Overlay deviation processing method, device, equipment, medium and system

Abstract

The invention relates to the technical field of semiconductors, in particular to an overlay deviation processing method, device, equipment, medium and system.The method comprises the steps that front and back process parameter characteristics and a deviation compensation model associated with a current layer are obtained and input into the deviation compensation model for deviation compensation processing, obtaining prediction deviation compensation data corresponding to front and back process parameter characteristics; under the condition that the predicted deviation compensation data corresponding to the first process parameter feature in the front and back process parameter features does not meet a preset condition, processing the wafers of the current batch based on the first process parameter feature, and performing post-development inspection and post-etching inspection on the processed wafers to generate real-time deviation compensation data; and the deviation compensation model is corrected according to the real-time deviation compensation data, so that the technological parameter characteristics of the previous and later processes except the photoetching process can be dynamically added into OVL correction, and the yield loss caused by OVL deviation in the mass production process is reduced.

Description

Technical Field

[0001] The present application relates to the field of semiconductor manufacturing, and in particular to a method, device, equipment, medium and system for processing overlay deviation. Background Art

[0002] As process scaling continues, the process window narrows, and overlay (OVL) must be controlled with increasing precision to meet process requirements. Traditional OVL control is limited to the photolithography process. This optimization involves optimizing alignment during exposure, ADIOVL measurement accuracy after exposure, and feedforward (FF) and feedback (FB) mechanisms for ADIOVL compensation between previous and next layers to achieve better OVL control.

[0003] In the development of advanced nodes, the impact of front-end and back-end processes other than photolithography, such as chemical-mechanical planarization (CMP), etching, and high-temperature annealing, on device OVL cannot be ignored. Figure 1 As shown in the figure, related techniques typically measure the ADIOVL and AEI OVL of several wafers, then statically compensate the average MTD of these wafers for the ADI OVL process, making the ADI OVL measurement closer to the actual AEIOVL measurement value. ADI refers to post-development inspection, AEI refers to post-etch inspection, and MTD refers to the offset compensation data between AEIOVL and ADIOVL.

[0004] However, the compensation method in the related art is a static compensation method, which cannot perform real-time compensation for MTD, and thus cannot reduce the yield loss caused by OVL deviation during high volume manufacturing (HVM). Summary of the Invention

[0005] The present application provides an overlay deviation processing method, device, equipment, medium and system to at least solve the above-mentioned problems existing in the related art.

[0006] In order to solve the above technical problems, the technical solutions of this application are as follows:

[0007] According to a first aspect of an embodiment of the present application, a method for processing overlay deviation is provided, the method comprising:

[0008] During the photolithography process of the current batch of wafers, front-end and back-end process parameter characteristics and a deviation compensation model associated with the current layer are obtained; the deviation compensation model is obtained by training an initial model based on preset process parameter characteristics and preset deviation compensation data corresponding to the preset process parameter characteristics;

[0009] Inputting the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics;

[0010] When the predicted deviation compensation data corresponding to the first process parameter feature among the front-end and back-end process parameter features does not meet a preset condition, the wafers of the current batch are processed based on the first process parameter feature, and a post-development inspection and a post-etching inspection are performed on the processed wafers, and real-time deviation compensation data corresponding to the first process parameter feature is generated according to the post-development inspection results and the post-etching inspection results;

[0011] The deviation compensation model is corrected according to the real-time deviation compensation data to obtain a corrected deviation compensation model; the corrected deviation compensation model is used to perform deviation compensation on the next batch of wafers.

[0012] In an optional embodiment, obtaining the front-end and back-end process parameter characteristics associated with the current layer includes:

[0013] Obtaining machine status information and front-layer process information of the back-end process; the front-layer process information is the process information of the front layer before the current layer is prepared, and the back-end process is the process after the current layer is photolithographically processed;

[0014] The front-end and back-end process parameter characteristics are generated according to the front-end process information and the machine status information of the back-end process.

[0015] In an optional embodiment, the training process of the deviation compensation model includes:

[0016] Acquiring the preset process parameter characteristics and the preset deviation compensation data;

[0017] Inputting the preset process parameter characteristics into the initial model for deviation compensation processing to obtain deviation compensation data output by the initial model;

[0018] The model parameters of the initial model are adjusted according to the difference between the deviation compensation data output by the initial model and the preset deviation compensation data until the difference meets the preset difference condition or the number of initial model training times meets the preset number condition, thereby obtaining the deviation compensation model.

[0019] In an optional embodiment, the obtaining of the preset process parameter characteristics and the preset deviation compensation data includes:

[0020] Determining at least two initial process parameter characteristics in a wafer processing process; the initial process parameter characteristics include machine status information of a preset back-end process and preset front-layer process information; the preset front-layer process information is process information of a preset front layer before preparing a preset current layer, the preset back-end process is a process after photolithography of the preset current layer, and the preset current layer is a current layer formed during photolithography of a preset current batch of wafers;

[0021] In a wafer processing process, adjusting any initial process parameter characteristic, fixing other initial process parameter characteristics, performing a post-development inspection process on the wafer after development processing, and performing a post-etching inspection process on the wafer after etching, and generating deviation compensation data corresponding to any of the initial process parameter characteristics based on the post-development inspection results and the post-etching inspection results;

[0022] The any one of the initial process parameter characteristics and the deviation compensation data corresponding to the any one of the initial process parameter characteristics are screened to obtain the preset process parameter characteristics and the preset deviation compensation data.

[0023] In an optional embodiment, the screening of any of the initial process parameter characteristics and the deviation compensation data corresponding to the any of the initial process parameter characteristics to obtain the preset process parameter characteristics and the preset deviation compensation data includes:

[0024] Obtaining the preset process parameter characteristics whose corresponding deviation compensation data satisfies the preset deviation condition from the at least two initial process parameter characteristics;

[0025] The preset deviation compensation data corresponding to the preset process parameter feature is obtained from the deviation compensation data corresponding to any one of the initial process parameter features.

[0026] In an optional embodiment, obtaining the status information of a downstream process tool includes:

[0027] When the tool status information of the back-end process is predetermined in the photolithography process of the current layer, the predetermined tool status information of the back-end process is acquired.

[0028] In an optional embodiment, obtaining the status information of a downstream process tool includes:

[0029] When the tool status information of the back-end process is not predetermined in the photolithography process of the current layer, the tool status information of all tools of the back-end process is obtained;

[0030] The tool status information of the back-end process is generated according to the tool status information of all the tools.

[0031] In an optional embodiment, inputting the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics includes:

[0032] The front-end and back-end process parameter characteristics are input into the deviation compensation model for deviation compensation processing to obtain a deviation compensation image, and the deviation compensation image is converted into vector information of a preset dimension to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics.

[0033] In an optional embodiment, inputting the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics includes:

[0034] Inputting the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing, and obtaining predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics and the confidence level of the predicted deviation compensation data;

[0035] When the predicted deviation compensation data corresponding to the first process parameter feature in the front-end and back-end process parameter features does not meet a preset condition, the wafers of the current batch are processed based on the first process parameter feature, and a post-development inspection and a post-etching inspection are performed on the processed wafers, and real-time deviation compensation data corresponding to the first process parameter feature is generated according to the post-development inspection results and the post-etching inspection results, including:

[0036] When the confidence level of the predicted deviation compensation data corresponding to the first process parameter characteristic does not meet the preset conditions, the current batch of wafers is processed based on the first process parameter characteristic, and the processed wafers are subjected to post-development inspection and post-etching inspection, and real-time deviation compensation data corresponding to the first process parameter characteristic is generated according to the difference between the post-development inspection results and the post-etching inspection results.

[0037] In an optional embodiment, the correcting the deviation compensation model according to the real-time deviation compensation data to obtain a corrected deviation compensation model includes:

[0038] Inputting the first process parameter characteristic and the real-time deviation compensation data into the deviation compensation model, so that the deviation compensation model performs deviation compensation processing on the first process parameter characteristic, and obtaining target deviation compensation data output by the deviation compensation model;

[0039] The model parameters of the deviation compensation model are adjusted according to the difference between the real-time deviation compensation data and the target deviation compensation data until the difference meets a preset difference condition or the number of deviation compensation model trainings meets a preset number condition, thereby obtaining a corrected deviation compensation model.

[0040] In an optional embodiment, the method further includes:

[0041] When the predicted deviation compensation data corresponding to the second process parameter feature in the front-end and back-end process parameter features meets a preset condition, processing the wafers of the current batch based on the second process parameter feature and the predicted deviation compensation data corresponding to the second process parameter feature, and performing a post-development inspection on the processed wafers to obtain a post-development inspection result corresponding to the second process parameter feature;

[0042] Among them, the second process parameter feature is a process parameter feature in the front-end and back-end process parameter features except the first process parameter feature, and the post-development inspection result corresponding to the second process parameter feature is used to indicate that the overlay deviation of the processed wafer meets the preset deviation threshold.

[0043] According to a second aspect of an embodiment of the present application, there is provided an apparatus for processing overlay deviation, the apparatus comprising:

[0044] A feature model acquisition module is used to acquire front-end and back-end process parameter features and a deviation compensation model associated with the current layer during the photolithography process of the current batch of wafers; the deviation compensation model is obtained by training an initial model based on preset process parameter features and preset deviation compensation data corresponding to the preset process parameter features;

[0045] a compensation processing module, configured to input the front-end and back-end process parameter characteristics into the deviation compensation model for performing deviation compensation processing, and obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics;

[0046] a real-time data generation module, configured to, if predicted deviation compensation data corresponding to a first process parameter characteristic among the front-end and back-end process parameter characteristics does not satisfy a preset condition, process the wafers of the current batch based on the first process parameter characteristic, perform a post-development inspection and a post-etching inspection on the processed wafers, and generate real-time deviation compensation data corresponding to the first process parameter characteristic based on the post-development inspection results and the post-etching inspection results;

[0047] The correction module is used to correct the deviation compensation model according to the real-time deviation compensation data to obtain a corrected deviation compensation model; the corrected deviation compensation model is used to perform deviation compensation on the next batch of wafers.

[0048] In an optional embodiment, the feature model acquisition module includes:

[0049] The status process information acquisition module is used to obtain the machine status information and front-layer process information of the back-end process; the front-layer process information is the process information of the front layer before preparing the current layer, and the back-end process is the process after the current layer is photolithographically processed.

[0050] The front-end and back-end process parameter generation module is used to generate the front-end and back-end process parameter characteristics according to the front-end process information and the machine status information of the back-end process.

[0051] In an optional embodiment, the device further comprises:

[0052] A feature compensation data acquisition module, used to acquire preset process parameter features and the preset deviation compensation data;

[0053] An input module, configured to input the preset process parameter characteristics into the initial model for deviation compensation processing, and obtain deviation compensation data output by the initial model;

[0054] An adjustment module is used to adjust the model parameters of the initial model according to the difference between the deviation compensation data output by the initial model and the preset deviation compensation data, until the difference meets the preset difference condition or the number of initial model training times meets the preset number condition, thereby obtaining the deviation compensation model.

[0055] In an optional embodiment, the feature compensation data acquisition module includes:

[0056] an initial process parameter characteristic determination unit, configured to determine at least two initial process parameter characteristics in a wafer processing process; the initial process parameter characteristics comprising machine status information of a preset back-end process and preset front-layer process information; the preset front-layer process information being process information of a preset front layer before preparing a preset current layer, the preset back-end process being a process after photolithography of a preset current layer, and the preset current layer being a current layer formed during photolithography of a preset current batch of wafers;

[0057] a variable control unit for adjusting any initial process parameter characteristic and fixing other initial process parameter characteristics during the wafer processing process, performing a post-development inspection process on the wafer after development processing, and performing a post-etching inspection process on the wafer after etching, and generating deviation compensation data corresponding to any of the initial process parameter characteristics based on the post-development inspection results and the post-etching inspection results;

[0058] The screening unit is used to screen any of the initial process parameter characteristics and the deviation compensation data corresponding to any of the initial process parameter characteristics to obtain the preset process parameter characteristics and the preset deviation compensation data.

[0059] In an optional embodiment, the screening unit includes:

[0060] a preset process parameter characteristic acquisition subunit, configured to acquire, from the at least two initial process parameter characteristics, the preset process parameter characteristics whose corresponding deviation compensation data satisfies a preset deviation condition;

[0061] The preset deviation compensation data acquisition subunit is used to acquire the preset deviation compensation data corresponding to the preset process parameter feature from the deviation compensation data corresponding to any initial process parameter feature.

[0062] In an optional embodiment, the state process information acquisition module includes:

[0063] The first machine status information acquiring unit is configured to acquire the predetermined machine status information of the back-end process when the machine status information of the back-end process is predetermined during the photolithography process of the current layer.

[0064] In an optional embodiment, the state process information acquisition module includes:

[0065] A second machine status information acquisition unit is configured to acquire machine status information of all machines in the back-end process when the machine status information of the back-end process is not predetermined in the photolithography process of the current layer;

[0066] The machine status information generating unit is used to generate the machine status information of the back-end process according to the machine status information of all the machines.

[0067] In an optional embodiment, the compensation processing module includes:

[0068] The first processing sub-unit is used to input the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain a deviation compensation image, convert the deviation compensation image into vector information of a preset dimension, and obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics.

[0069] In an optional embodiment, the compensation processing module includes:

[0070] The second processing sub-unit is used to input the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing, and obtain the predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics and the confidence of the predicted deviation compensation data.

[0071] Correspondingly, the real-time data generation module is used to process the current batch of wafers based on the first process parameter characteristics when the confidence level of the predicted deviation compensation data corresponding to the first process parameter characteristics does not meet the preset conditions, perform post-development inspection and post-etching inspection on the processed wafers, and generate real-time deviation compensation data corresponding to the first process parameter characteristics according to the difference between the post-development inspection results and the post-etching inspection results.

[0072] In an optional embodiment, the correction module includes:

[0073] a target deviation compensation data generating unit, configured to input the first process parameter characteristic and the real-time deviation compensation data into the deviation compensation model, so that the deviation compensation model performs deviation compensation processing on the first process parameter characteristic, and obtain target deviation compensation data output by the deviation compensation model;

[0074] A difference adjustment unit is used to adjust the model parameters of the deviation compensation model according to the difference between the real-time deviation compensation data and the target deviation compensation data until the difference meets a preset difference condition or the number of deviation compensation model training times meets a preset number condition, thereby obtaining a corrected deviation compensation model.

[0075] In an optional embodiment, the device further comprises:

[0076] a post-development inspection result generating module, configured to process the wafers of the current batch based on the second process parameter characteristic and the predicted deviation compensation data corresponding to the second process parameter characteristic in the front-end and back-end process parameter characteristics, and perform a post-development inspection on the processed wafers to obtain a post-development inspection result corresponding to the second process parameter characteristic, if the predicted deviation compensation data corresponding to the second process parameter characteristic in the front-end and back-end process parameter characteristics meets a preset condition;

[0077] Among them, the second process parameter feature is a process parameter feature in the front-end and back-end process parameter features except the first process parameter feature, and the post-development inspection result corresponding to the second process parameter feature is used to indicate that the overlay deviation of the processed wafer meets the preset deviation threshold.

[0078] According to a third aspect of an embodiment of the present application, an electronic device for processing overlay deviation is provided, the electronic device comprising a processor and a memory, the memory storing at least one instruction or at least one program, the at least one instruction or at least one program being loaded by the processor and executing the overlay deviation processing method as described in the above embodiment.

[0079] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, in which at least one instruction or at least one program is stored. The at least one instruction or at least one program is loaded and executed by a processor to implement the overlay deviation processing method as described in the above embodiment.

[0080] According to a fifth aspect of the present application, an overlay deviation processing system is provided, wherein the overlay deviation processing system includes:

[0081] Post-development inspection tools;

[0082] Post-etch inspection tools;

[0083] photolithography tools for fabricating the current and previous layers;

[0084] The controller is used to execute the overlay deviation processing method as described in the above embodiment.

[0085] The technical solutions provided by the embodiments of this application bring at least the following beneficial effects:

[0086] The embodiments of the present application provide an etching deviation processing method, device, equipment, medium and system, which, during the process of photolithography of the current batch of wafers, obtain the front-end and back-end process parameter characteristics and deviation compensation model associated with the current layer; the deviation compensation model is obtained by training an initial model based on preset process parameter characteristics and preset deviation compensation data corresponding to the preset process parameter characteristics; the front-end and back-end process parameter characteristics are input into the deviation compensation model for deviation compensation processing to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics; when the predicted deviation compensation data corresponding to the first process parameter characteristic in the front-end and back-end process parameter characteristics does not meet the preset conditions, the current batch of wafers is processed based on the first process parameter characteristic, and the processed wafers are subjected to post-development inspection and post-etching inspection, and real-time deviation compensation data corresponding to the first process parameter characteristic is generated according to the post-development inspection results and the post-etching inspection results; the deviation compensation model is corrected according to the real-time deviation compensation data to obtain a corrected deviation compensation model. In this way, the influence of front-end and back-end processes other than the lithography process can also be dynamically added to the OVL correction of the exposure machine, thereby reducing the influence of front-end and back-end processes on OVL to a certain extent; in addition, when the predicted deviation compensation data corresponding to the first process parameter feature in the front-end and back-end process parameter features does not meet the preset conditions, the deviation compensation model is corrected according to the real-time deviation compensation data, so that the front-end and back-end process parameter features other than the lithography process can be dynamically added to the OVL correction of the exposure machine, ensuring that each batch of wafers is produced with a better MTD compensation value under real-time data, achieving better OVL performance, and thus reducing the yield loss caused by OVL deviation during mass production.

[0087] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification are used to explain the principles of the present application, and do not constitute an improper limitation on the present application.

[0089] Figure 1 It is a flowchart of OVL control in related technologies.

[0090] Figure 2 A schematic diagram of a method for processing overlay deviation according to an exemplary embodiment is shown. Figure 1 .

[0091] Figure 3 A schematic diagram of a method for processing overlay deviation according to an exemplary embodiment is shown. Figure 2 .

[0092] Figure 4 This is a flow diagram of a training process of a deviation compensation model according to an exemplary embodiment. Figure 1 .

[0093] Figure 5 This is a flow diagram of a training process of a deviation compensation model according to an exemplary embodiment. Figure 2 .

[0094] Figure 6 FIG. 1 is a schematic diagram illustrating the influence of a back-end etching process of a current layer on deviation compensation data according to an exemplary embodiment.

[0095] Figure 7 FIG. 1 is a schematic diagram showing the influence of the deposition thickness of a front film on the deviation compensation data according to an exemplary embodiment.

[0096] Figure 8 FIG. 1 is a schematic diagram illustrating the influence of a front-layer CMP process on deviation compensation data according to an exemplary embodiment.

[0097] Figure 9 It is a schematic structural diagram of an initial model according to an exemplary embodiment.

[0098] Figure 10 A schematic diagram of a method for processing overlay deviation according to an exemplary embodiment is shown. Figure 3 .

[0099] Figure 11A schematic diagram of a method for processing overlay deviation according to an exemplary embodiment is shown. Figure 4 .

[0100] Figure 12 The figure is a block diagram of an apparatus for processing overlay deviation according to an exemplary embodiment. DETAILED DESCRIPTION

[0101] The following provides many different embodiments or examples for implementing the different features of the provided subject matter. The specific examples of the components and configurations described below are disclosed in a simplified manner. Of course, these components and configurations are merely examples and are not intended to be restrictive. For example, in the following description, the formation of a first feature above or on a second feature may include an embodiment in which the first and second features are formed in direct contact, and may also include an embodiment in which an additional feature may be formed between the first and second features so that the first and second features may not be in direct contact. In addition, the present application may repeat reference numbers and / or letters in various examples. This repetition is for the purpose of simplicity and clarity and does not itself indicate the relationship between the various embodiments and / or configurations discussed.

[0102] Additionally, spatially relative terms, such as "below," "beneath," "lower," "on," "upper," "front," "back," "above," and the like, may be used herein for ease of description to describe the relationship of one element or feature to another element or feature as illustrated in the figures. The spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures.

[0103] As described in the background art, static compensation in related art cannot reduce the yield loss caused by OVL deviation during mass production. For example, the static compensation method in related art has the following problems:

[0104] (1) Unable to compensate for the MTD differences caused by different Etch, CMP and high-temperature furnace process machines and semiconductor etching equipment (chambers), resulting in the OVL performance of wafers produced by one machine being good while the OVL performance of another machine is poor.

[0105] (2) Even for a single tool and chamber, it is impossible to compensate for the MTD differences caused by the real-time status changes of the tool, resulting in the OVL performance of wafers produced by the tool being good during a certain period of time, while the OVL performance is poor during other periods.

[0106] (3) The change of MTD depends on the delayed feedback of yield and slicing. When the problem is discovered, many wafers have already passed the problem site, which increases the production cost of wafers.

[0107] Based on this, in order to dynamically add the front-end and back-end process parameter characteristics outside the lithography process to the OVL correction of the exposure machine, ensure that each batch of wafers is produced with a better MTD compensation value under real-time data, achieve better OVL performance, and thereby reduce the yield loss caused by OVL deviation during mass production, the embodiments of the present application provide an overlay deviation processing method, device, equipment, medium and system.

[0108] Figure 2 A schematic diagram of a method for processing overlay deviation according to an exemplary embodiment is shown. Figure 1 ,like Figure 2 As shown, the method may include:

[0109] S11. During the photolithography process of the current batch of wafers, the front-end and back-end process parameter characteristics and the deviation compensation model associated with the current layer are obtained; the deviation compensation model is obtained by training the initial model based on the preset process parameter characteristics and the preset deviation compensation data corresponding to the preset process parameter characteristics.

[0110] In this embodiment, when the current batch of wafers arrives at the material coating station of the lithography process, the deviation compensation model (MTD model) is triggered to obtain the front-end and back-end process parameter characteristics and deviation compensation model associated with the current layer. The current layer refers to the image layer currently exposed during the lithography process.

[0111] Alternatively, research has found that MTD is influenced by two factors. The first is that the wafer edge mark, due to non-uniformity at the edge of the previous layer process, often exhibits asymmetric characteristics such as tilt, non-uniform critical dimensions (CD), and bowing, affecting the accuracy of subsequent measurements. The second factor is the positional variation during the transfer to the hard mask during post-lithography processes such as etching.

[0112] Therefore, based on the above two factors, the front-end and back-end process parameter characteristics may include the front-layer process information and the machine status information of the back-end process. The front-layer process information is for the front layer before the current layer, which may include the front-layer process information for preparing the front layer. Furthermore, the "front-layer process information" may refer to the process steps that affect the asymmetry of the front-layer OVL mask. Exemplarily, the "front-layer process information" may include but is not limited to: the site required to prepare the front layer, which machines need to be used, the parameter information used by each machine, the front-layer film deposition (film dep) thickness, the front-layer etching process (etch process), the front-layer high-temperature process, the front-layer CMP process, etc. The back-end process is for the current layer, which may include the process after the current layer lithography. Furthermore, the "back-end process" may refer to the process that affects the transfer of the current layer's graphics to the hard mask. The "machine status information" may refer to the status information of the machine used in the back-end process. For example, the "back-end process" may include, but is not limited to, etching processes, thin film deposition processes, interconnect processes, testing processes, packaging processes, etc. For example, the "tool status information" may refer to temperature information, operating hours information, etc. of tools used in the back-end process.

[0113] Optionally, a "previous layer" refers to an image layer that is exposed before the current layer and requires OVL control. In one embodiment, the "previous layer" and the current layer may be separated by several process steps. In another embodiment, the "previous layer" and the current layer may be adjacent layers that are in direct contact.

[0114] Optionally, the “back-end process tool status information” may include, but is not limited to: site information after photolithography, tools included in each site, parameters used by the tools included in each site, etc.

[0115] Accordingly, Figure 3 A schematic diagram of a method for processing overlay deviation according to an exemplary embodiment is shown. Figure 2 ,like Figure 3 As shown, in the above step S11, the above-mentioned acquisition of the front-end and back-end process parameter characteristics associated with the current layer may include:

[0116] S111. Obtain the machine status information of the back-end process and the front-end process information.

[0117] S112. Generate the front-end and back-end process parameter characteristics according to the front-end process information and the back-end process machine status information.

[0118] The machine status information of the back-end process and the front-end process information in the above S111 can be found in the above description and will not be repeated here. In one embodiment, in the above step S112, after obtaining the front-end process information and the machine status information of the back-end process, the front-end process information and the machine status information of the back-end process can be directly aggregated and fused to obtain the front-end and back-end process parameter characteristics. In another embodiment, the front-end process information and the machine status information of the back-end process can also be screened to filter out the front-end process information and the machine status information of the back-end process whose influence on the OVL MTD is greater than a preset threshold. For example, the front-end process information and the machine status information of the back-end process can be subjected to data preprocessing, feature extraction, feature selection and other processing to obtain the front-end process information and the machine status information of the back-end process that are ultimately used for subsequent model calculations.

[0119] Due to the non-uniformity of the previous layer process at the edge of the wafer edge mask, it often exhibits asymmetric performance such as tilt, non-uniform critical dimension (CD), and bowing, affecting the accuracy of subsequent measurements. Furthermore, post-litho processes such as etching can cause positional changes during the transfer to the hard mask. Therefore, obtaining the machine status information of the post-process and the previous layer process information and incorporating this information into dynamic MTD compensation can not only reduce the asymmetric performance of the wafer edge mask such as tilt, non-uniform critical dimension (CD), and bowing caused by the non-uniformity of the previous layer process at the edge, but also reduce the positional changes caused by the post-process during the transfer to the hard mask, thereby improving the OVL performance of the device and reducing the yield loss caused by OVL deviation during mass production.

[0120] Optionally, the deviation compensation model (MTD model) can be pre-trained using preset process parameter characteristics and preset deviation compensation data corresponding to the preset process parameter characteristics to establish a multi-dimensional association between the preset process parameters and the MTD to obtain the MTD model. The trained MTD model has the function of outputting deviation compensation data corresponding to the input process parameter characteristics based on the process parameter characteristics.

[0121] It should be noted that the MTD model can be various types of models, and the embodiments of the present application do not limit the type of the MTD model.

[0122] S12. Input the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics.

[0123] In this embodiment, since the trained model has the function of outputting the deviation compensation data corresponding to the input process parameter characteristics, the front-end and back-end process parameter characteristics can be input into the deviation compensation model for deviation compensation processing to obtain the predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics.

[0124] S13. When the predicted deviation compensation data corresponding to the first process parameter feature in the front-end and back-end process parameter features does not meet the preset conditions, the wafers of the current batch are processed based on the first process parameter feature, and the processed wafers are inspected after development and after etching, and the real-time deviation compensation data corresponding to the first process parameter feature are generated according to the results of the post-development inspection and the post-etching inspection.

[0125] S14. Correcting the deviation compensation model according to the real-time deviation compensation data to obtain a corrected deviation compensation model; the corrected deviation compensation model is used to perform deviation compensation on the next batch of wafers.

[0126] In this embodiment, the number of front-end and back-end process parameter features can be at least two. When the predicted deviation compensation data corresponding to some first process parameter features in the front-end and back-end process parameter features does not meet the preset conditions, it means that the first process parameter features were not used in the process of training the MTD model. Based on this, data can be re-collected to trigger the current batch of wafers to perform ADI and AEI in-station measurements under the first process parameters to collect real-time deviation compensation data (i.e., real-time MTD value) and correct the MTD model.

[0127] Since the first process parameter feature is a process parameter feature among the front-end and back-end process parameter features, the first process parameter feature may also include front-layer process information and back-end process machine status information. The front-layer process information and back-end process machine status information included in the first process parameter feature can be used to process the current batch of wafers, and the processed wafers can be subjected to post-development inspection and post-etching inspection. Based on the difference between the post-development inspection results and the post-etching inspection results, real-time deviation compensation data corresponding to the first process parameter feature is generated. Then, the deviation compensation model is corrected based on the real-time deviation compensation data to obtain a corrected deviation compensation model.

[0128] In an optional embodiment, in the above step S14, the correction processing of the deviation compensation model according to the real-time deviation compensation data to obtain the corrected deviation compensation model may include:

[0129] The first process parameter characteristic and the real-time deviation compensation data are input into the deviation compensation model, so that the deviation compensation model performs deviation compensation processing on the first process parameter characteristic, and obtains target deviation compensation data output by the deviation compensation model.

[0130] The model parameters of the deviation compensation model are adjusted according to the difference between the real-time deviation compensation data and the target deviation compensation data until the difference meets a preset difference condition or the number of deviation compensation model trainings meets a preset number condition, thereby obtaining a corrected deviation compensation model.

[0131] Optionally, the real-time deviation compensation data is equivalent to the actual deviation compensation data, which can be considered as the label data in the model training process. Then, correcting the deviation compensation model can refer to: inputting the first process parameter feature and the real-time deviation compensation data corresponding to the first process parameter feature as training data into the deviation compensation model, the deviation compensation model performing deviation compensation on the first process parameter feature, obtaining the target deviation compensation data output by the deviation compensation model, calculating the difference between the real-time deviation compensation data and the target deviation compensation data, and adjusting the model parameters of the deviation compensation model according to the difference until the difference meets the preset difference condition or the number of deviation compensation model trainings meets the preset number condition, thereby obtaining a corrected deviation compensation model. The corrected deviation compensation model is enabled to output the deviation compensation data corresponding to the first process parameter feature according to the first process parameter feature. In this way, the training data of the deviation compensation data can be enriched, so that the corrected deviation compensation data can accurately output the corresponding deviation compensation data for a large number of process parameter features, thereby improving the prediction range and prediction accuracy of the corrected model.

[0132] Optionally, the above-mentioned “prediction deviation compensation data does not meet the preset conditions” may mean that the prediction deviation compensation data is greater than a certain threshold, or the corresponding confidence level is not within an acceptable range, etc., and the implementation of this application does not make specific limitations on this.

[0133] Optionally, the revised deviation compensation model is used to perform deviation compensation processing on the next batch of wafers. For example, before the deviation compensation model was revised, it could not accurately predict the corresponding real-time deviation compensation data for certain machines and their parameters. After the revision, it can accurately predict the corresponding real-time deviation compensation data for certain machines and their parameters. Then, when the next batch of wafers arrives at the material coating station of the lithography (litho), the calculation of the revised deviation compensation model (MTD model) can be triggered. In this way, the revised deviation compensation model can predict certain machines and their parameters and obtain the corresponding real-time deviation compensation data.

[0134] In this way, the influence of front-end and back-end processes other than the lithography process can also be dynamically added to the OVL correction of the exposure machine, thereby reducing the influence of front-end and back-end processes on OVL to a certain extent; in addition, when the predicted deviation compensation data corresponding to the first process parameter feature in the front-end and back-end process parameter features does not meet the preset conditions, the deviation compensation model is corrected according to the real-time deviation compensation data, so that the front-end and back-end process parameter features other than the lithography process can be dynamically added to the OVL correction of the exposure machine, ensuring that each batch of wafers is produced with a better MTD compensation value under real-time data, achieving better OVL performance, and thus reducing the yield loss caused by OVL deviation during mass production.

[0135] First, the training process of the bias compensation model is introduced.

[0136] Figure 4 This is a flow diagram of a training process of a deviation compensation model according to an exemplary embodiment. Figure 1 ,like Figure 4 As shown, the training process of the deviation compensation model includes:

[0137] S01. Obtaining preset process parameter characteristics and the preset deviation compensation data.

[0138] Optionally, based on the MTD generation mechanism, factors affecting the deviation compensation data may be first determined, that is, at least two initial process parameter characteristics affecting the deviation compensation data may be determined. The at least two initial process parameter characteristics may include but are not limited to:

[0139] 1) Preset front layer process information. The preset front layer process information is for the preset front layer, which is the process information of the preset front layer before preparing the preset current layer. Furthermore, the preset front layer process information can be a process step that affects the asymmetry of the preset front layer OVL mask. Exemplarily, the preset front layer process information may include but is not limited to: the site required to prepare the preset front layer, which machines need to be used, the temperature data used by each machine, the preset front layer film deposition (film dep) thickness, the preset front layer etching process (etch process), the preset front layer high temperature process, the preset front layer CMP process, etc. Among them, the preset current layer is the current layer formed in the process of photolithography of the preset current batch of wafers.

[0140] 2) Machine status information of the preset back-end process. The preset back-end process is for the preset current layer, which is the process after the preset current layer is photolithographically processed. Furthermore, the preset back-end process may be a process that affects the transfer of the graphics of the preset current layer to the hard mask. The "machine status information" may refer to the status information of the machine used in the preset back-end process. Exemplarily, the "preset back-end process" may include but is not limited to: etching process, thin film deposition process, interconnection process, testing process, packaging process, etc. Exemplarily, the "machine status information" may refer to the temperature information, working time information, etc. of the machine used in the preset back-end process.

[0141] Since the number of at least two initial process parameter features is large, if each initial process parameter feature is involved in the model training process, the volume of model training will be large, thereby increasing the cost of MTD compensation and reducing the efficiency of MTD compensation. Based on this, the preset process parameter features whose correlation degree with the deviation compensation data is greater than the preset correlation threshold and the preset deviation compensation data corresponding to the preset process parameter features can be screened out from at least two initial process parameter features, and model training can be carried out based on this, thereby reducing the volume of the model, reducing the cost of MTD compensation and the efficiency of MTD compensation.

[0142] Figure 5 This is a flow diagram of a training process of a deviation compensation model according to an exemplary embodiment. Figure 2 ,like Figure 5 As shown, in an optional embodiment, the MTD data under different influencing factors can be collected by using the controlled variable method (DOE). Accordingly, in the above step S01, the above-mentioned acquisition of the preset process parameter characteristics and the preset deviation compensation data can include:

[0143] S011. Determine at least two initial process parameter characteristics in a wafer processing process.

[0144] S012. In the wafer processing process, adjust any initial process parameter feature, fix other initial process parameter features, perform post-development inspection on the wafer after development processing, and perform post-etching inspection on the wafer after etching, and generate deviation compensation data corresponding to any of the initial process parameter features based on the post-development inspection results and the post-etching inspection results.

[0145] S013. Screening any of the initial process parameter characteristics and the deviation compensation data corresponding to the any of the initial process parameter characteristics to obtain the preset process parameter characteristics and the preset deviation compensation data.

[0146] In this embodiment, after determining at least two initial process parameter characteristics, a control variable method can be used to sequentially adjust any of the initial process parameter characteristics while fixing the other initial process parameter characteristics to process the wafer. After the development process, the developed wafer is subjected to a post-development inspection process (ADI) to obtain a post-development inspection result (ADIOVL). After the etching process (AEI), the etched wafer is subjected to an etching process (AEIOVL). The difference between AEIOVL and ADIOVL is then calculated to obtain deviation compensation data corresponding to any of the initial process parameter characteristics.

[0147] Figure 6 is a schematic diagram showing the influence of the back-end etching process of the current layer on the deviation compensation data according to an exemplary embodiment. Figure 7 FIG. 1 is a schematic diagram showing the influence of the deposition thickness of a front film on the deviation compensation data according to an exemplary embodiment. Figure 8 FIG. 1 is a schematic diagram illustrating the influence of a front-layer CMP process on deviation compensation data according to an exemplary embodiment.

[0148] Take the deviation compensation data as the deviation compensation image (MTD map) as an example, Figure 6 As shown in the figure, when other initial process parameter characteristics are fixed and the radio frequency (RF) time of the back-end etching process of the current layer is adjusted, the MTD map gradually changes as the RF time increases, and the wafer edge gradually changes from an inward-contracting map to an outward-diverging map. Figure 7 As shown in Figure 2, when other initial process parameter characteristics are fixed and the thickness of the front film deposition is adjusted, the MTD map changes significantly with the thickness of the front film deposition. For example, Figure 7 When the film deposition thickness is 1, the value of the wafer edge is larger, and when the film deposition thickness is 2, the value of the wafer edge is smaller. Figure 8 As shown in Figure 2, when other initial process parameter characteristics are fixed and the front layer CMP process is adjusted, the MTD map changes significantly with the change of CMP optical critical dimension (OCD) performance, for example, Figure 8 Under the optical critical dimension 1, the value at the edge of the wafer is larger, and under the optical critical dimension 2, the value at the edge of the wafer is smaller.

[0149] Optionally, in order to reduce the size of the model, reduce the cost of MTD compensation and the efficiency of MTD compensation, any of the initial process parameter characteristics and the deviation compensation data corresponding to any of the initial process parameter characteristics can be screened to obtain the preset process parameter characteristics and the preset deviation compensation data.

[0150] Since the above-mentioned influencing factors are determined according to the generation mechanism of MTD, they include both preset front-layer process information and preset back-end process machine status information, so that the deviation compensation model can be trained based on the front-end and back-end process parameter characteristics other than the lithography process, so that the trained deviation compensation model can accurately predict the deviation compensation of the front-end and back-end process parameter characteristics during actual use, so as to dynamically add the front-end and back-end process parameter characteristics other than the lithography process to the OVL correction of the exposure machine, ensuring that each batch of wafers is produced with a better MTD compensation value under real-time data, achieving better OVL performance, and thus reducing the yield loss caused by OVL deviation in the mass production process; in addition, instead of involving each initial process parameter characteristic and its corresponding preset deviation compensation data in the model training process, the preset process parameter characteristics and the preset deviation compensation data are screened out for model training, thereby reducing the size of the model, reducing the cost of MTD compensation and the efficiency of MTD compensation.

[0151] In an exemplary embodiment, in the above step S013, the above screening of any of the initial process parameter characteristics and the deviation compensation data corresponding to any of the initial process parameter characteristics to obtain the preset process parameter characteristics and the preset deviation compensation data may include:

[0152] Obtaining corresponding deviation compensation data from the at least two initial process parameter characteristics to meet the preset process parameter characteristics of the preset deviation threshold; obtaining corresponding deviation compensation data from the at least two initial process parameter characteristics to meet the preset process parameter characteristics of the preset deviation condition.

[0153] In this embodiment, data preprocessing, feature extraction and feature selection may be performed on the at least two initial process parameter features in sequence.

[0154] Optionally, the data preprocessing may include, but is not limited to, cleaning, denoising, and processing the at least two initial process parameter features to ensure data accuracy and consistency, thereby obtaining the initial process parameter features after data preprocessing. Exemplary preprocessing methods may include deleting outliers, filling missing values, and data smoothing.

[0155] Optionally, the feature extraction may refer to extracting relevant features related to the process steps and MTD from the initial process parameter features after data preprocessing to obtain the extracted initial process parameter features. Exemplarily, the statistical features of the front-end and back-end process parameters may be extracted from the initial process parameter features. The statistical features of the front-end and back-end process parameters may include which processes were performed in the front-end layer, which processes were performed in the back-end process, time series features, etc. Among them, the time series features refer to the execution order of the initial process parameter features.

[0156] Optionally, feature selection may involve using statistical methods or machine learning-based methods to obtain process parameter features whose corresponding deviation compensation data satisfies a preset deviation threshold from the extracted initial process parameter features, thereby obtaining preset process parameter features. Statistical methods may include, but are not limited to, simple statistical methods, sampling techniques, discriminant analysis, regression analysis, factor analysis, etc.; machine learning methods may include, but are not limited to, regularized learning methods, decision trees, etc.

[0157] Optionally, “the corresponding deviation compensation data satisfies the preset deviation condition” may mean that the influence degree of “the corresponding deviation compensation data” on the OVL MTD is greater than a preset degree threshold. Figure 6 When other initial process parameter characteristics are fixed and the RF time of the back-end etching process of the current layer is adjusted, the MTD map gradually changes with the extension of the RF time, and the map of the wafer edge gradually changes from an inward-contracting map to an outward-diverging map, indicating that the "RF time of the back-end etching process of the current layer" has a greater impact on the OVL MTD and can be used as a preset process parameter characteristic. Similarly, according to Figure 7 and Figure 8 , the front layer deposition thickness and the front layer CMP process can be used as preset process parameter features.

[0158] Optionally, after the preset process parameter characteristics are determined, the preset deviation compensation data corresponding to the preset process parameter characteristics may be screened out from the existing deviation compensation data.

[0159] Therefore, according to the corresponding deviation compensation data, the preset process parameter characteristics and the corresponding preset deviation compensation data whose corresponding deviation compensation data meet the preset deviation conditions can be screened out from at least two initial process parameter characteristics, so that the preset process characteristics finally screened out are characteristics whose influence on OVL MTD is greater than the preset degree threshold, which can not only reduce the size of the model, but also reduce the cost and efficiency of MTD compensation.

[0160] S02. Input the preset process parameter characteristics into the initial model for deviation compensation processing to obtain deviation compensation data output by the initial model.

[0161] S03. Adjust the model parameters of the initial model according to the difference between the deviation compensation data output by the initial model and the preset deviation compensation data, until the difference meets the preset difference condition or the number of initial model training times meets the preset number condition, and obtain the deviation compensation model.

[0162] In this embodiment, after screening to obtain the preset process parameter characteristics and preset deviation compensation data, the preset process parameter characteristics and preset deviation compensation data can be input into the initial model. The initial model performs deviation compensation processing on the preset process parameter characteristics to obtain the deviation compensation data output by the initial model. Next, the difference between the deviation compensation data output by the initial model and the preset deviation compensation data is calculated. If the difference does not meet the preset difference condition or the number of training times does not meet the preset number condition, the parameters of the initial model are adjusted based on the difference until the difference meets the preset difference condition or the number of initial model training times meets the preset number condition, thereby obtaining a trained deviation compensation model.

[0163] Optionally, the “difference satisfies a preset difference condition” may mean that the difference is minimal.

[0164] In an exemplary embodiment, in the above step S02, inputting the preset process parameter characteristics into the initial model for deviation compensation processing to obtain the deviation compensation data output by the initial model may include:

[0165] Since the initial process parameter characteristics include the machine status information of the preset back-end process and the preset front-layer process information; the preset front-layer process information is the process information of the preset front layer before preparing the preset current layer, and the preset back-end process is the process after the preset current layer lithography. The preset process parameter characteristics are obtained by screening from the initial process parameter characteristics, so the preset process parameter characteristics can also include the machine status information of the preset back-end process and the preset front-layer process information. The preset process parameter characteristics can be divided into the following dimensions: 1) preset front-layer and preset current layer etch process parameters, such as etch machine, chamber, RF hours, etch critical dimension (etch CD), etch rate, etc.; 2) preset front-layer film thickness information, such as OCD value after film dep, OCD value after CMP; 3) etch tool (Etch Tool) and Chamber information used in the high-temperature steps experienced by the preset front and back layers.

[0166] Figure 9 is a structural diagram of an initial model according to an exemplary embodiment. Figure 9As shown, the structural diagram of the initial model may include an input layer, several hidden layers, and an output layer. The preset process parameter features after dimension division can be input into the input layer of the initial model for feature extraction and data conversion processing. Among them, feature extraction mainly uses some specific algorithms to extract useful features, and data conversion processing mainly converts the input data into vector or matrix form for subsequent calculation and processing. In addition, data conversion can preserve the structure and relationship of the data and provide more information to the network. The features output by the input layer are then input into the hidden layer. The hidden layer can map the features output by the input layer to another space through a function to perform deviation compensation processing on the features output by the input layer to obtain a deviation compensation image (MTD map). The hidden layer then converts the MTD map into vector information of a preset dimension (for example, m*1 dimension) that can be compensated by the scanner machine, thereby compressing the calculation time of the model and obtaining the final deviation compensation data. The output layer adjusts the model parameters of the initial model according to the difference between the deviation compensation data and the preset deviation compensation data until the difference meets the preset difference condition or the number of initial model training times meets the preset number condition.

[0167] Since the preset process parameter characteristics whose impact on OVL MTD is greater than the preset threshold and their corresponding preset deviation compensation data are used as data for training the deviation compensation model, the front-end and back-end process parameter characteristics other than the lithography process are added to the training process, so that the trained deviation compensation model can accurately predict the corresponding better deviation compensation data for different process parameter characteristics, thereby improving the prediction efficiency and accuracy of the deviation compensation data, and further ensuring that each batch of wafers is produced with the better MTD compensation value under real-time data, achieving better OVL performance, and reducing the yield loss caused by OVL deviation during mass production.

[0168] It should be noted that the above step S111 can be implemented in various ways, which are not specifically limited.

[0169] In one embodiment, in the step S111, the step of obtaining the status information of the back-end process machine may include:

[0170] When the tool status information of the back-end process is predetermined in the photolithography process of the current layer, the predetermined tool status information of the back-end process is acquired.

[0171] In another embodiment, in the step S111, the step of obtaining the status information of the downstream process machine may include:

[0172] In the case where the tool status information of the back-end process is not predetermined during the photolithography process of the current layer, the tool status information of all tools of the back-end process is obtained, and the tool status information of the back-end process is generated according to the tool status information of all tools.

[0173] In this embodiment, if the machine status information of the back-end process is predetermined during the photolithography process of the current layer, the predetermined machine status information of the back-end process is directly obtained. For example, if the etching device 2 in the etching tool 1 is predetermined to be used in the photolithography process of the current layer, the machine status information of the etching device 2 is directly obtained.

[0174] If the machine status information of the subsequent process is not predetermined during the lithography process of the current layer, the machine status information of all the machines that the wafer of the subsequent process will pass through can be obtained, and the mean of the machine status information of all the machines that the wafer will pass through can be calculated to obtain the machine status information of the subsequent process. Optionally, the mean may include but is not limited to: arithmetic mean, weighted mean, etc. For example, if the machine status information includes temperature information, the arithmetic mean, weighted mean, etc. of the temperature information of all the machines that the wafer will pass through can be calculated to obtain the machine status information of the subsequent process. In this way, the machine status information of the subsequent process can be directly determined during the lithography process of the previous layer, or the mean of the machine status information of all the machines that the wafer will pass through can be directly used as the machine status information of the subsequent process, which can improve the efficiency of obtaining the machine status information of the subsequent process, thereby improving the efficiency of the model calculation.

[0175] It should be noted that the above step S12 can be implemented in various ways, which are not specifically limited.

[0176] In an optional embodiment, in the above step S12, the inputting of the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain the predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics includes:

[0177] The front-end and back-end process parameter characteristics are input into the deviation compensation model for deviation compensation processing to obtain a deviation compensation image, and the deviation compensation image is converted into vector information of a preset dimension to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics.

[0178] In this embodiment, the front-end and back-end process parameter characteristics can be first divided into the following dimensions: 1) the etch process parameters of the front and current layers, such as the etch machine, chamber, RF hours, etch critical dimension (etch CD), etch rate, etc.; 1) the front layer film thickness information, such as the OCD value after film dep, the OCD value after CMP; 3) the tool and chamber information used in the high-temperature steps experienced by the front and back layers.

[0179] Similar to the initial model, the trained deviation compensation model can also include an input layer, several hidden layers and an output layer. The dimensionally divided front-end and back-end process parameter features can be input into the input layer of the deviation compensation model for feature extraction and data conversion processing. The features output by the input layer are then input into the hidden layer, which maps the features output by the input layer to another space through a function to perform deviation compensation processing on the features output by the input layer, and obtains a deviation compensation image (MTD map). The hidden layer then converts the MTD map into vector information of a preset dimension (for example, m*1 dimension) that can be compensated by the scanner machine, and obtains the predicted deviation compensation data corresponding to the front-end and back-end process parameter features. The output layer outputs the predicted deviation compensation data corresponding to the front-end and back-end process parameter features. Since the MTD map can be converted into vector information of a preset dimension that can be compensated by the scanner machine, the calculation time of the model can be compressed, thereby improving the calculation efficiency of the model.

[0180] In another optional embodiment, in the above step S12, the inputting of the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain the predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics includes:

[0181] The front-end and back-end process parameter characteristics are input into the deviation compensation model for deviation compensation processing to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics and the confidence level of the predicted deviation compensation data.

[0182] In this embodiment, the front-end and back-end process parameter characteristics can be first divided into the aforementioned dimensions. The dimensionally divided front-end and back-end process parameter characteristics are then sequentially applied to the input layer, several hidden layers, and output layer of the deviation compensation model to obtain predicted deviation compensation data and the confidence level of the predicted deviation compensation data. The confidence level is used to indicate the credibility of the predicted deviation compensation data.

[0183] Accordingly, in the above step S13, when the predicted deviation compensation data corresponding to the first process parameter feature in the front-end and back-end process parameter features does not meet the preset condition, the wafers of the current batch are processed based on the first process parameter feature, and the processed wafers are subjected to post-development inspection and post-etching inspection, and the real-time deviation compensation data corresponding to the first process parameter feature is generated according to the post-development inspection results and the post-etching inspection results, which may include:

[0184] When the confidence level of the predicted deviation compensation data corresponding to the first process parameter characteristic does not meet the preset conditions, the current batch of wafers is processed based on the first process parameter characteristic, and the processed wafers are subjected to post-development inspection and post-etching inspection, and real-time deviation compensation data corresponding to the first process parameter characteristic is generated according to the difference between the post-development inspection results and the post-etching inspection results.

[0185] In this embodiment, since the model not only outputs the predicted deviation compensation data, but also outputs the confidence of the predicted deviation compensation data, it is possible to directly determine whether the confidence meets the preset conditions. If not, it means that the first process parameter feature was not used in the process of training the MTD model. Based on this, data can be re-collected to trigger the current batch of wafers to perform ADI and AEI in-station measurements under the first process parameters, and calculate the difference between AEI and ADI to collect real-time deviation compensation data (i.e., real-time MTD value).

[0186] It should be noted that “the confidence level does not meet the preset conditions” may mean that the confidence level is not within a preset confidence level range, that is, the confidence level is unacceptable.

[0187] Since the confidence level can accurately and quickly determine whether the predicted deviation compensation data corresponding to the first process parameter feature meets the preset conditions, the accuracy and efficiency of generating real-time deviation compensation data can be improved, thereby improving the efficiency of model correction, ensuring that each batch of wafers is produced with the best MTD compensation value under real-time data, achieving better OVL performance, and thus reducing the yield loss caused by OVL deviation during mass production; and by judging whether it is necessary to generate real-time deviation compensation data through confidence level, the measurement time of AEIOVL can also be effectively shortened.

[0188] In an optional embodiment, the above method further includes:

[0189] When the predicted deviation compensation data corresponding to the second process parameter feature in the front-end and back-end process parameter features meets a preset condition, processing the wafers of the current batch based on the second process parameter feature and the predicted deviation compensation data corresponding to the second process parameter feature, and performing a post-development inspection on the processed wafers to obtain a post-development inspection result corresponding to the second process parameter feature;

[0190] Among them, the second process parameter feature is a process parameter feature in the front-end and back-end process parameter features except the first process parameter feature, and the post-development inspection result corresponding to the second process parameter feature is used to indicate that the overlay deviation of the processed wafer meets the preset deviation threshold.

[0191] In this embodiment, the number of front-end and back-end process parameter features can be at least two. When the predicted deviation compensation data corresponding to the second process parameter feature other than the first process parameter feature in the front-end and back-end process parameter features meets the preset conditions, it indicates that the second process parameter feature is used in the process of training the MTD model. Based on this, there is no need to re-collect data for the second process parameter feature, and the current batch of wafers can be processed directly based on the second process parameter feature. During the processing, the overlay deviation is controlled by the predicted deviation compensation data corresponding to the second process parameter feature, so that the overlay deviation of the wafer after deviation compensation meets the preset deviation threshold, thereby ensuring that each batch of wafers is produced with a better MTD compensation value under real-time data, achieving better OVL performance, and thereby reducing the yield loss caused by OVL deviation during mass production.

[0192] In another optional embodiment, in the above step S12, the inputting of the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain the predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics includes:

[0193] The front-end and back-end process parameter characteristics are input into the deviation compensation model for deviation compensation processing to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics and the confidence level of the predicted deviation compensation data.

[0194] Accordingly, when the predicted deviation compensation data corresponding to the second process parameter feature in the front-end and back-end process parameter features meets a preset condition, the wafers of the current batch are processed based on the second process parameter feature and the predicted deviation compensation data corresponding to the second process parameter feature, and the processed wafers are subjected to post-development inspection to obtain the post-development inspection results corresponding to the second process parameter feature, which may include:

[0195] If the confidence level of the predicted deviation compensation data corresponding to the second process parameter characteristic in the front-end and back-end process parameter characteristics does not meet a preset condition, the wafers of the current batch are processed based on the second process parameter characteristic and the predicted deviation compensation data corresponding to the second process parameter characteristic, and the processed wafers are subjected to a post-development inspection to obtain a post-development inspection result corresponding to the second process parameter characteristic. Because the confidence level can accurately and quickly determine whether the predicted deviation compensation data corresponding to the second process parameter characteristic meets the preset condition, the accuracy and efficiency of generating real-time deviation compensation data can be improved, thereby improving the efficiency of model correction. Furthermore, determining whether real-time deviation compensation data needs to be generated based on the confidence level can effectively reduce AEIOVL measurement time.

[0196] The following is an overall description of the above-mentioned method for processing overlay deviation:

[0197] Figure 10 A schematic diagram of a method for processing overlay deviation according to an exemplary embodiment is shown. Figure 3 ,like Figure 10 As shown, the overlay deviation processing method may include:

[0198] S21: Training bias compensation model.

[0199] 1) Based on the MTD generation mechanism, factors influencing the deviation compensation data are first determined, namely, at least two initial process parameter characteristics influencing the deviation compensation data are determined. These at least two initial process parameter characteristics may include, but are not limited to: preset front-end process information and preset back-end process machine status information. For details, please refer to the above-mentioned step S101 and will not be repeated here.

[0200] 2) Using the control variable method, sequentially adjust any initial process parameter characteristics while fixing the other initial process parameter characteristics to process the wafer. After the development process, perform a post-development inspection on the developed wafer to obtain a post-development inspection result. After the etching process, perform an etching process on the etched wafer. Next, calculate the difference between AEIOVL and ADIOVL to obtain deviation compensation data corresponding to any initial process parameter characteristics. Please refer to steps S011-S013 for details and will not be repeated here.

[0201] 3) Inputting the preset process parameter characteristics into the initial model for deviation compensation processing to obtain deviation compensation data output by the initial model. For details, please refer to step S02 and will not be repeated here.

[0202] 4) Adjusting the model parameters of the initial model based on the difference between the deviation compensation data output by the initial model and the preset deviation compensation data until the difference satisfies a preset difference condition or the number of initial model training times satisfies a preset number condition, thereby obtaining the deviation compensation model. For details, see step S03 and will not be repeated here.

[0203] S22: During the photolithography process for the current batch of wafers, the front-end and back-end process parameter characteristics and the deviation compensation model associated with the current layer are obtained. For details, please refer to step S11 and will not be repeated here.

[0204] S23: Input the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics. For details, please refer to step S12 and will not be repeated here.

[0205] S24: If the confidence level of the predicted deviation compensation data corresponding to a first process parameter characteristic among the front-end and back-end process parameter characteristics does not meet a preset condition, the current batch of wafers is processed based on the first process parameter characteristic, and the processed wafers are subjected to post-development inspection and post-etching inspection. Real-time deviation compensation data corresponding to the first process parameter characteristic is generated based on the post-development inspection and post-etching inspection results. For details, please refer to the above step S13 and will not be repeated here.

[0206] S25. Correct the deviation compensation model according to the real-time deviation compensation data to obtain a corrected deviation compensation model. For details, please refer to the above step S14, which will not be repeated here.

[0207] S26: When the confidence level of the predicted deviation compensation data corresponding to the second process parameter feature in the front-end and back-end process parameter features meets a preset condition, the wafers of the current batch are processed based on the second process parameter feature and the predicted deviation compensation data corresponding to the second process parameter feature, and the processed wafers are inspected after development to obtain a post-development inspection result corresponding to the second process parameter feature. The second process parameter feature is a process parameter feature in the linked process parameter feature other than the first process parameter feature. The post-development inspection result corresponding to the second process parameter feature is used to indicate that the overlay deviation of the processed wafer meets a preset deviation threshold.

[0208] The following example illustrates the above-mentioned method for processing overlay deviation:

[0209] Figure 11 A schematic diagram of a method for processing overlay deviation according to an exemplary embodiment is shown. Figure 4 ,like Figure 11As shown, the overlay deviation processing method can be executed by a cloud computing device.

[0210] The front-end and back-end process parameter characteristics associated with the current layer include the back-end process machine status information and the front-end process information.

[0211] The front layer process information may be process information for preparing the front layer, and specifically may include process information corresponding to Chamber 1 under Etch tool 1 and process information corresponding to Chamber 3 under Chemical Vapor Deposition (CVD) tool 1.

[0212] The back-end process tool status information is the tool status information of the back-end process for preparing the current layer and located after the current layer photolithography, and may specifically include: the tool status information of Chamber 4 under Etch tool 4.

[0213] During the process of photolithography on the current batch of wafers, the cloud computing device can obtain the "process information corresponding to Chamber 1 and Chamber 3" and the "machine status information of Chamber 4", input the "process information corresponding to Chamber 1 and Chamber 3" and the "machine status information of Chamber 4" into the deviation compensation model for deviation compensation processing, and obtain the predicted deviation compensation data corresponding to the "process information corresponding to Chamber 1 and Chamber 3" and the "machine status information of Chamber 4". When the predicted deviation compensation data corresponding to the first process parameter feature in the front-end and back-end process parameter features does not meet the preset conditions, the cloud computing device outputs an instruction to measure AEIOVL, controls the corresponding equipment to process the current batch of wafers based on the first process parameter feature, and performs post-development inspection and post-etching inspection on the processed wafers, and generates real-time deviation compensation data (for example, real-time MTD map) corresponding to the first process parameter feature according to the post-development inspection results and the post-etching inspection results. The cloud computing device obtains the real-time deviation compensation data, and corrects the deviation compensation model based on the real-time deviation compensation data to obtain a corrected deviation compensation model.

[0214] When the predicted deviation compensation data corresponding to the second process parameter feature in the front-end and back-end process parameter features meets the preset conditions, the cloud computing device controls the corresponding device to process the current batch of wafers based on the second process parameter feature and the predicted deviation compensation data corresponding to the second process parameter feature, and performs post-development inspection on the processed wafers to obtain the post-development inspection results corresponding to the second process parameter feature.

[0215] As a result, front-end and back-end process parameter characteristics outside of the lithography process can be dynamically incorporated into the exposure tool's OVL correction, ensuring that each batch of wafers is produced with the optimal MTD compensation value based on real-time data, achieving better OVL performance and thereby reducing yield losses caused by OVL deviations during mass production.

[0216] The embodiment of the present application also provides an overlay deviation processing device, Figure 12 FIG. 1 is a block diagram of an overlay deviation processing device according to an exemplary embodiment. Figure 12 As shown, the device includes:

[0217] The feature model acquisition module 31 is used to acquire the front-end and back-end process parameter features and the deviation compensation model associated with the current layer during the photolithography process of the current batch of wafers; the deviation compensation model is obtained by training an initial model based on the preset process parameter features and the preset deviation compensation data corresponding to the preset process parameter features;

[0218] A compensation processing module 32 is used to input the front-end and back-end process parameter characteristics into the deviation compensation model to perform deviation compensation processing, and obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics;

[0219] A real-time data generating module 33 is configured to process the wafers of the current batch based on the first process parameter characteristic among the front-end and back-end process parameter characteristics, and perform post-development inspection and post-etching inspection on the processed wafers, and generate real-time deviation compensation data corresponding to the first process parameter characteristic based on the post-development inspection and post-etching inspection results, if the predicted deviation compensation data corresponding to the first process parameter characteristic among the front-end and back-end process parameter characteristics do not meet a preset condition;

[0220] The correction module 34 is used to correct the deviation compensation model according to the real-time deviation compensation data to obtain a corrected deviation compensation model; the corrected deviation compensation model is used to perform deviation compensation on the next batch of wafers.

[0221] In an optional embodiment, the feature model acquisition module includes:

[0222] The status process information acquisition module is used to obtain the machine status information and front-layer process information of the back-end process; the front-layer process information is the process information of the front layer before preparing the current layer, and the back-end process is the process after the current layer is photolithographically processed.

[0223] The front-end and back-end process parameter generation module is used to generate the front-end and back-end process parameter characteristics according to the front-end process information and the machine status information of the back-end process.

[0224] In an optional embodiment, the device further comprises:

[0225] A feature compensation data acquisition module, used to acquire preset process parameter features and the preset deviation compensation data;

[0226] An input module, configured to input the preset process parameter characteristics into the initial model for deviation compensation processing, and obtain deviation compensation data output by the initial model;

[0227] An adjustment module is used to adjust the model parameters of the initial model according to the difference between the deviation compensation data output by the initial model and the preset deviation compensation data, until the difference meets the preset difference condition or the number of initial model training times meets the preset number condition, thereby obtaining the deviation compensation model.

[0228] In an optional embodiment, the feature compensation data acquisition module includes:

[0229] an initial process parameter characteristic determination unit, configured to determine at least two initial process parameter characteristics in a wafer processing process; the initial process parameter characteristics comprising machine status information of a preset back-end process and preset front-layer process information; the preset front-layer process information being process information of a preset front layer before preparing a preset current layer, the preset back-end process being a process after photolithography of a preset current layer, and the preset current layer being a current layer formed during photolithography of a preset current batch of wafers;

[0230] a variable control unit for adjusting any initial process parameter characteristic and fixing other initial process parameter characteristics during the wafer processing process, performing a post-development inspection process on the wafer after development processing, and performing a post-etching inspection process on the wafer after etching, and generating deviation compensation data corresponding to any of the initial process parameter characteristics based on the post-development inspection results and the post-etching inspection results;

[0231] The screening unit is used to screen any of the initial process parameter characteristics and the deviation compensation data corresponding to any of the initial process parameter characteristics to obtain the preset process parameter characteristics and the preset deviation compensation data.

[0232] In an optional embodiment, the screening unit includes:

[0233] a preset process parameter characteristic acquisition subunit, configured to acquire, from the at least two initial process parameter characteristics, the preset process parameter characteristics whose corresponding deviation compensation data satisfies a preset deviation condition;

[0234] The preset deviation compensation data acquisition subunit is used to acquire the preset deviation compensation data corresponding to the preset process parameter feature from the deviation compensation data corresponding to any initial process parameter feature.

[0235] In an optional embodiment, the state process information acquisition module includes:

[0236] The first machine status information acquiring unit is configured to acquire the predetermined machine status information of the back-end process when the machine status information of the back-end process is predetermined during the photolithography process of the current layer.

[0237] In an optional embodiment, the state process information acquisition module includes:

[0238] A second machine status information acquisition unit is configured to acquire machine status information of all machines in the back-end process when the machine status information of the back-end process is not predetermined in the photolithography process of the current layer;

[0239] The machine status information generating unit is used to generate the machine status information of the back-end process according to the machine status information of all the machines.

[0240] In an optional embodiment, the compensation processing module includes:

[0241] The first processing sub-unit is used to input the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain a deviation compensation image, convert the deviation compensation image into vector information of a preset dimension, and obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics.

[0242] In an optional embodiment, the compensation processing module includes:

[0243] The second processing sub-unit is used to input the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing, and obtain the predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics and the confidence of the predicted deviation compensation data.

[0244] Correspondingly, the real-time data generation module is used to process the current batch of wafers based on the first process parameter characteristics when the confidence level of the predicted deviation compensation data corresponding to the first process parameter characteristics does not meet the preset conditions, perform post-development inspection and post-etching inspection on the processed wafers, and generate real-time deviation compensation data corresponding to the first process parameter characteristics according to the difference between the post-development inspection results and the post-etching inspection results.

[0245] In an optional embodiment, the correction module includes:

[0246] a target deviation compensation data generating unit, configured to input the first process parameter characteristic and the real-time deviation compensation data into the deviation compensation model, so that the deviation compensation model performs deviation compensation processing on the first process parameter characteristic, and obtain target deviation compensation data output by the deviation compensation model;

[0247] A difference adjustment unit is used to adjust the model parameters of the deviation compensation model according to the difference between the real-time deviation compensation data and the target deviation compensation data until the difference meets a preset difference condition or the number of deviation compensation model training times meets a preset number condition, thereby obtaining a corrected deviation compensation model.

[0248] In an optional embodiment, the device further comprises:

[0249] a post-development inspection result generating module, configured to process the wafers of the current batch based on the second process parameter characteristic and the predicted deviation compensation data corresponding to the second process parameter characteristic in the front-end and back-end process parameter characteristics, and perform a post-development inspection on the processed wafers to obtain a post-development inspection result corresponding to the second process parameter characteristic, if the predicted deviation compensation data corresponding to the second process parameter characteristic in the front-end and back-end process parameter characteristics meets a preset condition;

[0250] Among them, the second process parameter feature is a process parameter feature in the front-end and back-end process parameter features except the first process parameter feature, and the post-development inspection result corresponding to the second process parameter feature is used to indicate that the overlay deviation of the processed wafer meets the preset deviation threshold.

[0251] In some embodiments, an embodiment of the present application also provides an electronic device for overlay deviation processing, the electronic device including a processor and a memory, the memory storing at least one instruction or at least one program, the at least one instruction or the at least one program being loaded by the processor and executing the overlay deviation processing method as described in the above embodiment.

[0252] In some embodiments, the embodiments of the present application also provide a computer-readable storage medium, which stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by a processor to implement the overlay deviation processing method described in the above embodiments.

[0253] In some embodiments, the present application further provides an overlay deviation processing system, the overlay deviation processing system comprising:

[0254] Post-development inspection tools;

[0255] Post-etch inspection tools;

[0256] photolithography tools for fabricating the current and previous layers;

[0257] The controller is used to execute the overlay deviation processing method as described in the above embodiment.

[0258] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0259] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for processing overlay deviation, characterized in that: The method comprises: During the photolithography process of the current batch of wafers, front-end and back-end process parameter characteristics and a deviation compensation model associated with the current layer are obtained; the deviation compensation model is obtained by training an initial model based on preset process parameter characteristics and preset deviation compensation data corresponding to the preset process parameter characteristics; Inputting the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics; When the predicted deviation compensation data corresponding to the first process parameter feature among the front-end and back-end process parameter features does not meet a preset condition, the wafers of the current batch are processed based on the first process parameter feature, and a post-development inspection and a post-etching inspection are performed on the processed wafers, and real-time deviation compensation data corresponding to the first process parameter feature is generated according to the post-development inspection results and the post-etching inspection results; The deviation compensation model is corrected according to the real-time deviation compensation data to obtain a corrected deviation compensation model; the corrected deviation compensation model is used to perform deviation compensation on the next batch of wafers.

2. The method for processing overlay deviation according to claim 1, wherein: The obtaining of the front-end and back-end process parameter characteristics associated with the current layer includes: Obtaining machine status information and front-layer process information of the back-end process; the front-layer process information is the process information of the front layer before the current layer is prepared, and the back-end process is the process after the current layer is photolithographically processed; The front-end and back-end process parameter characteristics are generated according to the front-end process information and the machine status information of the back-end process.

3. The method for processing overlay deviation according to claim 1, wherein: The training process of the bias compensation model includes: Acquiring the preset process parameter characteristics and the preset deviation compensation data; Inputting the preset process parameter characteristics into the initial model for deviation compensation processing to obtain deviation compensation data output by the initial model; The model parameters of the initial model are adjusted according to the difference between the deviation compensation data output by the initial model and the preset deviation compensation data until the difference meets the preset difference condition or the number of initial model training times meets the preset number condition, thereby obtaining the deviation compensation model.

4. The method for processing overlay deviation according to claim 3, wherein: The obtaining of the preset process parameter characteristics and the preset deviation compensation data includes: Determining at least two initial process parameter characteristics in a wafer processing process; the initial process parameter characteristics include machine status information of a preset back-end process and preset front-layer process information; the preset front-layer process information is process information of a preset front layer before preparing a preset current layer, the preset back-end process is a process after photolithography of the preset current layer, and the preset current layer is a current layer formed during photolithography of a preset current batch of wafers; In a wafer processing process, adjusting any initial process parameter characteristic, fixing other initial process parameter characteristics, performing a post-development inspection process on the wafer after development processing, and performing a post-etching inspection process on the wafer after etching, and generating deviation compensation data corresponding to any of the initial process parameter characteristics based on the post-development inspection results and the post-etching inspection results; The any one of the initial process parameter characteristics and the deviation compensation data corresponding to the any one of the initial process parameter characteristics are screened to obtain the preset process parameter characteristics and the preset deviation compensation data.

5. The method for processing overlay deviation according to claim 4, wherein: The screening of any of the initial process parameter characteristics and the deviation compensation data corresponding to the any of the initial process parameter characteristics to obtain the preset process parameter characteristics and the preset deviation compensation data includes: Obtaining the preset process parameter characteristics whose corresponding deviation compensation data satisfies the preset deviation condition from the at least two initial process parameter characteristics; The preset deviation compensation data corresponding to the preset process parameter feature is obtained from the deviation compensation data corresponding to any one of the initial process parameter features.

6. The method for processing overlay deviation according to claim 2, wherein: The acquisition of the machine status information of the back-end process includes: When the tool status information of the back-end process is predetermined in the photolithography process of the current layer, the predetermined tool status information of the back-end process is acquired.

7. The method for processing overlay deviation according to claim 2, wherein: The acquisition of the machine status information of the back-end process includes: When the tool status information of the back-end process is not predetermined in the photolithography process of the current layer, the tool status information of all tools of the back-end process is obtained; The tool status information of the back-end process is generated according to the tool status information of all the tools.

8. The method for processing overlay deviation according to any one of claims 1 to 7, characterized in that: The inputting the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics includes: The front-end and back-end process parameter characteristics are input into the deviation compensation model for deviation compensation processing to obtain a deviation compensation image, and the deviation compensation image is converted into vector information of a preset dimension to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics.

9. The method for processing overlay deviation according to any one of claims 1 to 7, characterized in that: The inputting the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing to obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics includes: Inputting the front-end and back-end process parameter characteristics into the deviation compensation model for deviation compensation processing, and obtaining predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics and the confidence level of the predicted deviation compensation data; When the predicted deviation compensation data corresponding to the first process parameter feature in the front-end and back-end process parameter features does not meet a preset condition, the wafers of the current batch are processed based on the first process parameter feature, and a post-development inspection and a post-etching inspection are performed on the processed wafers, and real-time deviation compensation data corresponding to the first process parameter feature is generated according to the post-development inspection results and the post-etching inspection results, including: When the confidence level of the predicted deviation compensation data corresponding to the first process parameter characteristic does not meet the preset conditions, the current batch of wafers is processed based on the first process parameter characteristic, and the processed wafers are subjected to post-development inspection and post-etching inspection, and real-time deviation compensation data corresponding to the first process parameter characteristic is generated according to the difference between the post-development inspection results and the post-etching inspection results.

10. The method for processing overlay deviation according to any one of claims 1 to 7, characterized in that: The step of correcting the deviation compensation model according to the real-time deviation compensation data to obtain a corrected deviation compensation model includes: Inputting the first process parameter characteristic and the real-time deviation compensation data into the deviation compensation model, so that the deviation compensation model performs deviation compensation processing on the first process parameter characteristic, and obtaining target deviation compensation data output by the deviation compensation model; The model parameters of the deviation compensation model are adjusted according to the difference between the real-time deviation compensation data and the target deviation compensation data until the difference meets a preset difference condition or the number of deviation compensation model trainings meets a preset number condition, thereby obtaining a corrected deviation compensation model.

11. The method for processing overlay deviation according to any one of claims 1 to 7, characterized in that: The method further comprises: When the predicted deviation compensation data corresponding to the second process parameter feature in the front-end and back-end process parameter features meets a preset condition, processing the wafers of the current batch based on the second process parameter feature and the predicted deviation compensation data corresponding to the second process parameter feature, and performing a post-development inspection on the processed wafers to obtain a post-development inspection result corresponding to the second process parameter feature; Among them, the second process parameter feature is a process parameter feature in the front-end and back-end process parameter features except the first process parameter feature, and the post-development inspection result corresponding to the second process parameter feature is used to indicate that the overlay deviation of the processed wafer meets the preset deviation threshold.

12. An overlay deviation processing device, characterized in that: The device comprises: A feature model acquisition module is used to acquire front-end and back-end process parameter features and a deviation compensation model associated with the current layer during the photolithography process of the current batch of wafers; the deviation compensation model is obtained by training an initial model based on preset process parameter features and preset deviation compensation data corresponding to the preset process parameter features; a compensation processing module, configured to input the front-end and back-end process parameter characteristics into the deviation compensation model for performing deviation compensation processing, and obtain predicted deviation compensation data corresponding to the front-end and back-end process parameter characteristics; a real-time data generation module, configured to, if predicted deviation compensation data corresponding to a first process parameter characteristic among the front-end and back-end process parameter characteristics does not satisfy a preset condition, process the wafers of the current batch based on the first process parameter characteristic, perform a post-development inspection and a post-etching inspection on the processed wafers, and generate real-time deviation compensation data corresponding to the first process parameter characteristic based on the post-development inspection results and the post-etching inspection results; The correction module is used to correct the deviation compensation model according to the real-time deviation compensation data to obtain a corrected deviation compensation model; the corrected deviation compensation model is used to perform deviation compensation on the next batch of wafers.

13. An electronic device for overlay deviation processing, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded by the processor and executes the overlay deviation processing method as described in any one of claims 1 to 11.

14. A computer scale storage medium, characterized in that The computer-readable storage medium stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by a processor to implement the overlay deviation processing method according to any one of claims 1 to 11.

15. A system for processing overlay deviation, characterized in that: include: Post-development inspection tools; Post-etch inspection tools; photolithography tools for fabricating the current and previous layers; A controller, configured to execute the overlay deviation processing method according to any one of claims 1 to 11.

Images