CN111999992B - Training method and device for determining lithography light source model - Google Patents

Training method and device for determining lithography light source model Download PDF

Info

Publication number
CN111999992B
CN111999992B CN202010988995.2A CN202010988995A CN111999992B CN 111999992 B CN111999992 B CN 111999992B CN 202010988995 A CN202010988995 A CN 202010988995A CN 111999992 B CN111999992 B CN 111999992B
Authority
CN
China
Prior art keywords
light source
layout
specific size
determining
neural network
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010988995.2A
Other languages
Chinese (zh)
Other versions
CN111999992A (en
Inventor
马乐
韦亚一
张利斌
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Institute of Microelectronics of CAS
Original Assignee
Institute of Microelectronics of CAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Institute of Microelectronics of CAS filed Critical Institute of Microelectronics of CAS
Priority to CN202010988995.2A priority Critical patent/CN111999992B/en
Publication of CN111999992A publication Critical patent/CN111999992A/en
Application granted granted Critical
Publication of CN111999992B publication Critical patent/CN111999992B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G03—PHOTOGRAPHY; CINEMATOGRAPHY; ANALOGOUS TECHNIQUES USING WAVES OTHER THAN OPTICAL WAVES; ELECTROGRAPHY; HOLOGRAPHY
    • G03F—PHOTOMECHANICAL PRODUCTION OF TEXTURED OR PATTERNED SURFACES, e.g. FOR PRINTING, FOR PROCESSING OF SEMICONDUCTOR DEVICES; MATERIALS THEREFOR; ORIGINALS THEREFOR; APPARATUS SPECIALLY ADAPTED THEREFOR
    • G03F7/00—Photomechanical, e.g. photolithographic, production of textured or patterned surfaces, e.g. printing surfaces; Materials therefor, e.g. comprising photoresists; Apparatus specially adapted therefor
    • G03F7/70—Microphotolithographic exposure; Apparatus therefor
    • G03F7/70483—Information management; Active and passive control; Testing; Wafer monitoring, e.g. pattern monitoring
    • G03F7/70491—Information management, e.g. software; Active and passive control, e.g. details of controlling exposure processes or exposure tool monitoring processes
    • G03F7/705—Modelling or simulating from physical phenomena up to complete wafer processes or whole workflow in wafer productions
    • G—PHYSICS
    • G03—PHOTOGRAPHY; CINEMATOGRAPHY; ANALOGOUS TECHNIQUES USING WAVES OTHER THAN OPTICAL WAVES; ELECTROGRAPHY; HOLOGRAPHY
    • G03F—PHOTOMECHANICAL PRODUCTION OF TEXTURED OR PATTERNED SURFACES, e.g. FOR PRINTING, FOR PROCESSING OF SEMICONDUCTOR DEVICES; MATERIALS THEREFOR; ORIGINALS THEREFOR; APPARATUS SPECIALLY ADAPTED THEREFOR
    • G03F7/00—Photomechanical, e.g. photolithographic, production of textured or patterned surfaces, e.g. printing surfaces; Materials therefor, e.g. comprising photoresists; Apparatus specially adapted therefor
    • G03F7/70—Microphotolithographic exposure; Apparatus therefor
    • G03F7/70058—Mask illumination systems
    • G03F7/70091—Illumination settings, i.e. intensity distribution in the pupil plane or angular distribution in the field plane; On-axis or off-axis settings, e.g. annular, dipole or quadrupole settings; Partial coherence control, i.e. sigma or numerical aperture [NA]
    • G—PHYSICS
    • G03—PHOTOGRAPHY; CINEMATOGRAPHY; ANALOGOUS TECHNIQUES USING WAVES OTHER THAN OPTICAL WAVES; ELECTROGRAPHY; HOLOGRAPHY
    • G03F—PHOTOMECHANICAL PRODUCTION OF TEXTURED OR PATTERNED SURFACES, e.g. FOR PRINTING, FOR PROCESSING OF SEMICONDUCTOR DEVICES; MATERIALS THEREFOR; ORIGINALS THEREFOR; APPARATUS SPECIALLY ADAPTED THEREFOR
    • G03F7/00—Photomechanical, e.g. photolithographic, production of textured or patterned surfaces, e.g. printing surfaces; Materials therefor, e.g. comprising photoresists; Apparatus specially adapted therefor
    • G03F7/70—Microphotolithographic exposure; Apparatus therefor
    • G03F7/70058—Mask illumination systems
    • G03F7/70125—Use of illumination settings tailored to particular mask patterns
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Exposure And Positioning Against Photoresist Photosensitive Materials (AREA)

Abstract

The application discloses a model training method and device for determining a photoetching light source, wherein a convolutional neural network model is constructed, a specific size layout and a light source corresponding to the specific size layout are used as a training set of the convolutional neural network model, a mapping relation between the specific size layout and the light source corresponding to the specific size layout is learned, model parameters of the convolutional neural network model are determined according to the mapping relation, and therefore training of the convolutional neural network model is completed. Therefore, the accuracy of determining the photoetching light source model can be improved by using the specific size layout and the corresponding light source as the training set of the convolutional neural network model. The convolutional neural network method obtained by the model training method can determine the lithography light source of the layout to be subjected to lithography, does not need to calculate the lithography process window of each representative graph in the layout by adopting simulation software every time, and improves the efficiency of determining the lithography light source.

Description

Training method and device for determining lithography light source model
The application provides a divisional application for Chinese patent application with application number of 201811259558.6 and application date of 2018, 10 and 26, and the name of the application is 'method and device for determining lithography light source and model training method and device'.
Technical Field
The present disclosure relates to the field of semiconductor lithography, and in particular, to a training method and apparatus for determining a lithography light source model.
Background
In a semiconductor lithography process, in order to ensure a large lithography process window, a suitable light source needs to be selected according to the layout.
Currently, the existing method for determining the optimal light source of the lithography process generally adopts simulation software to calculate a process window of a representative graph in the layout, and then determines the lithography light source of the layout according to the calculation result of the process window.
However, the number of patterns included in the whole layout is huge, the calculation amount of the process window calculation of the representative patterns in the layout by the simulation software is huge, and the calculation process needs to consume too much calculation time, so that the efficiency of determining the lithography light source is low.
Disclosure of Invention
In view of this, the present application provides a training method and apparatus for determining a model of a lithography light source, so as to improve the efficiency of determining the lithography light source.
In order to solve the technical problems, the application adopts the following technical scheme:
a training method for determining a model of a lithographic light source, comprising:
constructing a convolutional neural network model;
acquiring a training set of the convolutional neural network model, wherein the training set comprises a specific size layout and a corresponding light source thereof;
learning the mapping relation between the layout with the specific size and the corresponding light source;
and determining model parameters of the convolutional neural network model according to the mapping relation.
Optionally, the acquiring the training set of the convolutional neural network model includes:
performing simulation calculation on the specific size layout to obtain a photoetching process window corresponding to the specific size layout;
determining an initial corresponding light source of the layout with the specific size according to the photoetching process window;
classifying the initial corresponding light sources of the specific size layout according to a preset discretization light source to obtain the final corresponding light sources of the specific size layout;
and the specific size layout and the final corresponding light source of the specific size layout are training sets of the convolutional neural network model.
Optionally, the classifying the initial corresponding light source of the specific size layout according to the preset discretized light source to obtain the final corresponding light source of the specific size layout includes:
judging whether the light source size parameter of the initial corresponding light source of the specific size layout is within the light source size parameter range corresponding to the preset discretized light source, and if so, taking the preset discretized light source as the final corresponding light source of the specific size layout.
Optionally, the method further comprises:
performing parameter discretization on a photoetching light source of a preset type to obtain a plurality of preset discretized light sources.
Optionally, when the lithography light source of the pre-known type is a ring light source, the performing parameter discretization on the lithography light source of the pre-known type to obtain a plurality of preset discretized light sources specifically includes:
and discretizing the outer corner radius and the inner corner radius of the annular light source respectively, and combining the discretized outer corner radius and inner corner radius into a plurality of preset discretized light sources.
Optionally, the outer corner radius and the inner corner radius of the annular light source are discretized respectively, and the discretized outer corner radius and inner corner radius are combined into a plurality of preset discretized light sources, which specifically includes:
setting the range of the external angle radius sigma_out of the annular light source as [ a, b ]]If the step size is the first step size s1, the outside corner radius σ_out is discretely divided into N outside corner radii, denoted as [ k ], according to the following equation (1) 1 ，k 2 …k i …k N-1 ，k N ，1≤i≤N]；
Setting the discretized outer angle radius k i The corresponding inner angle radius sigma_in is in the range of x i ,y i ]The step size is a second step size s2, and the inside corner radius sigma_in is discretely divided into the following formula (2) ni The inner corner radii;
the discretized outside corner radius and inside corner radius are combined into
Figure BDA0002690204760000031
The preset discretized light sources are arranged;
wherein, formula (1) is:
Figure BDA0002690204760000032
the formula (2) is as follows:
Figure BDA0002690204760000033
Figure BDA0002690204760000034
representing an upward rounding.
Optionally, after determining the model parameters of the convolutional neural network model according to the mapping relationship, the method further includes:
processing layout blocks of the to-be-photoetched layout according to the convolutional neural network model to respectively obtain corresponding light sources of each layout block; the pattern block is obtained by dividing the pattern to be photoetched according to a specific size;
and determining the photoetching light source of the layout to be subjected to photoetching according to the corresponding light source of each version of the block.
Optionally, determining the lithography light source of the layout to be lithographically according to the corresponding light source of each version block specifically includes:
and determining the same light sources in the corresponding light sources of each plate block, determining the number of the light sources in various same light sources, and taking the same light source with the largest number of light sources as the photoetching light source of the layout to be photoetched.
A training apparatus for determining a model of a lithographic light source, comprising:
the construction unit is used for constructing a convolutional neural network model;
the second acquisition unit is used for acquiring a training set of the convolutional neural network model, wherein the training set comprises a specific size layout and a corresponding light source thereof;
the learning unit is used for learning the mapping relation between the layout with the specific size and the corresponding light source;
and the second determining unit is used for determining model parameters of the convolutional neural network model according to the mapping relation.
Compared with the prior art, the application has the following beneficial effects: according to the model training method for determining the lithography light source, a convolutional neural network model is constructed, a specific size layout and the light source corresponding to the specific size layout are used as a training set of the convolutional neural network model, the mapping relation between the specific size layout and the light source corresponding to the specific size layout is learned, and model parameters of the convolutional neural network model are determined according to the mapping relation, so that training of the convolutional neural network model is completed. Therefore, the accuracy of determining the photoetching light source model can be improved by using the specific size layout and the corresponding light source as the training set of the convolutional neural network model. The convolutional neural network method obtained by the model training method can determine the lithography light source of the layout to be subjected to lithography, does not need to calculate the lithography process window of each representative graph in the layout by adopting simulation software every time, and improves the efficiency of determining the lithography light source.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIGS. 1 a-1 c are schematic views of different illumination modes employed in the art;
FIG. 2 is a flowchart of a method for determining a lithographic light source according to an embodiment of the present application;
FIG. 3 is a flowchart of a training method for determining a model of a lithographic light source according to an embodiment of the present application;
FIG. 4 is a schematic diagram of a layout provided in an embodiment of the present application;
FIG. 5 is a flowchart of another method for determining a lithographic light source according to an embodiment of the present application;
FIG. 6 is a schematic structural diagram of an apparatus for determining a lithographic light source according to an embodiment of the present disclosure;
fig. 7 is a schematic structural diagram of a model training device according to an embodiment of the present application.
Detailed Description
In the photoetching process, the imaging resolution of the photoetching machine is related to the photoetching light source, so that when the photoetching process is carried out on different layouts, the illumination condition of the photoetching light source needs to be adjusted according to the patterns of the layouts, so that the layouts have the largest process window (common process window) when exposed.
The process window is a range of exposure dose and focus depth that can achieve the best lithography effect, and the larger the process window is, the better the process window is.
It is additionally noted that the lithographic light sources are generally symmetrically arranged, e.g. disk-shaped light sources, horizontal dipole light sources, vertical dipole light sources and ring-shaped light sources, which are located on the main optical axis.
Referring to fig. 1a to 1c, there are schematic illustrations of different illumination sources employed in the art.
Fig. 1a is a schematic view of illumination of a disc-shaped light source on a main optical axis, wherein the first area 11 is an illumination area of the disc-shaped light source;
fig. 1b is an illumination schematic diagram of a vertical bipolar light source, wherein the second region 121 and the third region 122 are an illumination region corresponding to a first pole and an illumination region corresponding to a second pole of the vertical bipolar light source, respectively;
since the illumination of the horizontal bipolar light source is the same as that of the vertical bipolar light source, only the light source is positioned differently, and thus, for brevity, description thereof will not be repeated here.
Fig. 1c is a schematic view of illumination of an annular light source, wherein the fourth region 13 is an illumination region of the annular light source, which region is annular.
Referring to fig. 2, a flowchart of a method for determining a lithography light source in a lithography process is provided in an embodiment of the present application. The method specifically comprises the following steps:
s201: and obtaining a representative part from the layout to be subjected to photoetching.
The representative part refers to a pattern part such as repeated patterns which are largely appeared in the layout, patterns which are easy to cause problems in lithography, key function patterns in a circuit and the like.
S202: a preliminary lithography model is built in simulation software, a plurality of different conventional light sources are utilized to illuminate the representative portion, and a process window of the representative portion under illumination of the different conventional light sources is obtained.
As an example, a variety of conventional light sources may include: conventional round light sources, dipole light sources, quadrupole light sources, on-axis light sources and annular light sources.
In particular, the types of dipole light sources are numerous, for example, both horizontal and vertical dipole light sources are one type of dipole light source.
There are many kinds of on-axis light sources, for example, a disc-shaped light source located on the main optical axis is an on-axis light source.
S203: the optimal process window is determined based on the process window under illumination from a plurality of different conventional light sources.
The optimal process window is the best illumination condition for overlapping the photolithographic process windows of all patterns.
S204, determining the optimal photoetching light source for the layout to be photoetched according to the optimal process window.
It should be noted that the lithography light source determined in S204 may include a light source type and a light source attribute parameter (such as, a diameter, an inner and outer radius, etc.).
Also, it is worth explaining that since the graphics in each layout may include X-direction lines, Y-direction lines, and other-direction lines. Thus, the prior art generally uses a ring light source for illumination. In particular, the patterns on the contact hole layer layout or the through hole layer layout are square and rectangle with specific side length, and the directions of lines are random, and then an annular light source is generally selected for illumination.
Because the method for determining the optimal light source of the photoetching process needs to calculate the process windows of all the representative graphics in the layout by adopting simulation software, and the number of the graphics contained in the whole layout is huge, the calculation amount of the simulation software for calculating the process windows of the representative graphics in the layout is huge, and the calculation process needs to consume too much calculation time, so that the efficiency for determining the photoetching light source is lower.
To solve this problem, embodiments of the present application further provide a method of determining a lithographic light source, the method comprising: obtaining a layout to be subjected to photoetching; dividing the layout to be subjected to photoetching according to a specific size to form M edition blocks; the specific size is a layout size used in convolutional neural network model training; processing m layout blocks according to the trained convolutional neural network model to respectively obtain corresponding light sources of each layout block; and determining the photoetching light source of the layout to be subjected to photoetching according to the corresponding light source of each version of the block. Wherein M is more than or equal to 2 and less than or equal to M, and M and M are integers.
In the method for determining the photoetching light source provided by the embodiment of the application, firstly, a layout to be subjected to photoetching is segmented according to a specific size to form a plurality of layout blocks, and then each layout block is respectively processed by using a trained convolutional neural network model to obtain the light source respectively corresponding to each layout block; and finally, determining the photoetching light source of the layout to be subjected to photoetching according to the light sources respectively corresponding to the plate blocks. Therefore, in the method provided by the application, the photoetching process window of each representative graph in the layout is not required to be calculated by adopting simulation software, and the photoetching light source of the layout to be photoetched can be determined by utilizing the convolutional neural network model. Therefore, the method solves the problem that the efficiency of determining the lithography light source is low because the simulation software consumes too much calculation time for calculating the lithography process windows of each representative graph, and improves the efficiency of determining the lithography light source.
In the embodiment of the application, the type of the lithography light source (such as an annular light source) and the size parameter of the light source (such as the size of the inner radius and the outer radius of the annular light source) of the layout to be subjected to lithography can be obtained simultaneously by using the trained convolutional neural network model. Thus, the light source determined by the method includes not only the light source type but also the light source size parameter.
In order to make the present application solution better understood by those skilled in the art, the following description will clearly and completely describe the technical solution in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
For ease of explanation and understanding of the method for determining a lithographic light source provided herein, a training method for determining a model of a lithographic light source will be described first with reference to the accompanying drawings.
Referring to fig. 3, a flowchart of a training method for determining a model of a lithographic light source according to an embodiment of the present application is shown.
The training method for determining the lithography light source model provided by the embodiment of the application comprises the following steps:
s301: and constructing a convolutional neural network model.
The convolutional neural network (Convolutional Neural Networks, CNN) may comprise multiple layers. For example, a 9-layer convolutional neural network includes: an input layer, a first convolution layer, a first activation function layer, a normalization layer, a second convolution layer, a second activation function layer, a full connection layer, a de-maximum layer, and a classification layer.
S302: and obtaining a training set of the convolutional neural network model, wherein the training set comprises a specific size layout and a corresponding light source thereof.
The specific size is a predetermined size, for example, a predetermined specific size of 2 μm by 2 μm.
The plurality of layouts within the training set may be of a wide variety, for example, in order to increase the accuracy of determining the model of the lithography light source, the training set of the model may comprise a variety of representative layouts. The representative layout can be a layout comprising patterns with higher repetition rate, a layout comprising patterns which are easy to make mistakes in the photoetching process, or a layout comprising patterns with key functions in a circuit.
Referring to fig. 4, a schematic layout diagram is provided in an embodiment of the present application.
In fig. 4, the first layout 401 includes 7 representative graphs 401a to 401h, and when the training set is obtained from the first layout 401, the first layout 401 may be cut based on the representative graphs 401a to 401h, and the cut layout including the representative graphs 401a to 401h may be added to the training set.
In order to improve the accuracy of determining the lithography light source model, a layout with a specific size may be selected from a plurality of layouts including a representative graph as a training set (of course, the training set also includes light sources corresponding to the layout with the specific size). For example, the first training set includes 300 specific size layouts and light sources corresponding to the specific size layouts, wherein 20 specific size layouts including the representative pattern are acquired from the first layout to the tenth layout, 40 specific size layouts including the representative pattern are acquired from the eleventh layout, and 60 layouts including the representative pattern are acquired from the twelfth layout.
S303: and learning the mapping relation between the layout with the specific size and the corresponding light source.
S304: and determining model parameters of the convolutional neural network model according to the mapping relation.
The above process is a process of determining a convolutional neural network model.
As an example, S302 may be specifically:
s302a: and performing simulation calculation on the specific size layout to obtain a photoetching process window corresponding to the specific size layout.
S302b: and determining an initial corresponding light source of the layout with the specific size according to the photoetching process window.
Specifically, the steps are firstly to determine the optimal process window in the corresponding photoetching process window by carrying out simulation calculation on the layout with the specific size, and then to determine the photoetching light source corresponding to the optimal process window, namely the initial corresponding light source of the layout with the specific size.
S302c: classifying the initial corresponding light sources of the specific size layout according to the preset discretized light sources to obtain the final corresponding light sources of the specific size layout.
The final corresponding light sources of the specific size layout and the specific size layout are training sets of the convolutional neural network model.
Specifically, the preset discretized light source is a discretized light source obtained by performing parameter discretization on a photoetching light source of a preset type in advance. The total number of preset discretized light sources is limited.
It should be noted that, by simulating the layout with a specific size, the determined initial corresponding light source is a light source with very refined size parameters, that is, if the data are continuous within the allowable range of the lithography machine, there may be an infinite number of initial corresponding light sources with such refined sizes. Considering that the initial corresponding light source cannot be effectively learned as a training set, the initial corresponding light source is classified according to a limited number of preset discretized light sources. For example, when the type of the light source is a ring light source, the initial corresponding light source has an outer diameter of 50.125 μm and an inner diameter of 36.889 μm, and the light source may be classified into a preset discretized light source having an outer diameter of 50.0 μm and an inner diameter of 37.0 μm; while the initial corresponding light source has an outer diameter of 49.5 μm and an inner diameter of 37.2 μm, it is also possible to categorize the light source into a preset discretized light source having an outer diameter of 50.0 μm and an inner diameter of 37.0 μm.
As an example, S302c may be specifically: comparing the light source size parameter of the initial corresponding light source of the specific size layout with the light source size parameter range corresponding to the preset discretized light source, and if the light source size parameter of the initial corresponding light source falls within the light source size parameter range corresponding to a certain preset discretized light source, attributing the initial corresponding light source to the preset discretized light source, wherein the preset discretized light source is the final corresponding light source. Illustrating: setting the light source size parameter of the initial light source corresponding to the specific size layout to be 11.5 μm, and setting the light source size of the light source 1 after the pre-discretization to be 11.7 μm, wherein the parameter range corresponding to the light source 1 is 11.2 μm to 12.2 μm (namely, more than or equal to 11.2 μm and less than 12.2 μm), and attributing the initial corresponding light source to the light source 1, namely, setting the light source 1 as the final corresponding light source.
Furthermore, in combination with the foregoing method flow, in order to obtain the discretized light source, the training method provided in the present application may further include:
performing parameter discretization on a photoetching light source of a preset type to obtain a plurality of preset discretized light sources.
It is noted that the lithography light source of the pre-known type may be of a type commonly used in the art, for example at least one of a ring light source, a conventional light source, a bi-polar light source, a quadrupole light source, an on-axis quadrupole light source, etc.
As an implementation manner, when the lithography light source of the pre-known type is an annular light source, the outer corner radius and the inner corner radius of the annular light source are respectively discretized, and the discretized outer corner radius and inner corner radius are combined into a plurality of preset discretized light sources. Moreover, the specific process can be as follows: and discretizing the outer corner radius and the inner corner radius of the annular light source respectively, and combining the discretized outer corner radius and inner corner radius into a plurality of light sources.
In order to improve the accuracy of determining the model of the lithography light source, the outer corner radius and the inner corner radius of the annular light source are respectively discretized, which can be specifically:
first, the outer angle radius sigma_out of the ring light source is set to be [ a, b]If the step size is the first step size s1, the outside corner radius σ_out is discretely divided into N outside corner radii, denoted as [ k ], according to the following equation (1) 1 ，k 2 …k i …k N-1 ，k N ，1≤i≤N]；
Wherein, formula (1) is:
Figure BDA0002690204760000091
in the formula (1),
Figure BDA0002690204760000092
representing an upward rounding.
The method for rounding upwards specifically comprises the following steps: the number after the decimal point is firstly truncated, and then the number before the decimal point is added by one.
For example, if n=21.1 and N is rounded up, N is eventually 22; if n=21 and N is rounded up, N is eventually 21; if n=21.8 and N is rounded up, N is eventually 22.
Then, the i-th outside corner radius k after discretization is set i The corresponding inner angle radius sigma_in is in the range of x i ,y i ]The step size is a second step size s2, and the inside corner radius sigma_in is discretely divided into the following formula (2) ni The inner corner radii;
the formula (2) is as follows:
Figure BDA0002690204760000101
in the formula (2),
Figure BDA0002690204760000102
representing an upward rounding.
Finally, the discretized outside corner radius and inside corner radius are combined into
Figure BDA0002690204760000103
And each preset discretized light source.
In the specific implementation manner of the method for determining the training of the lithography light source model provided in the embodiment of the application, a plurality of specific size layouts and pre-discretized light sources to which corresponding light sources belong are used as training sets of the convolutional neural network model, mapping relations between the specific size layouts and the pre-discretized light sources to which the light sources corresponding to the specific size layouts belong are learned, and finally model parameters of the convolutional neural network model are determined according to the mapping relations, so that training of the convolutional neural network model is completed.
According to the trained convolutional neural network model, the photoetching light source required by the layout to be subjected to photoetching for obtaining the layout by using the convolutional neural network model can be conveniently determined.
Based on the convolutional neural network model obtained by the training method for determining the lithography light source model provided by the embodiment, the embodiment of the application provides a method for determining the lithography light source, and the method for determining the lithography light source will be described below with reference to the accompanying drawings.
Referring to fig. 5, a flowchart of a method for determining a lithographic light source according to an embodiment of the present application is provided.
The method for determining the photoetching light source provided by the embodiment of the application comprises the following steps:
s501: and obtaining the layout to be subjected to photoetching.
S502: dividing the layout to be subjected to photoetching according to a specific size to form M edition blocks; the specific size is a layout size used in convolutional neural network model training; m is more than or equal to 2, and M is an integer.
The specific size refers to the size of the layout in the training set used to train the convolutional neural network model.
For example, when the size of the layout in the training set for training the convolutional neural network model is 2 μm by 2 μm, the specific size is 2 μm by 2 μm.
The M pattern blocks may be obtained from the whole of the layout to be lithographically or from a part of the layout to be lithographically, and these two obtaining modes will be described in sequence.
The first acquisition mode is as follows: the M version blocks are obtained from the whole of the layout to be lithographically.
The number M of version blocks is an integer and is obtained by downward rounding, and the method for obtaining M may specifically be:
firstly, dividing the area of a layout to be photoetched by the area of a plate block with a specific size to obtain a number initial value M1.
For example, if the area of the layout to be lithographically is S and the specific dimension is k μm, the area of the pattern block of the specific dimension is k 2 The number initial value m1=s/k 2 。
And then, rounding down the number initial value M1 to obtain the number M of the layout, wherein the number M is rounded down to a number after the decimal point of M1 is directly truncated.
For example, if m1=21.1, then m=21; if m1=21, then m=21; if m1=21.8, then m=21.
As an example, when the size of the layout in the training set of the training convolutional neural network model is 2 μm by 2 μm and the area of the layout to be lithographically etched is 101 μm 2 When the step S502 may be specifically: firstly, calculating the area of a plate block with a specific size to be 4 mu m 2 The method comprises the steps of carrying out a first treatment on the surface of the Then the area of the layout to be photoetched is as follows101μm 2 Divided by 4 μm 2 M1=25.25 is obtained; then, rounding down M1, the number M of layouts is obtained as 25.
The first acquisition mode has the advantages that: because the M layout blocks selected in the first mode are from each part of the layout to be subjected to photoetching, the light source types corresponding to the M layout blocks respectively can comprehensively represent the light source requirements of different parts of the layout to be subjected to photoetching, so that the comprehensiveness of the M layout blocks is improved.
The second acquisition mode is as follows: m pattern blocks are obtained from the part of the pattern to be photoetched.
Two methods are available for obtaining pattern blocks including representative patterns in a pattern to be etched, one is to cut a portion of the representative patterns in the pattern to be etched into pattern blocks of a specific size; the other is to cut all the layout to be photoetched, and then select the layout blocks including the representative graph from all the layout blocks.
The second acquisition mode has the advantages that: because the number of representative patterns in the layout to be photoetched is smaller, the number of M version blocks acquired according to the second acquisition mode is also smaller, so that the efficiency of determining the photoetching light source is improved.
Different acquisition modes have different advantages, so that in an actual scene, an appropriate acquisition mode can be selected according to actual needs.
S503: processing m layout blocks according to the trained convolutional neural network model to respectively obtain corresponding light sources of each layout block; m is more than or equal to 2 and less than or equal to M, and M is an integer.
All the M layout pieces may be processed, or a part of the M layout pieces selected from the M layout pieces may be processed more typically, which is not limited herein.
As an example, S503 may specifically be: and taking each layout block as the input of a trained convolutional neural network model, processing the layout block by utilizing the convolutional neural network model, and finally taking the light source corresponding to the layout block as the output of the convolutional neural network model so as to obtain the light source corresponding to the layout block.
S504: and determining the photoetching light source of the layout to be subjected to photoetching according to the corresponding light source of each version of the block.
In step S504, there are various ways of determining the lithography light source of the layout to be lithographically according to the corresponding light sources of the different versions of the tiles. In order to improve the effectiveness of a lithography light source of a layout to be lithographically etched, the embodiment of the application provides a determination method, which may specifically be:
and determining the same light sources in the corresponding light sources of each plate block, determining the number of various same light sources, and taking the same light source with the largest number of light sources as the photoetching light source of the layout to be photoetched.
As an example, when m=50, if 30 layout blocks each correspond to a first light source, 10 version blocks each correspond to a second light source, 6 version blocks each correspond to a third light source, and 4 version blocks each correspond to a fourth light source. Specifically, S504 may be performed as: firstly, respectively searching the corresponding light sources of all the plate blocks, namely, a first light source, a second light source, a third light source and a fourth light source, and then determining the number of plate blocks (also counted as the number of each light source) corresponding to the first light source to the fourth light source, wherein the number of the plate blocks corresponding to the first light source is 30, the number of the plate blocks corresponding to the second light source is 10, the number of the plate blocks corresponding to the third light source is 6 and the number of the plate blocks corresponding to the fourth light source is 4; and then, taking the first light source with the largest number of the layout blocks as a photoetching light source of the layout to be photoetched.
In the method for determining the photoetching light source provided by the embodiment of the application, firstly, a layout to be subjected to photoetching is segmented according to a specific size to form a plurality of plate blocks, and then each plate block is processed by using a trained convolutional neural network model to obtain a corresponding light source of each plate block; and finally, determining the lithography light source of the layout to be subjected to lithography according to the corresponding light source of each version of the block. Therefore, in the method provided by the application, the photoetching process window of each representative graph in the layout is not required to be calculated by adopting simulation software every time, and the photoetching light source of the layout to be photoetched can be determined by utilizing the convolutional neural network model. The method solves the problem that the efficiency of determining the lithography light source is low because the simulation software is required to calculate the lithography light source representing each pattern every time and too much calculation time is consumed.
Based on the training method for determining the lithography light source model and the method for determining the lithography light source, the embodiment of the application also provides a device for determining the lithography light source, and the device for determining the lithography light source is explained and described below with reference to the accompanying drawings.
Referring to fig. 6, a schematic structural diagram of an apparatus for determining a lithographic light source according to an embodiment of the present application is shown.
The device for determining a lithography light source provided by the embodiment of the application comprises:
the first obtaining unit 601 is configured to obtain a layout to be lithographically etched.
The segmentation unit 602 is configured to segment the layout to be lithographically according to a specific size to form M version blocks; the specific size is a layout size used in convolutional neural network model training;
and the processing unit 603 is configured to process the M layout blocks according to the trained convolutional neural network model, so as to obtain corresponding light sources of each layout block respectively.
The first determining unit 604 is configured to determine a lithography light source of the layout to be lithographically according to the corresponding light sources of the respective version blocks.
Wherein M is more than or equal to 2 and less than or equal to M, and M and M are integers.
In order to improve the efficiency of determining the lithography light source, the first determining unit 604 is specifically configured to determine the same light sources in the corresponding light sources of each pattern block, determine the number of light sources in the same light sources, and use the same light source with the largest number of light sources as the lithography light source of the layout to be lithographically etched.
In order to improve the efficiency of determining the lithographic light source, the apparatus for determining the lithographic light source may further comprise:
and the training unit is used for training the convolutional neural network model.
The training unit specifically comprises:
the construction subunit is used for constructing a convolutional neural network model;
the training set acquisition subunit is used for acquiring a training set of the convolutional neural network model, wherein the training set comprises a specific size layout and a corresponding light source thereof;
the learning subunit is used for learning the mapping relation between the layout with the specific size and the corresponding light source;
and the determining subunit is used for determining the model parameters of the convolutional neural network model according to the mapping relation.
To improve the efficiency of determining the lithography light source, the training set acquisition subunit, as an example, comprises:
and the simulation module is used for performing simulation calculation on the layout with the specific size to obtain a photoetching process window corresponding to the layout with the specific size.
The determining module is used for determining an initial corresponding light source of the layout with the specific size according to the photoetching process window;
the classifying module is used for classifying the initial corresponding light source of the specific size layout according to a preset discretization light source so as to obtain the final corresponding light source of the specific size layout;
and the specific size layout and the final corresponding light source of the specific size layout are training sets of the convolutional neural network model.
In order to improve the efficiency of determining the lithographic light source, the apparatus for determining the lithographic light source may further comprise:
and the discrete subunit is used for carrying out parameter discretization on the photoetching light source of a preset type in advance so as to obtain a plurality of preset discretized light sources.
When the lithographic light source of the pre-known type is a ring light source, the discrete subunits are in particular: the method is used for discretizing the outer corner radius and the inner corner radius of the annular light source respectively, and the discretized outer corner radius and inner corner radius are combined into a plurality of preset discretized light sources.
In order to improve the efficiency of determining the lithography light source, when the light source is preset to be a ring light source, the discrete subunit specifically includes:
a first setting module for setting the range of the external angle radius sigma_out of the annular light source to be [ a, b ]]If the step size is the first step size s1, the outside corner radius sigma_out is discretely divided into N outside corner radii, which is denoted as [ k ], according to the formula (1) 1 ，k 2 …k i …k N-1 ，k N ，1≤i≤N]。
A second setting module for setting the discretized outer angle radius k i The corresponding inner angle radius sigma_in is in the range of x i ,y i ]The step size is a second step size s2, and the inside corner radius sigma_in is discretized and divided into according to the formula (2) ni The inner corner radii;
the discretized outside corner radius and inside corner radius are combined into
Figure BDA0002690204760000141
And each preset discretized light source.
The device for determining the lithography light source provided by the embodiment of the application comprises a first acquisition unit 601, a segmentation unit 602, a processing unit 603 and a first determination unit 604, wherein the device can determine the lithography light source of the layout to be lithographically determined by using a convolutional neural network model without adopting simulation software to calculate lithography process windows of various representative graphs in the layout. Therefore, the device solves the problem that the efficiency of determining the lithography light source is low because the simulation software consumes too much calculation time for calculating the lithography process windows of each representative graph, and improves the efficiency of determining the lithography light source.
Based on the above training method for determining the lithography light source model, the embodiment of the application further provides a training device for determining the lithography light source model, and the training device for determining the lithography light source model will be explained and described below with reference to the accompanying drawings.
Referring to fig. 7, a schematic structural diagram of a training device for determining a model of a lithography light source according to an embodiment of the present application is shown.
The training device for determining a lithography light source model provided by the embodiment of the application is characterized by comprising:
a construction unit 701, configured to construct a convolutional neural network model;
a second obtaining unit 702, configured to obtain a training set of the convolutional neural network model, where the training set includes a layout with a specific size and a corresponding light source;
a learning unit 703, configured to learn a mapping relationship between the layout of the specific size and the corresponding light source;
and a second determining unit 704, configured to determine model parameters of the convolutional neural network model according to the mapping relationship.
As an example, the second acquisition unit 702 may specifically include:
the simulation subunit is used for performing simulation calculation on the layout with the specific size to obtain a lithography process window corresponding to the layout with the specific size;
a determining subunit, configured to determine an initial corresponding light source of the layout with a specific size according to the photolithography process window;
the classifying subunit is used for classifying the initial corresponding light source of the specific size layout according to a preset discretization light source so as to obtain the final corresponding light source of the specific size layout;
and the specific size layout and the final corresponding light source of the specific size layout are training sets of the convolutional neural network model.
The training device for determining a lithography light source model provided by the embodiment of the application comprises: the device comprises a model construction unit 701, a second acquisition unit 702 and a model training unit 703, wherein a plurality of specific size layouts and corresponding light sources thereof are used as training sets of a convolutional neural network model, the mapping relation between the specific size layouts and the corresponding light sources of the specific size layouts is learned, and finally model parameters of the convolutional neural network model are determined according to the mapping relation, so that training of the convolutional neural network model is completed.
The foregoing is a specific implementation of the present application.

Claims (7)

1. A training method for determining a model of a lithographic light source, comprising:
constructing a convolutional neural network model;
acquiring a training set of the convolutional neural network model, wherein the training set comprises a specific size layout and a corresponding light source thereof;
learning the mapping relation between the layout with the specific size and the corresponding light source;
and determining model parameters of the convolutional neural network model according to the mapping relation.
2. The method of claim 1, wherein the obtaining the training set of convolutional neural network models comprises:
performing simulation calculation on the specific size layout to obtain a photoetching process window corresponding to the specific size layout;
determining an initial corresponding light source of the layout with the specific size according to the photoetching process window;
classifying the initial corresponding light sources of the specific size layout according to a preset discretization light source to obtain the final corresponding light sources of the specific size layout;
and the specific size layout and the final corresponding light source of the specific size layout are training sets of the convolutional neural network model.
3. The method according to claim 2, wherein said classifying the initial corresponding light source of a particular size layout according to a preset discretized light source to obtain the final corresponding light source of the particular size layout comprises:
judging whether the light source size parameter of the initial corresponding light source of the specific size layout is within the light source size parameter range corresponding to the preset discretized light source, and if so, taking the preset discretized light source as the final corresponding light source of the specific size layout.
4. The method according to claim 2, wherein the method further comprises:
performing parameter discretization on a photoetching light source of a preset type to obtain a plurality of preset discretized light sources.
5. The method according to claim 4, wherein when the lithography light source of the pre-known type is a ring light source, the performing the parameter discretization on the lithography light source of the pre-known type to obtain a plurality of the pre-set discretized light sources, specifically comprises:
and discretizing the outer corner radius and the inner corner radius of the annular light source respectively, and combining the discretized outer corner radius and inner corner radius into a plurality of preset discretized light sources.
6. The method according to claim 5, wherein the outer corner radius and the inner corner radius of the annular light source are discretized respectively, and the discretized outer corner radius and inner corner radius are combined into a plurality of the preset discretized light sources, specifically comprising:
setting the range of the external angle radius sigma_out of the annular light source as [ a, b ]]If the step size is the first step size s1, the outside corner radius σ_out is discretely divided into N outside corner radii, denoted as [ k ], according to the following equation (1) 1 ，k 2 …k i …k N-1 ，k N ，1≤i≤N]；
Setting the discretized outer angle radius k i The corresponding inner angle radius sigma_in is in the range of x i ,y i ]The step size is a second step size s2, and the inside corner radius sigma_in is discretely divided into n according to the following formula (2) i The inner corner radii;
the discretized outside corner radius and inside corner radius are combined into
Figure FDA0004094755680000021
The preset discretized light sources are arranged;
wherein, formula (1) is:
Figure FDA0004094755680000022
the formula (2) is as follows:
Figure FDA0004094755680000023
/>
Figure FDA0004094755680000024
representing an upward rounding.
7. A training apparatus for determining a model of a lithographic light source, comprising:
the construction unit is used for constructing a convolutional neural network model;
the second acquisition unit is used for acquiring a training set of the convolutional neural network model, wherein the training set comprises a specific size layout and a corresponding light source thereof;
the learning unit is used for learning the mapping relation between the layout with the specific size and the corresponding light source;
and the second determining unit is used for determining model parameters of the convolutional neural network model according to the mapping relation.
CN202010988995.2A 2018-10-26 2018-10-26 Training method and device for determining lithography light source model Active CN111999992B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010988995.2A CN111999992B (en) 2018-10-26 2018-10-26 Training method and device for determining lithography light source model

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811259558.6A CN109143796B (en) 2018-10-26 2018-10-26 Method and device for determining photoetching light source and method and device for training model
CN202010988995.2A CN111999992B (en) 2018-10-26 2018-10-26 Training method and device for determining lithography light source model

Related Parent Applications (1)

Application Number Title Priority Date Filing Date
CN201811259558.6A Division CN109143796B (en) 2018-10-26 2018-10-26 Method and device for determining photoetching light source and method and device for training model

Publications (2)

Publication Number Publication Date
CN111999992A CN111999992A (en) 2020-11-27
CN111999992B true CN111999992B (en) 2023-05-02

Family

ID=64809853

Family Applications (2)

Application Number Title Priority Date Filing Date
CN202010988995.2A Active CN111999992B (en) 2018-10-26 2018-10-26 Training method and device for determining lithography light source model
CN201811259558.6A Active CN109143796B (en) 2018-10-26 2018-10-26 Method and device for determining photoetching light source and method and device for training model

Family Applications After (1)

Application Number Title Priority Date Filing Date
CN201811259558.6A Active CN109143796B (en) 2018-10-26 2018-10-26 Method and device for determining photoetching light source and method and device for training model

Country Status (1)

Country Link
CN (2) CN111999992B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110187609B (en) * 2019-06-05 2020-08-21 北京理工大学 Deep learning method for calculating photoetching
CN114326313B (en) * 2020-09-29 2024-01-23 长鑫存储技术有限公司 Method for simultaneously monitoring multiple lighting conditions
CN112506008A (en) * 2020-11-20 2021-03-16 深圳晶源信息技术有限公司 Integrated circuit manufacturing light source optimization method and electronic equipment
CN116306452B (en) * 2023-05-17 2023-08-08 华芯程(杭州)科技有限公司 Photoresist parameter acquisition method and device and electronic equipment

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6115108A (en) * 1998-12-04 2000-09-05 Advanced Micro Devices, Inc. Illumination modification scheme synthesis using lens characterization data
CN105425532A (en) * 2015-11-20 2016-03-23 中国科学院微电子研究所 Light source mask collaborative optimization method
CN106773541A (en) * 2016-12-20 2017-05-31 中国科学院微电子研究所 A kind of photoetching solution Forecasting Methodology based on the matching of domain geometric properties
WO2017215767A1 (en) * 2016-06-17 2017-12-21 Huawei Technologies Co., Ltd. Exposure-related intensity transformation
CN107766893A (en) * 2017-11-03 2018-03-06 电子科技大学 Target identification method based on label multilevel coding neutral net
CN108535952A (en) * 2018-04-09 2018-09-14 北京理工大学 A kind of calculating photolithography method based on model-driven convolutional neural networks

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH04319935A (en) * 1991-04-18 1992-11-10 Fuji Photo Film Co Ltd Photograph printing exposure quantity deciding device
JP2006301779A (en) * 2005-04-18 2006-11-02 Konica Minolta Photo Imaging Inc Image processing system, image processing method, and image processing program
KR20080018203A (en) * 2005-06-24 2008-02-27 코닌클리케 필립스 일렉트로닉스 엔.브이. Methods and devices for characterizing polarization of illumination system
CN102789134B (en) * 2011-05-18 2014-10-29 上海微电子装备有限公司 Method for optimizing performances of exposure system
CN107797391B (en) * 2017-11-03 2020-04-24 上海集成电路研发中心有限公司 Optical proximity correction method
CN108665060B (en) * 2018-06-12 2022-04-01 上海集成电路研发中心有限公司 Integrated neural network for computational lithography

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6115108A (en) * 1998-12-04 2000-09-05 Advanced Micro Devices, Inc. Illumination modification scheme synthesis using lens characterization data
CN105425532A (en) * 2015-11-20 2016-03-23 中国科学院微电子研究所 Light source mask collaborative optimization method
WO2017215767A1 (en) * 2016-06-17 2017-12-21 Huawei Technologies Co., Ltd. Exposure-related intensity transformation
CN106773541A (en) * 2016-12-20 2017-05-31 中国科学院微电子研究所 A kind of photoetching solution Forecasting Methodology based on the matching of domain geometric properties
CN107766893A (en) * 2017-11-03 2018-03-06 电子科技大学 Target identification method based on label multilevel coding neutral net
CN108535952A (en) * 2018-04-09 2018-09-14 北京理工大学 A kind of calculating photolithography method based on model-driven convolutional neural networks

Also Published As

Publication number Publication date
CN109143796A (en) 2019-01-04
CN111999992A (en) 2020-11-27
CN109143796B (en) 2021-02-12

Similar Documents

Publication Publication Date Title
CN111999992B (en) Training method and device for determining lithography light source model
US20230394203A1 (en) Generating integrated circuit floorplans using neural networks
US11651259B2 (en) Neural architecture search for convolutional neural networks
TW201918791A (en) Integrated circuit fabricating method
US20140256144A1 (en) Semiconductor fin formation method and mask set
CN1869818A (en) Method, program product and apparatus for improving calibration of resist models used in critical dimension calculation
CN101930484A (en) Method and system of placing printing assist feature for random mask layout
CN108153995B (en) Test pattern selection method and device and method and device for building photoetching model
KR20210028716A (en) Deep learning model training method, device, electronic device and storage medium
CN106950795A (en) The forming method of secondary graphics
CN104991415A (en) Optical proximity effect correction method aiming at specific repeated patterns
US7711504B2 (en) Method and system for performing optical proximity correction with process variations considerations
CN115293075A (en) OPC modeling method, OPC modeling device and electronic equipment
CN103744265B (en) Improve the optical proximity correction method of process window
US9026954B2 (en) Surface topography enhanced pattern (STEP) matching
CN103901713B (en) Self-adaption optical proximity effect correction method adopting kernel regression technology
CN1495533A (en) Analogue method and system for designing aperture of exposure device and medium for recording analogue method
CN111369063B (en) Test paper model training method, test paper combining method and related device
CN104216235A (en) Figure pretreatment method and method for measuring figure density
US10001698B2 (en) Layout hierachical structure defined in polar coordinate
CN108021985A (en) A kind of model parameter training method and device
CN101738848B (en) Method for establishing OPC model based on variable light acid diffusion length
CN117518726A (en) Correction method of optical proximity effect
US20140237434A1 (en) Photolithography mask design simplification
JP2007200322A (en) Method and system for analyzing layout of semiconductor integrated circuit device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant