss value can include: calculating the square of the difference between the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label to obtain a first photoacid concentration distribution loss term (i.e., the prediction loss value Fit Loss of the photoacid concentration distribution described above); and calculating a second photoacid concentration distribution loss term (i.e., the correction loss value PEB Loss of the photoacid concentration distribution described above) according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution prediction result; and then calculating the photoacid concentration distribution loss value according to the first photoacid concentration distribution loss term and the second photoacid concentration distribution loss term. The second photoacid concentration distribution loss term is the loss value of the post-exposure chemical diffusion correction calculation process, which can be represented by L PEB-A .

[0208] In an example, the process of calculating the photoacid concentration distribution loss value can be shown in the following formula:

[0209] L A = L fit-A + βL PEB-A = ||PEB(A ML )-A​​​T || 2 +βL PEB-A Equation (17).

[0210] In Equation (17), L A represents the photoacid concentration distribution loss value, L fit-A represents the first photoacid concentration distribution loss term, L PEB-A represents the second photoacid concentration distribution loss term, PEB(A ML ) represents the remaining photoacid concentration distribution correction result, A T represents the remaining photoacid concentration distribution label, and β is an adjustable parameter. The larger β is, the more the impact of the post-bake process physical information needs to be considered in the model training. The calculation process of the second photoacid concentration distribution loss term L PEB-A will be described in detail below, which is omitted here.

[0211] The above embodiment has the benefit that, on the basis of being able to calculate the photoacid concentration distribution loss value, the photoacid concentration distribution loss value can reflect the post-bake process physical information, so that the accuracy of the trained model is higher.

[0212] For step 820, in an embodiment, the process of calculating the base concentration distribution loss value can include: calculating the square of the difference between the base concentration distribution correction result and the base concentration distribution label to obtain a first base concentration distribution loss term; and taking the first base concentration distribution loss term as the base concentration distribution loss value. For example, the first base concentration distribution loss term L fit-Q = ||PEB(Q ML )-Q T || 2 , where PEB(Q ML ) represents the remaining photoacid concentration distribution correction result, and Q T represents the remaining photoacid concentration distribution label.

[0213] In another embodiment, the process of calculating the base concentration distribution loss value can include: calculating the square of the difference between the base concentration distribution correction result and the base concentration distribution label to obtain a first base concentration distribution loss term (i.e., the prediction loss value Fit Loss of the base concentration distribution described above); and calculating a second base concentration distribution loss term (i.e., the correction loss value PEB Loss of the base concentration distribution described above) according to the base concentration distribution correction result and the base concentration distribution prediction result; and then calculating the base concentration distribution loss value according to the first base concentration distribution loss term and the second base concentration distribution loss term. The second base concentration distribution loss term is the loss value of the post-bake chemical diffusion correction calculation process, which can be represented by L PEB-Q .

[0214] In an example, the process of calculating the base concentration distribution loss value can be shown in the following formula:

[0215] L Q =L fit-Q +βL PEB-Q =||PEB(Q ML )-Q T || 2 +βL PEB-Q Equation (18).

[0216] In Equation (18), L Q represents the base concentration distribution loss value, L fit-Q represents the first base concentration distribution loss term, L PEB-Q represents the second base concentration distribution loss term, PEB(Q ML ) represents the base concentration distribution correction result, Q T represents the base concentration distribution label, and β is an adjustable parameter. The larger β is, the more the impact of the post-baking process physical information needs to be considered in the model training. The calculation process of the second base concentration distribution loss term L PEB-Q will be described in detail below, which is omitted here.

[0217] The above embodiment has the benefit that, on the basis of being able to calculate the base concentration distribution loss value, the base concentration distribution loss value can reflect the post-baking process physical information, so that the accuracy of the trained model is higher.

[0218] For step 830, the process of calculating the polymer concentration distribution loss value can include: calculating the square of the difference between the polymer concentration distribution correction result and the polymer concentration distribution label to obtain the first polymer concentration distribution loss term; and taking the first polymer concentration distribution loss term as the polymer concentration distribution loss value. For example, the first polymer concentration distribution loss term L fit-P ||PEB(P ML )-P T || 2 , wherein PEB(P ML ) represents the polymer concentration distribution correction result, and P T represents the polymer concentration distribution label.

[0219] In another embodiment, the process of calculating the polymer concentration distribution loss value can include: calculating the square of the difference between the polymer concentration distribution correction result and the polymer concentration distribution label to obtain a first polymer concentration distribution loss term (i.e., the aforementioned prediction loss value Fit Loss of the polymer concentration distribution); and calculating a second polymer concentration distribution loss term (i.e., the aforementioned correction loss value PEB Loss of the polymer concentration distribution) according to the polymer concentration distribution correction result and the polymer concentration distribution prediction result; and then calculating the polymer concentration distribution loss value according to the first polymer concentration distribution loss term and the second polymer concentration distribution loss term. The second polymer concentration distribution loss term is the loss value of the post-baking chemical diffusion correction calculation process, which can be represented by L PEB-P .

[0220] In an example, the process of calculating the polymer concentration distribution loss value can be shown in the following formula:

[0221] L P = L fit-P + βL PEB-P = ||PEB(P ML ) - P T || 2 + βL PEB-P Equation (19).

[0222] In Equation (19), L P represents the polymer concentration distribution loss value, L fit-P represents the first polymer concentration distribution loss term, L PEB-P represents the second polymer concentration distribution loss term, PEB(Q ML ) represents the polymer concentration distribution correction result, P T represents the polymer concentration distribution label, and β is an adjustable parameter. The greater the β, the more the post-baking process physical information needs to be considered in the model training. The calculation process of the second polymer concentration distribution loss term L PEB-P will be described in detail below, which is omitted here.

[0223] The above-mentioned embodiments have the benefit that the polymer concentration distribution loss value can reflect the post-baking process physical information on the basis of being able to calculate the polymer concentration distribution loss value, so that the accuracy of the trained model is higher.

[0224] It should be noted that, Figure 3 the prediction loss value Fit Loss shown in Equation (18) includes: the first photoacid concentration distribution loss term L fit-A , the first base concentration distribution loss term L fit-Q , and the first polymer concentration distribution loss term L fit-P .Figure 3 The modified loss value PEB Loss shown in the middle includes: a second photoacid concentration distribution loss term L PEB-A , a second base concentration distribution loss term L PEB-Q , and a second polymer concentration distribution loss term L PEB-P .

[0225] For step 840, in an embodiment, the number of photoresist post exposure bake prediction models can be multiple, including a first photoresist post exposure bake prediction model RFNO_A, a second photoresist post exposure bake prediction model RFNO_Q, and a third photoresist post exposure bake prediction model RFNO_P. Then step 240 can include: modifying the model parameters of the first photoresist post exposure bake prediction model RFNO_A according to the photoacid concentration distribution loss value; modifying the model parameters of the second photoresist post exposure bake prediction model RFNO_Q according to the base concentration distribution loss value; and modifying the model parameters of the third photoresist post exposure bake prediction model RFNO_P according to the polymer concentration distribution loss value.

[0226] In another embodiment, the number of photoresist post exposure bake prediction models is one, but it can include a photoresist post exposure bake first prediction sub-model, a photoresist post exposure bake second prediction sub-model, and a photoresist post exposure bake third prediction sub-model. Then step 240 can include: calculating a total distribution loss value according to the photoacid concentration distribution loss value, the base concentration distribution loss value, and the polymer concentration distribution loss value; and jointly modifying the model parameters of the photoresist post exposure bake first prediction sub-model, the model parameters of the photoresist post exposure bake second prediction sub-model, and the model parameters of the photoresist post exposure bake third prediction sub-model according to the total distribution loss value.

[0227] In an embodiment, in the process of modifying the model parameters of the photoresist post exposure bake prediction model, the gradients of the photoacid concentration distribution loss value, the base concentration distribution loss value, and the polymer concentration distribution loss value are defined as follows:

[0228]

[0229]

[0230]

[0231] In formulas (20) to (22), w A represents the model parameters of the first photoresist post exposure bake prediction model RFNO_A, w Q represents the model parameters of the second photoresist post exposure bake prediction model RFNO_Q, and w P represents the model parameters of the third photoresist post exposure bake prediction model RFNO_P. and The gradient of the post-baking chemical diffusion equation is obtained through automatic differential calculation. Therefore, by correcting the model parameters using the above chain rule, the physical information of the post-baking process can be implicitly contained in the trained model parameters.

[0232] Furthermore, according to formula (20), The second photoacid concentration distribution loss term L PEB-A For A ML Find the derivative of A, where A ML Q represents the predicted distribution of residual photoacid concentration. ML This represents the predicted alkali concentration distribution. Because... It is the term on the right-hand side of the equals sign in the above formula (13). Therefore, by combining formula (13) and formula (20), the second photoacid concentration distribution loss term L can be calculated. PEB-A .

[0233] Furthermore, according to formula (21), The second alkali concentration distribution loss term L PEB-Q For Q ML Find the derivative of A, where A ML Q represents the predicted distribution of residual photoacid concentration. ML This represents the predicted alkali concentration distribution. Because... It is the term on the right-hand side of the equals sign in the above formula (14). Therefore, by combining formula (14) and formula (21), the second alkali concentration distribution loss term L can be calculated. PEB-Q .

[0234] Similarly, according to formula (22), (k P A ML P ML ) is the second polymer concentration distribution loss term L PEB-P For P ML Find the derivative of A, where A ML P represents the predicted distribution of residual photoacid concentration. ML This represents the predicted polymer concentration distribution. Since (k P A ML P ML The term L is the right-hand side of the equation in formula (15). Therefore, by combining formula (15) and formula (22), the second polymer concentration distribution loss term L can be calculated. PEB-P .

[0235] The advantage of the embodiments of steps 810 to 840 above is that the model parameters in the trained photoresist post-bake prediction model implicitly contain the physical information of the post-bake process, thereby improving the accuracy of model prediction.

[0236] After the post-exposure bake prediction model of the photoresist is trained through the above embodiments, in an embodiment, the target photoacid concentration distribution can be obtained first; secondly, the post-exposure bake prediction of the target photoacid concentration distribution is performed by using the trained post-exposure bake prediction model of the photoresist to obtain the target residual photoacid concentration distribution prediction result, the target base concentration distribution prediction result and the target polymer concentration distribution prediction result; then the development model is used to perform the development conversion on the target polymer concentration distribution prediction result to obtain the development rate distribution; finally, the development rate distribution is converted into the development time profile, and the final pattern profile is obtained.

[0237] The target photoacid concentration distribution refers to the data sample about the photoacid concentration collected in a certain experiment or research. The target photoacid concentration distribution is similar to the photoacid concentration distribution sample in the above, except that one is used in the model use stage and the other is used in the model training stage. The acquisition method of the target photoacid concentration distribution can refer to the acquisition method of the photoacid concentration distribution sample in the above, which will not be described here. The specific process of the post-exposure bake prediction of the trained post-exposure bake prediction model of the photoresist can refer to the process of the post-exposure bake prediction of the photoresist before the parameter correction in the above, which will not be described here. The difference is that the model parameters of the trained post-exposure bake prediction model of the photoresist implicitly contain the physical information of the post-exposure bake process, so that the target residual photoacid concentration distribution prediction result, the target base concentration distribution prediction result and the target polymer concentration distribution prediction result obtained by prediction have higher accuracy.

[0238] In an embodiment, the development model can be shown in the following formula:

[0239]

[0240]

[0241] In formula (23) to formula (24), R(x, y, z) represents the development rate distribution corresponding to the grid point with coordinates (x, y, z), R max represents the maximum development rate, R min represents the minimum development rate, n represents the development factor, m th represents the development threshold, and P represents the polymer concentration distribution (i.e. the target polymer concentration distribution prediction result). By inputting the target polymer concentration distribution prediction result into the development model for calculation, the development rate distribution can be obtained. The target polymer concentration distribution prediction result obtained above can be converted into the development rate distribution by using the development model, which has high accuracy.

[0242] To verify the effectiveness of the algorithm provided in this application (i.e., the model training method described above) and its application to the photolithography mask design of chip layouts, widely used public photolithography mask datasets were selected, such as GAN-OPC: Mask optimization with lithography-guided generative adversarialnets.DOI and Bentian Jiang, et al. 2020. Neural-ILT: Migrating ILT to neural networks for mask printability and complexity co-optimization. These two datasets contain a total of 10271 chip layouts and corresponding masks, with the chip layouts meeting the generation requirements of a 32nm process node and certain design rules. The light intensity distribution corresponding to each mask was calculated using the Hopkins optical model, employing a 193nm wavelength ring light source.

[0243] like Figure 9 As shown, Figure 9 This paper illustrates the error (Development Mean Square Error) of the polymer concentration distribution predicted by the algorithm of this application and other algorithms under different measurement indices (Gauge Index), corresponding to the development rate distribution and the results of the finite difference algorithm. For example, when the measurement index is 10, the algorithm is sorted from largest to smallest Development MSE as follows: CGAN (Conditional Generative Adversarial Network), DeePEB (Deep Post-Bake), FNO (Fourier Operator Network), RFNO (Reduced Fourier Operator Network), and PEBINN (the algorithm of this application). Similarly, when the measurement index is 20, the algorithm is sorted from largest to smallest Development MSE as follows: CGAN, DeePEB, RFNO, FNO, and PEBINN (the algorithm of this application). It is evident that regardless of the measurement index, the algorithm of this application has the smallest root mean square error, meaning that the algorithm of this application has the best accuracy.

[0244] Figure 10 The training time of the algorithm in this application embodiment and other algorithms is shown. For example... Figure 10 As shown, the training time of the algorithm in this embodiment is basically the same as that of other algorithms. Figure 10 In this context, FDM(gpu) and FDM(cpu) represent the execution time of the finite difference algorithm on a V100 CPU machine and a 32-core CPU machine, respectively. Figure 11The inference time of the algorithm of the embodiment of the application and other algorithms is shown. As shown in Figure 11 The inference time of the algorithm of the embodiment of the application is basically consistent with the inference time of other algorithms. Figure 11 FDM(gpu) and FDM(cpu) in the table respectively represent the running time of running the finite difference algorithm on the CPU machine of V100 and the 32-core CPU machine. It needs to be particularly pointed out that, compared with the finite difference algorithm on the same GPU machine, the inference time of the algorithm of the embodiment of the application is shortened by about 1 / 15, and the calculation efficiency is greatly improved.

[0245] The algorithm (a deep learning mask optimization algorithm embedding process parameters and lithography physical model information) provided by the embodiment of the application proposes to use a deep learning algorithm embedding post-baking physical information to simulate photoresist, design the post-baking chemical diffusion equation correction layer as a layer of deep neural network, and continuously correct the photoresist polymer concentration distribution generated by the deep learning network, so that the post-baking process physical information is embedded in the deep learning model. Compared with other mainstream deep learning visual algorithms such as Fourier operator neural network in the related art, the algorithm provided by the embodiment of the application has a significant improvement in precision while keeping the training time and inference time basically consistent. Compared with the finite difference algorithm, the calculation efficiency of the algorithm provided by the embodiment of the application can be improved by more than ten times.

[0246] With reference to Figure 12 The embodiment of the application also discloses a model training device. The model training device 1200 can realize the model training method in the foregoing embodiment, and the model training device 1200 comprises:

[0247] A sample acquisition unit 1210 is configured to acquire photoacid concentration distribution samples and sample labels corresponding to the photoacid concentration distribution samples, and the sample labels comprise residual photoacid concentration distribution labels, alkali concentration distribution labels and polymer concentration distribution labels.

[0248] A post-baking prediction unit 1220 is configured to call a photoresist post-baking prediction model to perform post-baking prediction on the photoacid concentration distribution samples, and obtain residual photoacid concentration distribution prediction results, alkali concentration distribution prediction results and polymer concentration distribution prediction results.

[0249] A result correction unit 1230 is configured to perform post-baking chemical diffusion correction on the residual photoacid concentration distribution prediction results, the alkali concentration distribution prediction results and the polymer concentration distribution prediction results, respectively, and obtain residual photoacid concentration distribution correction results, alkali concentration distribution correction results and polymer concentration distribution correction results.

[0250] The parameter correction unit 1240 is configured to correct model parameters of a photoresist post-baking prediction model according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label, the base concentration distribution correction result and the base concentration distribution label, and the polymer concentration distribution correction result and the polymer concentration distribution label.

[0251] In an embodiment, the sample acquisition unit 1210 is specifically configured to:

[0252] acquire an initial light intensity distribution and an initial photoacid concentration distribution in a photoresist of a wafer;

[0253] perform exposure simulation processing according to the initial light intensity distribution and the initial photoacid concentration distribution to obtain a post-exposure photoacid concentration distribution;

[0254] calculate a difference between the post-exposure photoacid concentration distribution and the initial photoacid concentration distribution to obtain a photoacid concentration distribution sample.

[0255] In an embodiment, the sample acquisition unit 1210 is specifically configured to:

[0256] execute an exposure simulation calculation process with the initial light intensity distribution and the initial photoacid concentration distribution as input parameters, wherein the exposure simulation calculation process includes: calculating a light intensity change rate with depth according to the initial light intensity distribution and the initial photoacid concentration distribution; calculating a light intensity distribution at a next time according to the light intensity change rate with depth and the initial light intensity distribution; calculating a photoacid concentration change rate with time according to the light intensity distribution at the next time and the initial photoacid concentration distribution; and calculating a photoacid concentration distribution at the next time according to the photoacid concentration change rate with time and the initial photoacid concentration distribution;

[0257] repeat execution of the exposure simulation calculation process with the light intensity distribution at the next time and the photoacid concentration distribution at the next time as input parameters until a photoacid concentration distribution at an exposure time is obtained;

[0258] take the photoacid concentration distribution at the exposure time as the post-exposure photoacid concentration distribution.

[0259] In an embodiment, the sample acquisition unit 1210 is specifically configured to:

[0260] acquire an initial base concentration distribution and an initial polymer concentration distribution;

[0261] perform post-baking chemical diffusion simulation processing according to the photoacid concentration distribution sample, the initial base concentration distribution, and the initial polymer concentration distribution to obtain a remaining photoacid concentration distribution simulation result, a base concentration distribution simulation result, and a polymer concentration distribution simulation result;

[0262] take the remaining photoacid concentration distribution simulation result as a remaining photoacid concentration distribution label;

[0263] simulate the base concentration distribution as the base concentration distribution label;

[0264] simulate the polymer concentration distribution as the polymer concentration distribution label.

[0265] In an embodiment, the sample obtaining unit 1210 is specifically configured to:

[0266] take the photoacid concentration distribution sample, the initial base concentration distribution, and the initial polymer concentration distribution as input parameters, and perform a post-baking chemical diffusion simulation calculation process, wherein the post-baking chemical diffusion simulation calculation process comprises: calculating a photoacid concentration distribution at a next time according to the photoacid concentration distribution sample, the initial base concentration distribution, a preset photoacid diffusion rate, and a preset acid-base neutralization rate; calculating a base concentration distribution at the next time according to the photoacid concentration distribution sample, the initial base concentration distribution, a preset base diffusion rate, and the acid-base neutralization rate; and calculating a polymer concentration distribution at the next time according to the photoacid concentration distribution sample, the initial polymer concentration distribution, and a preset polymer deprotection reaction rate;

[0267] take the photoacid concentration distribution at the next time, the base concentration distribution at the next time, and the polymer concentration distribution at the next time as input parameters, and repeatedly perform the post-baking chemical diffusion simulation calculation process until a photoacid concentration distribution at a post-baking time, a base concentration distribution at the post-baking time, and a polymer concentration distribution at the post-baking time are obtained;

[0268] take the photoacid concentration distribution at the post-baking time, the base concentration distribution at the post-baking time, and the polymer concentration distribution at the post-baking time as corresponding remaining photoacid concentration distribution simulation results, base concentration distribution simulation results, and polymer concentration distribution simulation results.

[0269] In an embodiment, the result correcting unit 1230 is specifically configured to:

[0270] take the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, and the polymer concentration distribution prediction result as input parameters, and perform a post-baking chemical diffusion correction calculation process, wherein the post-baking chemical diffusion correction calculation process comprises: calculating a photoacid concentration distribution at a first iteration according to the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, a preset photoacid diffusion rate, and a preset acid-base neutralization rate; calculating a base concentration distribution at the first iteration according to the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, a preset base diffusion rate, and the acid-base neutralization rate; and calculating a polymer concentration distribution at the first iteration according to the remaining photoacid concentration distribution prediction result, the polymer concentration distribution prediction result, and a preset polymer deprotection reaction rate;

[0271] repeating the post-bake chemical diffusion correction calculation process with the first-iteration photoacid concentration distribution, the first-iteration base concentration distribution, and the first-iteration polymer concentration distribution as input parameters until a photoacid concentration distribution at an iteration number, a base concentration distribution at the iteration number, and a polymer concentration distribution at the iteration number are obtained;

[0272] the photoacid concentration distribution at the iteration number, the base concentration distribution at the iteration number, and the polymer concentration distribution at the iteration number correspond to a remaining photoacid concentration distribution correction result, a base concentration distribution correction result, and a polymer concentration distribution correction result.

[0273] In an embodiment, the parameter correction unit 1240 is specifically configured to:

[0274] calculate a photoacid concentration distribution loss value according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label;

[0275] calculate a base concentration distribution loss value according to the base concentration distribution correction result and the base concentration distribution label;

[0276] calculate a polymer concentration distribution loss value according to the polymer concentration distribution correction result and the polymer concentration distribution label;

[0277] correct model parameters of the post-bake photoresist prediction model according to the photoacid concentration distribution loss value, the base concentration distribution loss value, and the polymer concentration distribution loss value.

[0278] In an embodiment, the parameter correction unit 1240 is specifically configured to:

[0279] calculate a square of a difference between the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label to obtain a first photoacid concentration distribution loss term;

[0280] calculate a second photoacid concentration distribution loss term according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution prediction result;

[0281] calculate the photoacid concentration distribution loss value according to the first photoacid concentration distribution loss term and the second photoacid concentration distribution loss term.

[0282] In an embodiment, the parameter correction unit 1240 is specifically configured to:

[0283] calculate a square of a difference between the base concentration distribution correction result and the base concentration distribution label to obtain a first base concentration distribution loss term;

[0284] calculate a second base concentration distribution loss term according to the base concentration distribution correction result and the base concentration distribution prediction result;

[0285] According to the first base concentration distribution loss term and the second base concentration distribution loss term, a base concentration distribution loss value is calculated.

[0286] In an embodiment, the parameter correction unit 1240 is specifically configured to:

[0287] Calculate a square of a difference between the polymer concentration distribution correction result and the polymer concentration distribution label to obtain a first polymer concentration distribution loss term;

[0288] According to the polymer concentration distribution correction result and the polymer concentration distribution prediction result, a second polymer concentration distribution loss term is calculated.

[0289] According to the first polymer concentration distribution loss term and the second polymer concentration distribution loss term, a polymer concentration distribution loss value is calculated.

[0290] In an embodiment, the post-baking prediction unit 1220 is specifically configured to:

[0291] Local feature extraction is performed on the photoacid concentration distribution sample to obtain a sample local feature;

[0292] The photoacid concentration distribution sample is subjected to dimension expansion processing to obtain a first feature, and a dimension number of the first feature is greater than a dimension number of the sample local feature;

[0293] The first feature is subjected to Fourier transform processing to obtain a second feature;

[0294] The second feature and model weight parameters of the photoresist post-baking prediction model are subjected to complex convolution processing to obtain a third feature;

[0295] The third feature is subjected to inverse Fourier transform processing to obtain a fourth feature;

[0296] The fourth feature is subjected to dimension reduction processing to obtain a sample global feature, and a dimension number of the sample global feature is consistent with a dimension number of the sample local feature;

[0297] Based on the sample global feature and the sample local feature, post-baking prediction is performed to obtain a post-baking prediction result, wherein the post-baking prediction result includes one of a residual photoacid concentration distribution prediction result, a base concentration distribution prediction result, or a polymer concentration distribution prediction result.

[0298] Referring to Figure 13 , the embodiments of the present application further disclose an electronic device, and the electronic device 1300 includes:

[0299] At least one processor 1310;

[0300] At least one memory 1320 for storing at least one program;

[0301] When at least one program is executed by the at least one processor 1310, the model training method as described above is implemented.

[0302] The embodiment of the present application further discloses a computer readable storage medium, wherein a computer program executable by a processor is stored, and the computer program executable by the processor is used to implement the model training method as described above when executed by the processor.

[0303] The embodiment of the present application further discloses a computer program product, comprising a computer program or computer instructions, the computer program or computer instructions are stored in a computer readable storage medium, and a processor of an electronic device reads the computer program or computer instructions from the computer readable storage medium, and the processor executes the computer program or computer instructions, so that the electronic device executes the model training method as described above.

[0304] The terms "first", "second", "third", "fourth" and the like used in the description of the specification and the above drawings (if any) are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented, for example, in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0305] It should be understood that in the present application, "at least one" refers to one or more, and "multiple" refers to two or more. "And / or" is used to describe the association relationship of the associated objects, which means that there can be three relationships, for example, "A and / or B" can represent three cases of only A, only B and A and B existing at the same time, wherein A and B can be singular or plural. The character " / " generally represents that the associated objects before and after are in an "or" relationship. "At least one of the following" or similar expressions means any combination of these items, including any combination of single item or multiple items. For example, at least one of a, b or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b and c can be single or multiple.

[0306] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the described device embodiments are merely illustrative. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0307] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program with a predetermined function, and works together with other related parts to achieve a predetermined target, and can be implemented entirely or partially by using software, hardware (such as a processing circuit or a memory) or a combination thereof. Similarly, one processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be a part of an integral module or unit that includes the functions of the module or unit.

[0308] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or they can be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments of the present application.

[0309] In addition, each functional unit in the various embodiments of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0310] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0311] For the step numbers in the above method embodiments, only for the convenience of explanation, the order between the steps is not limited, and the execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

Claims

1. A model training method, characterized in that, The method comprises the following steps: obtaining a photoacid concentration distribution sample and a sample label corresponding to the photoacid concentration distribution sample, the sample label comprising a residual photoacid concentration distribution label, an alkali concentration distribution label and a polymer concentration distribution label; performing post-baking prediction on the photoacid concentration distribution sample by calling a photoresist post-baking prediction model to obtain a residual photoacid concentration distribution prediction result, an alkali concentration distribution prediction result and a polymer concentration distribution prediction result; performing post-baking chemical diffusion correction on the residual photoacid concentration distribution prediction result, the alkali concentration distribution prediction result and the polymer concentration distribution prediction result respectively to obtain a residual photoacid concentration distribution correction result, an alkali concentration distribution correction result and a polymer concentration distribution correction result; correcting model parameters of the photoresist post-baking prediction model according to the residual photoacid concentration distribution correction result and the residual photoacid concentration distribution label, the alkali concentration distribution correction result and the alkali concentration distribution label, and the polymer concentration distribution correction result and the polymer concentration distribution label.

2. The method of claim 1, wherein, The photoacid concentration distribution sample is obtained according to the following steps: obtaining an initial light intensity distribution and an initial photoacid concentration distribution in a photoresist of a wafer; performing exposure simulation processing according to the initial light intensity distribution and the initial photoacid concentration distribution to obtain a post-exposure photoacid concentration distribution; calculating a difference between the post-exposure photoacid concentration distribution and the initial photoacid concentration distribution to obtain the photoacid concentration distribution sample.

3. The method of claim 2, wherein, The exposure simulation processing according to the initial light intensity distribution and the initial photoacid concentration distribution to obtain a post-exposure photoacid concentration distribution comprises: taking the initial light intensity distribution and the initial photoacid concentration distribution as input parameters to perform an exposure simulation calculation process, wherein the exposure simulation calculation process comprises: calculating a light intensity change rate with depth according to the initial light intensity distribution and the initial photoacid concentration distribution; calculating a light intensity distribution at a next time according to the light intensity change rate with depth and the initial light intensity distribution; calculating a photoacid concentration change rate with time according to the light intensity distribution at the next time and the initial photoacid concentration distribution; and calculating a photoacid concentration distribution at the next time according to the photoacid concentration change rate with time and the initial photoacid concentration distribution; taking the light intensity distribution at the next time and the photoacid concentration distribution at the next time as input parameters to repeatedly perform the exposure simulation calculation process until a photoacid concentration distribution at an exposure time is obtained; taking the photoacid concentration distribution at the exposure time as the post-exposure photoacid concentration distribution.

4. The method of claim 2, wherein, The sample label is obtained according to the following steps: obtaining an initial alkali concentration distribution and an initial polymer concentration distribution; performing post-baking chemical diffusion simulation processing according to the photoacid concentration distribution sample, the initial alkali concentration distribution and the initial polymer concentration distribution to obtain a residual photoacid concentration distribution simulation result, an alkali concentration distribution simulation result and a polymer concentration distribution simulation result; taking the residual photoacid concentration distribution simulation result as the residual photoacid concentration distribution label; simulate the base concentration distribution as the base concentration distribution label; simulate the polymer concentration distribution as the polymer concentration distribution label.

5. The method of claim 4, wherein, The post-baking chemical diffusion simulation process is performed according to the photoacid concentration distribution sample, the initial base concentration distribution, and the initial polymer concentration distribution to obtain a remaining photoacid concentration distribution simulation result, a base concentration distribution simulation result, and a polymer concentration distribution simulation result, including: The post-baking chemical diffusion simulation calculation process includes: calculating a photoacid concentration distribution at the next time according to the photoacid concentration distribution sample, the initial base concentration distribution, a preset photoacid diffusion rate, and a preset acid-base neutralization rate; calculating a base concentration distribution at the next time according to the photoacid concentration distribution sample, the initial base concentration distribution, a preset base diffusion rate, and the acid-base neutralization rate; and calculating a polymer concentration distribution at the next time according to the photoacid concentration distribution sample, the initial polymer concentration distribution, and a preset polymer deprotection reaction rate; The post-baking chemical diffusion simulation calculation process is repeatedly performed with the photoacid concentration distribution at the next time, the base concentration distribution at the next time, and the polymer concentration distribution at the next time as input parameters until a photoacid concentration distribution at a post-baking time, a base concentration distribution at a post-baking time, and a polymer concentration distribution at a post-baking time are obtained; The photoacid concentration distribution at the post-baking time, the base concentration distribution at the post-baking time, and the polymer concentration distribution at the post-baking time correspond to the remaining photoacid concentration distribution simulation result, the base concentration distribution simulation result, and the polymer concentration distribution simulation result.

6. The method of claim 1, wherein, The post-baking chemical diffusion correction calculation process includes: calculating a photoacid concentration distribution at the first iteration according to the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, a preset photoacid diffusion rate, and a preset acid-base neutralization rate; calculating a base concentration distribution at the first iteration according to the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, a preset base diffusion rate, and the acid-base neutralization rate; and calculating a polymer concentration distribution at the first iteration according to the remaining photoacid concentration distribution prediction result, the polymer concentration distribution prediction result, and a preset polymer deprotection reaction rate; The post-baking chemical diffusion correction calculation process includes: calculating a photoacid concentration distribution at the first iteration according to the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, a preset photoacid diffusion rate, and a preset acid-base neutralization rate; calculating a base concentration distribution at the first iteration according to the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, a preset base diffusion rate, and the acid-base neutralization rate; and calculating a polymer concentration distribution at the first iteration according to the remaining photoacid concentration distribution prediction result, the polymer concentration distribution prediction result, and a preset polymer deprotection reaction rate; repeating the post-bake chemical diffusion correction calculation process with the first-iteration photoacid concentration distribution, the first-iteration base concentration distribution, and the first-iteration polymer concentration distribution as input parameters until a photoacid concentration distribution at an iteration number, a base concentration distribution at the iteration number, and a polymer concentration distribution at the iteration number are obtained; the photoacid concentration distribution at the iteration number, the base concentration distribution at the iteration number, and the polymer concentration distribution at the iteration number correspond to a remaining photoacid concentration distribution correction result, a base concentration distribution correction result, and a polymer concentration distribution correction result.

7. The method of claim 1, wherein, The model parameters of the photoresist post-bake prediction model are corrected according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label, the base concentration distribution correction result and the base concentration distribution label, and the polymer concentration distribution correction result and the polymer concentration distribution label, including: According to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label, a photoacid concentration distribution loss value is calculated. According to the base concentration distribution correction result and the base concentration distribution label, a base concentration distribution loss value is calculated. According to the polymer concentration distribution correction result and the polymer concentration distribution label, a polymer concentration distribution loss value is calculated. The model parameters of the photoresist post-bake prediction model are corrected according to the photoacid concentration distribution loss value, the base concentration distribution loss value, and the polymer concentration distribution loss value.

8. The method of claim 7, wherein, The model parameters of the photoresist post-bake prediction model are corrected according to the photoacid concentration distribution loss value, the base concentration distribution loss value, and the polymer concentration distribution loss value. The photoacid concentration distribution loss value is calculated according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label, including: The square of the difference between the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label is calculated to obtain a first photoacid concentration distribution loss term. A second photoacid concentration distribution loss term is calculated according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution prediction result.

9. The method of claim 7, wherein, The photoacid concentration distribution loss value is calculated according to the first photoacid concentration distribution loss term and the second photoacid concentration distribution loss term. The base concentration distribution loss value is calculated according to the base concentration distribution correction result and the base concentration distribution label, including: The square of the difference between the base concentration distribution correction result and the base concentration distribution label is calculated to obtain a first base concentration distribution loss term. A second base concentration distribution loss term is calculated according to the base concentration distribution correction result and the base concentration distribution prediction result.

10. The method of claim 7, wherein, The base concentration distribution loss value is calculated according to the first base concentration distribution loss term and the second base concentration distribution loss term. The polymer concentration distribution loss value is calculated according to the polymer concentration distribution correction result and the polymer concentration distribution label, including: The square of the difference between the polymer concentration distribution correction result and the polymer concentration distribution label is calculated to obtain a first polymer concentration distribution loss term. A second polymer concentration distribution loss term is calculated according to the polymer concentration distribution correction result and the polymer concentration distribution prediction result. The polymer concentration distribution loss value is calculated according to the first polymer concentration distribution loss term and the second polymer concentration distribution loss term. According to the first polymer concentration distribution loss term and the second polymer concentration distribution loss term, a polymer concentration distribution loss value is calculated.

11. The method of claim 1, wherein, The photoresist post-baking prediction model performs post-baking prediction on the photoacid concentration distribution sample, and the process includes the following steps: Local feature extraction is performed on the photoacid concentration distribution sample to obtain sample local features. Dimension expansion processing is performed on the photoacid concentration distribution sample to obtain first features, and the number of dimensions of the first features is greater than the number of dimensions of the sample local features. Fourier transform processing is performed on the first features to obtain second features. Complex convolution processing is performed on the second features and the model weight parameters of the photoresist post-baking prediction model to obtain third features. Inverse Fourier transform processing is performed on the third features to obtain fourth features. Dimension reduction processing is performed on the fourth features to obtain sample global features, and the number of dimensions of the sample global features is consistent with the number of dimensions of the sample local features. Based on the sample global features and the sample local features, post-baking prediction is performed to obtain a post-baking prediction result, wherein the post-baking prediction result includes one of the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, or the polymer concentration distribution prediction result.

12. A model training apparatus, comprising: It includes: A sample acquisition unit is configured to acquire a photoacid concentration distribution sample and a sample label corresponding to the photoacid concentration distribution sample, wherein the sample label includes a remaining photoacid concentration distribution label, a base concentration distribution label, and a polymer concentration distribution label. A post-baking prediction unit is configured to call a photoresist post-baking prediction model to perform post-baking prediction on the photoacid concentration distribution sample to obtain a remaining photoacid concentration distribution prediction result, a base concentration distribution prediction result, and a polymer concentration distribution prediction result. A result correction unit is configured to perform post-baking chemical diffusion correction on the remaining photoacid concentration distribution prediction result, the base concentration distribution prediction result, and the polymer concentration distribution prediction result, respectively, to obtain a remaining photoacid concentration distribution correction result, a base concentration distribution correction result, and a polymer concentration distribution correction result. A parameter correction unit is configured to correct model parameters of the photoresist post-baking prediction model according to the remaining photoacid concentration distribution correction result and the remaining photoacid concentration distribution label, the base concentration distribution correction result and the base concentration distribution label, and the polymer concentration distribution correction result and the polymer concentration distribution label.

13. An electronic device, comprising: It includes: At least one processor; At least one memory for storing at least one program; When at least one of the programs is executed by at least one of the processors, the model training method according to any one of claims 1 to 11 is implemented.

14. A computer-readable storage medium, characterized in that, The computer program executable by the processor is stored in the computer readable storage medium, and the computer program executable by the processor is executed by the processor to implement the model training method according to any one of claims 1 to 11.

15. A computer program product comprising computer programs or computer instructions, characterized in that, The computer program or the computer instruction is stored in a computer readable storage medium, and a processor of the electronic device reads the computer program or the computer instruction from the computer readable storage medium, and the processor executes the computer program or the computer instruction, so that the electronic device executes the model training method in any one of claims 1 to 11.

Images