A mask phase type defect identification method, device, equipment and medium

Abstract

The present disclosure relates to a mask phase type defect identification method, device, equipment and medium, the method comprises: collecting the light field reflection amplitude distribution of the mask phase type defect, converting the light field reflection amplitude distribution into spatial image sample data of the mask defect; the spatial image sample data is divided into spatial image sample data training set, spatial image sample data test set according to pre-set proportion;The spatial image sample data training set is labeled for defect type;The corresponding relationship between the result of the defect type labeling and the spatial image sample data training set is trained to build a defect identification model;The spatial image sample data test set is input into the defect identification model to obtain a defect identification result.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of integrated circuits, and more particularly, to a mask phase type defect identification method, device, equipment and medium. BACKGROUND

[0002] A mask is a patterned master used in photolithography process in integrated circuit manufacturing. Mask defects seriously affect the yield of chip manufacturing in photolithography process.

[0003] Extreme ultraviolet lithography is currently internationally recognized as the most promising photolithography technology. Multilayer film defects are unique defects in extreme ultraviolet lithography masks. During mask manufacturing, phase type defects will be generated at the bottom of the mask substrate and the multilayer film due to particle contamination, physical deformation or incomplete chemical reactions in the material preparation process.

[0004] Phase type defects are divided into two types of protrusions and pits. In the spatial image intensity distribution, the edge of the protrusion defect enhances the center and weakens it, and the pit defect is the opposite, and there are some intensity oscillations at the outer edge of the defect. The size parameters of the phase type defect include the half-peak width w and the peak height h. By monitoring the type and size information of the phase type defect, the level of the mask manufacturing process can be effectively evaluated, and it can be determined whether the mask can be used in the production environment.

[0005] The spatial image of the phase type defect obtained by using the actinic detection device has the characteristics of small chromatic aberration change and blurred edge. Due to the influence of the current dark noise in the data acquisition process, there will be some noise points on the final image. Using the image classification algorithm based on traditional machine vision, it is difficult to determine a unified classification threshold to identify the defect image information of different sizes. The neural network uses the back propagation algorithm to automatically update the weight of each neuron, and contains a nonlinear activation function, which is suitable for processing multi-dimensional nonlinear classification problems. In the prior art, the defect type needs to be determined in advance, and then the corresponding neural network model is used to identify the size of the pit defect and the protrusion defect, which will occupy more computer memory. The existing technology also has the problems of large neural network model size and large amount of data required for training the neural network, which leads to slow computer program running speed and long sample data acquisition cycle. SUMMARY

[0006] To solve the technical problem that the defect identification method of the prior art cannot meet the user's needs.

[0007] To achieve the above technical purpose, the present disclosure provides a mask phase type defect identification method, which comprises:

[0008] Collecting the light field reflection amplitude distribution of the mask phase type defect, and converting the light field reflection amplitude distribution into spatial image sample data of the mask defect;

[0009] The spatial image sample data is divided into a spatial image sample data training set and a spatial image sample data test set according to a preset ratio;

[0010] The spatial image sample data is labeled by defect type;

[0011] The correspondence between the result of the defect type labeling and the spatial image sample data training set is trained to build a defect recognition model;

[0012] The spatial image sample data test set is input into the defect recognition model to obtain a defect recognition result.

[0013] Further, the conversion of the light field reflection amplitude distribution into the spatial image sample data of the mask defect further comprises:

[0014] The spatial image sample data is augmented by using an image enhancement algorithm;

[0015] The image enhancement algorithm comprises image flipping, image rotation, brightness adjustment and / or color adjustment.

[0016] Further, the defect recognition model is a convolutional neural network composed of at least three convolutional layers, at least three pooling layers, at least three fully connected layers and a softmax layer;

[0017] Batch normalization processing is performed between each of the convolutional layers and the pooling layers.

[0018] Further, the defect recognition model is a convolutional neural network composed of five convolutional layers, five pooling layers, at least three fully connected layers and a softmax layer;

[0019] The convolution kernel size of the convolutional layer is 3x3;

[0020] The pooling layer is an average pooling.

[0021] Further, the pooling layer after the third convolutional layer is an average pooling with a pooling window of 3x3 and a step of 3;

[0022] The pooling layers after the remaining four convolutional layers are all average poolings with a pooling window of 2x2 and a step of 2;

[0023] The at least three fully connected layers comprise:

[0024] a defect category fully connected layer, a defect half-height-width fully connected layer and a defect peak-height fully connected layer;

[0025] A Softmax function is added after the defect category fully connected layer as a regression function of the defect category;

[0026] Use the full-linking layer of defect half-height peak width as the regression function of defect half-height peak width.

[0027] Use the full-linking layer of defect peak height as the regression function of defect peak height.

[0028] Further, the light field reflection amplitude distribution of the mask phase defect is specifically collected as follows:

[0029] The collector collects the light field reflection light path of the working light source after the illumination element, irradiates to the to-be-detected mask, and then passes through the wave band sheet imaging magnification to obtain the light field reflection intensity amplitude distribution with the mask defect characteristics.

[0030] Further, the spatial image sample data test set specifically includes:

[0031] The spatial image sample data test subset and the spatial image sample data verification set;

[0032] The input of the spatial image sample data test set into the defect identification model to obtain the defect identification result is specifically:

[0033] The input of the spatial image sample data test subset into the defect identification model to obtain the defect identification result;

[0034] After the input of the spatial image sample data test subset into the defect identification model to obtain the defect identification result, the method further includes:

[0035] Judging whether the current defect identification model converges, if the model loss value of the current defect identification result is higher than a preset threshold, continuing to fine-tune the defect identification model and re-judging whether the current defect identification model converges;

[0036] If the model loss value of the current defect identification result is lower than the preset threshold, considering that the current defect identification model converges, stopping the fine-tuning of the defect identification model, and outputting the final defect identification model.

[0037] Further, the model loss value is specifically obtained by the following method:

[0038] For the classification problem of defects, using a cross-entropy function as a loss function to obtain a first loss value;

[0039] For the regression problem of defect half-height peak width and peak height, using a mean square error equation as a loss function to obtain two second loss values;

[0040] The first loss value and the two second loss values are weighted and averaged to obtain the model loss value.

[0041] Further,

[0042] If the current defect recognition model converges, the inputting the spatial image sample data test set into the defect recognition model to obtain a defect recognition result further includes:

[0043] Inputting the spatial image sample data verification set into the converged defect recognition model to obtain a final defect recognition result.

[0044] To achieve the above technical purposes, the present disclosure can also provide a computer storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the steps of the mask phase-type defect recognition method.

[0045] To achieve the above technical purposes, the present disclosure also provides an electronic device including a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the mask phase-type defect recognition method when executing the computer program.

[0046] The present disclosure has the following beneficial effects:

[0047] The present disclosure proposes a small-volume neural network model, and performs data enhancement on the collected data set, thereby reducing the data collection period. Meanwhile, the existing defect recognition method belongs to a two-stage structure, that is, the type of phase-type defect is first judged, and then the corresponding neural network model is selected for size recognition according to the judgment result. The neural network model proposed by the present disclosure does not need to judge the defect type in advance, the input layer of the network is the spatial image of the phase-type defect, and the output result is the type and size of the phase-type defect, thereby reducing the algorithm complexity and the computer program memory occupation, and in the neural network training process, random noise is added to the training data, which is closer to the real data collection situation in the industrial environment, thereby improving the robustness of the algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 A flowchart of the method of embodiment 1 of the present disclosure is shown;

[0049] Figure 2 A structural diagram of the defect recognition model of embodiment 1 of the present disclosure is shown;

[0050] Figure 3 A structural diagram of embodiment 3 of the present disclosure is shown. DETAILED DESCRIPTION

[0051] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In addition, in the following description, the description of well-known structures and techniques is omitted to avoid unnecessary confusion of the concept of the present disclosure.

[0052] Various structural diagrams according to embodiments of the present disclosure are shown in the drawings. These diagrams are not drawn to scale in which certain details are exaggerated for clarity and others omitted. The shapes and relative sizes of the various regions, layers, and their relative positions illustrated in the drawings are merely exemplary and may deviate in actuality due to manufacturing tolerances or technical limitations, and regions / layers with different shapes, sizes, and relative positions can be additionally designed according to actual needs by those skilled in the art.

[0053] Embodiment one:

[0054] As Figure 1 shown:

[0055] The present disclosure provides a mask phase type defect identification method, comprising:

[0056] S101: Collecting the light field reflection amplitude distribution of the mask phase type defect, and converting the light field reflection amplitude distribution into spatial image sample data of the mask defect;

[0057] S102: Dividing the spatial image sample data into a spatial image sample data training set and a spatial image sample data test set according to a preset ratio;

[0058] S103: Defect type labeling of the spatial image sample data;

[0059] S104: Training the corresponding relationship between the result of the defect type labeling and the spatial image sample data training set to build a defect identification model;

[0060] S105: Inputting the spatial image sample data test set into the defect identification model to obtain a defect identification result.

[0061] Further, the conversion of the light field reflection amplitude distribution into the spatial image sample data of the mask defect further comprises:

[0062] Sample data augmentation and enhancement of the spatial image sample data using an image enhancement algorithm;

[0063] The image enhancement algorithm includes image flipping, image rotation, brightness adjustment, and / or color adjustment.

[0064] Further, the defect identification model is a convolutional neural network composed of at least three convolutional layers, at least three pooling layers, at least three fully connected layers, and a softmax layer.

[0065] Batch normalization processing is performed between each of the convolutional layers and the pooling layers.

[0066] Further, the defect identification model is a convolutional neural network composed of five convolutional layers, five pooling layers, at least three fully connected layers, and a softmax layer.

[0067] The convolution kernel size of the convolutional layer is 3x3.

[0068] The pooling layer is average pooling.

[0069] Further, the pooling layer after the third convolutional layer is average pooling with a pooling window size of 3x3 and a step size of 3.

[0070] The pooling layers after the remaining four convolutional layers are all average pooling with a pooling window size of 2x2 and a step size of 2.

[0071] The at least three fully connected layers include:

[0072] a defect type fully connected layer, a defect half-height peak width fully connected layer, and a defect peak height fully connected layer.

[0073] A Softmax function is added after the defect type fully connected layer as a regression function for defect type.

[0074] The defect half-height peak width fully connected layer is used as a regression function for defect half-height peak width.

[0075] The defect peak height fully connected layer is used as a regression function for defect peak height.

[0076] Further, the collected phase-type mask defect light field reflection amplitude distribution specifically includes:

[0077] The collector collects the light field reflection path of the working light source after being irradiated to the mask to be detected through the illumination element and then imaged and enlarged by the waveband sheet, to obtain the light field reflection intensity amplitude distribution with mask defect characteristics.

[0078] Further, the spatial image sample data test set specifically includes:

[0079] a spatial image sample data test subset and a spatial image sample data validation set.

[0080] The input of the spatial image sample data test set into the defect identification model to obtain a defect identification result specifically includes:

[0081] The input of the spatial image sample data test subset into the defect identification model to obtain a defect identification result.

[0082] After the input of the spatial image sample data test subset into the defect identification model to obtain a defect identification result, it further includes:

[0083] determining whether the current defect recognition model converges, if the model loss value of the current defect recognition result is higher than a preset threshold, continuing to fine-tune the defect recognition model and re-determining whether the current defect recognition model converges;

[0084] If the model loss value of the current defect recognition result is lower than the preset threshold, it is considered that the current defect recognition model converges, the fine-tuning of the defect recognition model is stopped, and the final defect recognition model is output.

[0085] Further, the model loss value is specifically obtained by the following method:

[0086] For the classification problem of defects, a cross-entropy function is used as a loss function to obtain a first loss value;

[0087] For the regression problem of defect half-height peak width and peak height, a mean square error equation is used as a loss function to obtain two second loss values;

[0088] The first loss value and the second loss value are weighted and averaged to obtain the model loss value.

[0089] Further,

[0090] If the current defect recognition model converges, the method further comprises the following steps after inputting the spatial image sample data test set into the defect recognition model to obtain a defect recognition result:

[0091] Inputting a spatial image sample data verification set into the converged defect recognition model to obtain a final defect recognition result.

[0092] It should be stated that the defect attribute recognition method can be applied to all mask defect detection devices based on spatial image imaging principle. In the embodiment, an extreme ultraviolet mask defect detection system based on waveband plate is taken as an example.

[0093] Embodiment two:

[0094] The present disclosure can also provide a computer storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the mask phase type defect recognition method.

[0095] The computer storage medium of the present disclosure can be implemented by a semiconductor memory, a magnetic core memory, a magnetic drum memory or a magnetic disk memory.

[0096] Semiconductor memory, the main semiconductor storage element for computer, mainly has Mos and bipolar type. Mos element has high integration, simple process but slow speed. Bipolar element has complex process, large power consumption, low integration but fast speed. NMos and CMos make Mos memory start to occupy the main position in semiconductor memory. NMos has fast speed, such as 1K bit static random memory of Intel company has access time of 45ns. CMos has low power consumption, 4K bit CMos static memory has access time of 300ns. The above semiconductor memory is random access memory (RAM), that is, it can randomly read and write new content in working process. The semiconductor read only memory (ROM) can randomly read in working process but cannot write, which is used to store solidified program and data. ROM is divided into non-rewritable fuse type read only memory-PROM and rewritable read only memory-EPROM.

[0097] Magnetic core memory, which has low cost and high reliability, has more than 20 years of practical experience. Before the mid-1970s, magnetic core memory was widely used as main memory. Its storage capacity can reach more than 10 bits, and the fastest access time is 300ns. The typical magnetic core memory capacity in the world is 4MS-8MB, and the access period is 1.0-1.5μs. After the rapid development of semiconductor memory to replace magnetic core memory as the main memory, magnetic core memory can still be used as large capacity expansion memory.

[0098] Magnetic drum memory, an external storage of magnetic recording. Although its capacity is small, it is gradually replaced by magnetic disk memory, but it is still used as external storage for real-time process control computers and medium and large computers. In order to meet the needs of small and microcomputers, super small magnetic drums have appeared, which have small size, light weight, high reliability and easy use.

[0099] Magnetic disk memory, an external storage of magnetic recording. It has the advantages of both magnetic drum and magnetic tape memory, that is, its storage capacity is larger than that of magnetic drum, and its access speed is faster than that of magnetic tape memory, and it can be stored offline, so it is widely used as large capacity external storage in various computer systems. Magnetic disks are generally divided into two categories: hard disk and soft disk memory.

[0100] There are many types of hard disk storage devices. Structurally, they are divided into two types: replaceable and fixed. Replaceable disks have interchangeable platters, while fixed disks have fixed platters. Both replaceable and fixed disks have multi-platter and single-platter structures, and can be further divided into fixed-head and movable-head types. Fixed-head disks have smaller capacities, lower recording densities, and higher access speeds, but are more expensive. Movable-head disks have higher recording densities (up to 1000-6250 bits / inch), resulting in larger capacities, but their access speeds are relatively lower than fixed-head disks. Disk products can have storage capacities of several hundred megabytes, with a bit density of 6250 bits / inch and a track density of 475 tracks / inch. Multi-platter replaceable disk storage devices, due to their replaceable platters, offer very large independent capacity, and with high speeds, can store large amounts of information and are widely used in online information retrieval systems and database management systems.

[0101] Example 3:

[0102] This disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the mask phase-type defect identification method described above.

[0103] Figure 3 This is a schematic diagram of the internal structure of an electronic device in one embodiment. For example... Figure 3 As shown, the electronic device includes a processor, storage medium, memory, and network interface connected via a system bus. The storage medium stores an operating system, a database, and computer-readable instructions. The database may store control information sequences. When executed by the processor, the computer-readable instructions enable the processor to implement a mask-based phase-type defect identification method. The processor provides computational and control capabilities, supporting the operation of the entire computer device. The memory stores computer-readable instructions, which, when executed by the processor, enable the processor to perform a mask-based phase-type defect identification method. The network interface of the computer device is used for communication with a terminal. Those skilled in the art will understand that… Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0104] This electronic device includes, but is not limited to, smartphones, computers, tablets, wearable smart devices, artificial intelligence devices, and power banks.

[0105] The processor can be composed of integrated circuits in some embodiments, for example, can be composed of a single packaged integrated circuit, or can be composed of multiple packaged integrated circuits with the same function or different functions, including one or more central processing units (CPU), microprocessors, digital processing chips, graphics processors, combinations of various control chips, etc. The processor is the control core of the electronic device. It connects various components of the entire electronic device through various interfaces and lines, executes programs or modules stored in the memory (such as executing remote data read-write programs, etc.), and calls data stored in the memory to perform various functions and process data of the electronic device.

[0106] The bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize the connection and communication between the memory, the at least one processor, etc.

[0107] Figure 3 Only the electronic device with components is shown, and those skilled in the art can understand that, Figure 3 The structure shown does not constitute a limitation on the electronic device, and can include fewer or more components than shown, or combine certain components, or different component arrangements.

[0108] For example, although not shown, the electronic device can also include a power supply (such as a battery) for powering various components. Preferably, the power supply can be logically connected to the at least one processor through a power management device, so as to realize functions such as charge management, discharge management, and power consumption management through the power management device. The power supply can also include one or more direct current or alternating current power supplies, recharging devices, power supply fault detection circuits, power supply converters or inverters, power supply status indicators, etc. The electronic device can also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described here.

[0109] Further, the electronic device can also include a network interface. Optionally, the network interface can include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is usually used to establish a communication connection between the electronic device and other electronic devices.

[0110] Optionally, the electronic device can further include a user interface, which can be a display, an input unit such as a keyboard, and optionally, a standard wired interface, a wireless interface. Optionally, in some embodiments, the display can be an LED display, a liquid crystal display, a touch liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, or the like. The display can also be appropriately referred to as a display screen or a display unit, and is used to display information processed in the electronic device and to display a visualized user interface.

[0111] Further, the computer usable storage medium can mainly include a storage program area and a storage data area, wherein the storage program area can store an operating system, an application program required by at least one function, and the like; and the storage data area can store data created according to the use of the blockchain node, and the like.

[0112] In several embodiments provided by the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the above-described device embodiments are merely illustrative, and the division of the modules is merely a logical function division. In actual implementation, another division manner can be used.

[0113] The modules described as separated components can or can not be physically separated, and the components displayed as modules can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments.

[0114] In addition, each functional module in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of hardware plus software functional modules.

[0115] The above describes the embodiments of the present disclosure. However, these embodiments are only for illustrative purposes, and are not intended to limit the scope of the present disclosure. The scope of the present disclosure is defined by the appended claims and their equivalents. Without departing from the scope of the present disclosure, those skilled in the art can make various substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A reticle phase-type defect identification method, characterized by, The method comprises the following steps: Collecting a light field reflection amplitude distribution of a mask phase defect, and converting the light field reflection amplitude distribution into spatial image sample data of the mask defect; The step of collecting the light field reflection amplitude distribution of the mask phase defect specifically comprises: a collector collecting a light field reflection light path of a working light source after the working light source irradiates a to-be-detected mask and then passes through a wave band sheet imaging magnification, so as to obtain a light field reflection intensity amplitude distribution with a mask defect feature; The spatial image sample data is divided into a spatial image sample data training set and a spatial image sample data test set according to a preset ratio; Defect type labeling is performed on the spatial image sample data; A correspondence between a result of the defect type labeling and the spatial image sample data training set is trained, so as to build a defect recognition model; The spatial image sample data test set is input into the defect recognition model to obtain a defect recognition result; The defect recognition model is a convolutional neural network composed of five convolutional layers, five pooling layers, at least three full-link layers and a softmax layer; a convolution kernel size of the convolutional layers is 3*3; and the pooling layers are average pooling; A pooling window of a pooling layer after a third convolutional layer is 3*3, and a step length is 3; pooling windows of pooling layers after the other four convolutional layers are all 2*2, and the step length is 2; The at least three full-link layers comprise a defect type full-link layer, a defect half-height peak width full-link layer and a defect peak value height full-link layer; a Softmax function is added after the defect type full-link layer as a regression function of the defect type; The defect half-height peak width full-link layer is used as a regression function of the defect half-height peak width; and the defect peak value height full-link layer is used as a regression function of the defect peak value height.

2. The method of claim 1, wherein, After the step of converting the light field reflection amplitude distribution into the spatial image sample data of the mask defect, the following steps are further included: Sample data expansion and enhancement are performed on the spatial image sample data by using an image enhancement algorithm; The image enhancement algorithm comprises image flipping, image rotation, brightness adjustment and / or color adjustment.

3. The method of claim 1, wherein, The spatial image sample data test set specifically comprises: A spatial image sample data test subset and a spatial image sample data verification set; The step of inputting the spatial image sample data test set into the defect recognition model to obtain a defect recognition result specifically comprises: The spatial image sample data test subset is input into the defect recognition model to obtain a defect recognition result; It is judged whether the current defect recognition model converges; if a model loss value of the current defect recognition result is higher than a preset threshold, fine tuning training of the defect recognition model is continued, and it is re-judged whether the current defect recognition model converges; If the model loss value of the current defect recognition result is lower than the preset threshold, it is considered that the current defect recognition model converges, fine tuning training of the defect recognition model is stopped, and a final defect recognition model is output.

4. The method of claim 3, wherein, The model loss value is specifically obtained in the following manner: For a classification problem of the defect, a cross-entropy function is used as a loss function to obtain a first loss value; For a regression problem of the defect half-height peak width and the peak value height, a mean square error equation is used as a loss function to obtain two second loss values; and The model loss value is obtained by weighted average of the first loss value and the second loss value.

5. The method of claim 3 or 4, wherein, If the current defect identification model converges, the inputting of the aerial image sample data test set into the defect identification model to obtain a defect identification result further comprises: Inputting an aerial image sample data verification set into the converged defect identification model to obtain a final defect identification result.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps corresponding to the mask phase-type defect identification method in any one of claims 1-5 when executing the computer program.

7. A computer storage medium having stored thereon computer program instructions, wherein the computer program instructions are executable by a computer to cause the computer to perform the method according to any one of claims 1 to 6. The program instructions are executed by the processor to implement the steps corresponding to the mask phase-type defect identification method in any one of claims 1-5. The program instructions are executed by the processor to implement the steps corresponding to the mask phase-type defect identification method in any one of claims 1-5.

Images