Deep learning model for determining mask design related to semiconductor manufacturing

Abstract

A method of determining a mask design is described. The method includes generating a continuous multi-modal representation of a probability distribution of the target design in at least a portion of the potential space. The potential space includes a distribution of feature variables that can be used to generate a mask design based on the target design. The method includes selecting a variable from a continuous multi-modal representation in a potential space. The variables include potential spatial representations of one or more features used to determine the mask design. The method includes determining a mask design based on a target design and a variable.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to U.S. application No. 63 / 390,359, filed on July 19, 2022, and the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present disclosure generally relates to determining photolithography mask designs associated with semiconductor manufacturing. Background Art

[0004] For example, lithographic projection equipment can be used to manufacture integrated circuits (ICs). A patterning device (e.g., a mask) may include or provide a pattern corresponding to a single layer (a "design layout") of the IC, and the pattern may be transferred to a target portion (e.g., including one or more dies) on a substrate (e.g., a silicon wafer) coated with a layer of radiation-sensitive material ("resist") by methods such as irradiating the target portion with the pattern on the patterning device. Typically, a single substrate contains multiple adjacent target portions, and the lithographic projection equipment transfers the pattern to these target portions successively, one target portion at a time.

[0005] Before the pattern is transferred from the pattern forming device to the substrate, the substrate can undergo various procedures such as primer, resist coating and soft baking. After exposure, the substrate can undergo other procedures ("post-exposure procedures"), such as post-exposure baking (PEB), development, hard baking and measurement / inspection of the transferred pattern. This series of procedures serves as the basis for manufacturing a single layer of a device (e.g., an IC). The substrate can then undergo various processes such as etching, ion implantation (doping), metallization, oxidation, chemical mechanical polishing, etc., all of which are intended to complete a single layer of the device. If multiple layers are required in the device, the entire procedure or a variation thereof is repeated for each layer. Ultimately, a device will be present in each target portion on the substrate. These devices are then separated from each other by techniques such as cutting or sawing so that the individual devices can be mounted on a carrier, connected to pins, etc.

[0006] The manufacture of devices (such as semiconductor devices) typically involves the use of a variety of manufacturing processes to process a substrate (e.g., a semiconductor wafer) to form various features and multiple layers of the device. These layers and features are typically manufactured and processed using, for example, deposition, lithography, etching, chemical mechanical polishing, and ion implantation. Multiple devices can be manufactured on multiple dies on a substrate and then separated into individual devices. The device manufacturing process can be considered a patterning process. The patterning process involves a patterning step, such as optical and / or nanoimprint lithography using a pattern forming device in a lithographic apparatus to transfer the pattern on the pattern forming device to the substrate, and typically but optionally involves one or more related pattern processing steps, such as resist development by a developing device, baking the substrate using a baking tool, etching using the pattern using an etching device, etc.

[0007] Photolithography is a core step in the manufacture of devices such as integrated circuits (ICs), where patterns formed on a substrate define the functional elements of the device, such as microprocessors, memory chips, etc. Similar photolithography techniques are also used to form flat-panel displays, microelectromechanical systems (MEMS), and other devices.

[0008] As semiconductor manufacturing processes continue to advance, the size of functional elements continues to shrink. At the same time, the number of functional elements (such as transistors) per device has been steadily increasing, a trend often referred to as "Moore's Law." At the current state of the art, device layers are manufactured using photolithography projection equipment that projects the design layout onto a substrate using illumination from an illumination source, creating individual functional elements with dimensions well below 100 nanometers.

[0009] Such processes with printed feature sizes smaller than the traditional resolution limit of the lithographic projection equipment are often referred to as low-k1 lithography, according to the resolution formula CD = k1 × λ / NA, where λ is the wavelength of the radiation used (currently 248nm or 193nm in most cases), NA is the numerical aperture of the projection optical system in the lithographic projection equipment, CD is the "critical dimension" - usually the minimum feature size printed - and k1 is the empirical resolution factor. In general, the smaller the k1, the more difficult it is to reproduce on the substrate a pattern similar in shape and size to that planned by the designer to achieve specific electrical functions and performance. In order to overcome these difficulties, complex fine-tuning steps need to be applied to the lithographic projection equipment, the design layout or the pattern forming device. These include, but are not limited to, for example, optimization of NA and optical coherence settings, customized illumination schemes, use of phase-shifted pattern forming devices, optical proximity correction (OPC, sometimes also called "optical and process correction") in the design layout, source mask optimization (SMO) or other methods generally defined as "resolution enhancement technology" (RET). Summary of the invention

[0010] When generating training data from a target design to train a prediction model to predict mask designs for a semiconductor manufacturing process, similar target design patterns may result in different predicted mask designs, thereby generating different training data (e.g., even though similar target design patterns may be almost identical). Inconsistent training data causes typical machine learning models to predict average mask design features, which leads to ambiguity in feature extraction and often results in variations in defect predictions for a given semiconductor manufacturing process.

[0011] The present disclosure describes a model that learns a continuous multimodal distribution of mask features that result in effective wafer imaging and selects (optimal) variables from the continuous multimodal distribution. The variables include a latent space representation of one or more features used to determine the mask design. The model determines the mask design based on the target design and the variables. This provides consistent training data, thereby reducing ambiguity in feature extraction and enhancing defect prediction for a given semiconductor manufacturing process, among other advantages.

[0012] According to one embodiment, a method for determining a mask design is provided. The method includes generating a continuous multimodal representation of a probability distribution of a target design in at least a portion of a latent space. The latent space includes a distribution of feature variables that can be used to generate a mask design based on the target design. The method includes selecting variables from the continuous multimodal representation in the latent space. The variables include a latent space representation of one or more features used to determine the mask design. The method includes determining the mask design based on the target design and the variables.

[0013] In some embodiments, selecting the variables includes selecting a mode from a multi-modal probability distribution and sampling the variables from the selected mode.

[0014] In some embodiments, the generating, selecting, and determining are performed by an encoder structure and a generating structure having a conditional mapping sub-model.

[0015] In some embodiments, the encoder structure and the generator structure form a U-net type deep learning model.

[0016] In some embodiments, a deep learning model with a conditional mapping submodel includes a first neural network block, a second neural network block, and a third neural network block, wherein the first neural network block is configured to generate a continuous multimodal representation of a probability distribution of a target design in a latent space, the second neural network block is configured to select variables during training, and the third neural network block is configured to determine a mask design based on the target design and the variables.

[0017] In some embodiments, the first neural network block, the second neural network block, and the third neural network block are trained jointly.

[0018] In some embodiments, the second neural network block is trained to generate a distribution of feature variables present in the input sub-resolution assist feature (SRAF) and / or optical proximity correction (OPC) data.

[0019] In some embodiments, during training, the selected variables are used as ground truth to train a third neural network block to generate a mask design based on the input target design and the pattern selection options given the selected variables.

[0020] In some embodiments, the variables include information content from the mask domain or propagation of that information from the second neural network block to the latent space. In some embodiments, the variables include information content from the OPC and / or SRAF domains or propagation of that information from the second neural network block to the latent space. The mask, OPC, and / or SRAF domains can be and / or include data, calculations, manufacturing operations, and / or other information associated with the mask, OPC, and / or SRAF.

[0021] In some embodiments, the method further includes training the first neural network block, the second neural network block, and the third neural network block by classifying the output mask design as fake or real using an adversarial training sub-model, such that after training, the adversarial sub-model cannot distinguish the output of the third neural network block from the true reference data.

[0022] In some embodiments, the method further includes applying additional regularization / loss costs during training of the first neural network block, the second neural network block, and the third neural network block.

[0023] In some embodiments, applying the regularization / loss cost includes applying a cost term that penalizes the number of jagged edges in the determined mask design, reweighting the cost term that penalizes the number of jagged edges, applying a cost term that prioritizes binary pixel values ​​in an image associated with the determined mask design, applying a fixed selection option to select the best mask design, and / or applying regularization on the difference between two versions of the mask design.

[0024] In some embodiments, the target design includes an intended wafer pattern and / or intermediate data associated with the intended wafer pattern, the intermediate data including continuous transmission mask (CTM) data, a CTM image, and / or an intermediate mask design.

[0025] In some embodiments, determining a mask design based on a target design and variables includes (1) mapping the target design, CTM data and / or CTM image and / or intermediate mask design to a mask design, and / or (2) mapping the target design to CTM data and / or CTM image.

[0026] In some embodiments, latent space modeling may be used to generate distributions of feature variables of mask designs via variational Bayesian inference techniques.

[0027] In some embodiments, the feature includes a shape or structure associated with a reticle design of a target and / or semiconductor device.

[0028] In some embodiments, the method further includes performing forward consistency sub-modeling configured to ensure that the determined mask design will create a desired semiconductor wafer structure corresponding to the target design.

[0029] In some embodiments, forward consistent sub-modeling is performed by a fixed physics model and / or a parametric model that approximates the physics of the semiconductor manufacturing process.

[0030] In some embodiments, determining the mask design includes determining sub-resolution assist feature (SRAF) and / or optical proximity correction (OPC) data of the mask design.

[0031] In some embodiments, SRAF data and OPC data are determined as separate contributions.

[0032] In some embodiments, the target design is a target substrate design for a semiconductor wafer.

[0033] In some embodiments, the determined mask design includes an image.

[0034] In some embodiments, the method further includes sampling the resulting conditional latent space by generating a plurality of selection options; and evaluating a process window key performance indicator of the resulting mask design to determine the most robust mask that the pre-trained model can produce.

[0035] According to another embodiment, a method for determining a semiconductor mask design is provided, the method using a model that learns a multimodal distribution of mask features and selects variables that result in effective semiconductor wafer imaging. The method includes using a first neural network block of the model to generate a continuous multimodal representation of a probability distribution of a wafer target design in at least a portion of a latent space. The latent space includes a distribution of feature variables that can be used to generate a mask design based on the target design. The method includes using a second neural network block of the model and selecting variables from the continuous multimodal representation in the latent space during model training. The variables include a latent space representation of one or more features for determining the mask design. The selection includes selecting a pattern from the multimodal representation of the probability distribution and sampling variables from the selected pattern. The method includes using a third neural network block of the model to determine the mask design based on the target design and the variables. For example, the model can be a U-net type deep learning model with a conditional mapping submodel.

[0036] According to another embodiment, a non-transitory computer-readable medium having instructions thereon is provided, which, when executed by a computer, causes the computer to perform any of the operations of the above method.

[0037] According to another embodiment, a system is provided, which includes one or more processors configured to perform any operation of the above method.

[0038] Other advantages of the embodiments of the present disclosure will become apparent from the following description taken in conjunction with the accompanying drawings which set forth certain exemplary embodiments by way of illustration and example. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings are incorporated in and constitute a part of this specification to illustrate one or more embodiments and together with the specification explain these embodiments. Embodiments will now be described by way of example only with reference to the accompanying drawings, in which corresponding reference characters indicate corresponding parts, wherein:

[0040] Figure 1 is a schematic diagram of a lithographic projection apparatus according to an embodiment.

[0041] Figure 2 A schematic diagram of a lithography cell according to one embodiment is shown.

[0042] Figure 3 A schematic diagram of overall lithography is shown according to one embodiment, representing the cooperation between three technologies to optimize semiconductor manufacturing.

[0043] Figure 4 An exemplary flow chart for simulating lithography according to one embodiment is shown.

[0044] Figure 5 An encoder-decoder architecture according to one embodiment is shown.

[0045] Figure 6 An encoder-decoder architecture within a neural network is shown according to one embodiment.

[0046] Figure 7 An overview of the operation of one embodiment of the present method for determining a mask design according to one embodiment is shown.

[0047] Figure 8 A generalized high-level representation of a model associated with some of the ideas described herein, including an encoder structure, a generative structure, and a conditional mapping sub-model, according to one embodiment is shown.

[0048] Fig. 9 A more specific representation of a model according to one embodiment is shown, including an encoder structure, a generation structure, and a conditional mapping sub-model.

[0049] Fig.10 Adversarial sub-models that may be included in a model and / or used to train a model are shown according to one embodiment.

[0050] Fig.11 A forward consistency sub-model that may be included in a model and / or used to train a model is shown according to one embodiment.

[0051] Fig.12 An embodiment of this model is shown, according to one embodiment, where optical proximity correction (OPC) and sub-resolution assist feature (SRAF) contributions are processed separately and then combined to generate a mask design.

[0052] Fig.13 The iterations involved in finding a joint solution for training optimization associated with a model according to one embodiment are schematically illustrated.

[0053] Fig.14 Equations for training a model according to one embodiment are shown.

[0054] Fig.15 An example of using a trained model to infer (or otherwise determine) a mask design (with OPC / SRAF features) is shown according to one embodiment.

[0055] Fig.16 Two different possible example options for reconfiguring a model for training a fixed potential selection to achieve a target performance level with respect to a predefined key performance indicator (KPI) of a semiconductor manufacturing process are shown according to one embodiment.

[0056] Fig.17 An example embodiment of configuring a model to account for lithography scanner focus disturbances according to one embodiment is shown.

[0057] Fig.18 is a diagram of an example computer system that may be used for one or more operations described herein, according to one embodiment. DETAILED DESCRIPTION

[0058] The design of photolithography masks involves finding a solution to an inverse problem: that is, given a set of target features on a substrate (e.g., a wafer), one needs to determine the equivalent mask features required to accurately expose the pattern on the substrate. Traditionally, this inverse lithography task has been solved as a series of optimization problems, namely finding the best mask design given the multiple requirements of the semiconductor manufacturing process (e.g., for the mask design itself, and for the process windows associated with forming the various features in the substrate).

[0059] Typically, current approaches to solving this inverse problem include a series of subtasks: a) a physics-based model is used to characterize the physical system (e.g., a scanner mask optical model and a resist model); the physics-based model is deployed as a forward model in the optimization task, aiming to partially solve the inverse problem from the target design to an intermediate continuous representation of the desired mask (e.g., a continuous transmission mask CTM); b) a deep learning model is constructed and trained to reproduce the results of the inverse problem, and after training, an appropriate CTM can be quickly evaluated; c) a series of discretization and post-processing operations are performed to convert the CTM obtained from the physics-based model or the deep learning model into an appropriate mask design (e.g., using optical proximity correction (OPC) and / or sub-resolution assist features (SRAF)) that meets the manufacturability criteria and the desired target design.

[0060] However, using current methods, the mapping from CTM to SRAF / OPC features is unstable. For example, for small perturbations in the CTM, the generated OPC / SRAF features in the mask design may differ significantly. Such differences are undesirable because they can introduce unnecessary differences and / or different features in the generated mask design, which makes final semiconductor manufacturing process control difficult. Additionally, this instability does not allow for the direct creation of machine learning models to automate and accelerate the mapping between the CTM and the target design because the models are trained using unstable data. Finally, current methods do not directly incorporate performance criteria for the resulting mask design. For example, when mapping the CTM to OPC / SRAF features, this criterion is inherent to the discrete steps taken.

[0061] For example, in order to train the model, ground truth images are required. Using current methods, generating ground truth images given similar input (target) images results in output (SRAF+OPC) images that are drastically different (e.g., due to the instabilities described above). This difference in output images (from prior / classical / naive prediction models - prior to the current models described in this paper) causes the model to learn an "average" of the output (SRAF+OPC) images, resulting in blurry and inappropriate mask designs that may not image correctly or cannot be manufactured.

[0062] Compared to the (multiple) models used in current methods, this paper describes a new deep learning model configured to solve the inverse mapping problem. The model can be built based on the concepts described below. The model is configured to learn the multimodal distribution of mask features, thereby producing a target design that can be manufactured on a substrate (such as a wafer). The model includes several sub-models that are trained as a single monolithic model, as described below.

[0063] The new deep learning model is configured to accept the variance in the ground truth output (SRAF+OPC) and explicitly learn the distribution of output (SRAF+OPC) images that are likely to come from similar inputs (target designs). This distribution (probability density function) can be modeled in a low-dimensional, real, and continuous latent space via methods such as variational Bayes. Given an input (target) image, samples from the latent space probability density function will each generate its own mask variable. Each of these mask variables will represent the ground truth image used to train the network and is no longer ambiguous. Since the latent space is variational, functions such as σ_ 先验 Parameters such as provide information about how the output (SRAF+OPC) image changes for this particular input (target). This information can also be used to guide the training of the model.

[0064] Embodiments of the present disclosure are described in detail with reference to the accompanying drawings, which are provided as illustrative examples of the present disclosure to enable those skilled in the art to practice the present disclosure. The following figures and examples are not intended to limit the scope of the present disclosure to a single embodiment, but other embodiments may be implemented by interchanging some or all of the described or shown elements. When certain elements of the present disclosure can be implemented in part or in whole using known components, only the parts of these known components necessary for understanding the present disclosure will be described, and detailed descriptions of other parts of these known components will be omitted to avoid confusing the present disclosure. The embodiments described as being implemented in software should not be limited thereto, but may include embodiments implemented in hardware, or a combination of software and hardware, or vice versa, which is obvious to those skilled in the art, unless otherwise specified herein. In this specification, embodiments showing a single component should not be considered restrictive; on the contrary, the present disclosure is intended to cover other embodiments including multiple identical components, or vice versa, unless otherwise explicitly stated herein. In addition, the applicant does not intend to give any term in the specification or claims an uncommon or special meaning unless explicitly specified. The present disclosure covers existing and future known equivalents of the known components mentioned herein by way of example.

[0065] Although IC manufacturing may be specifically mentioned herein, it should be clearly understood that the description herein has many other applications. For example, it can be used to manufacture integrated optical systems, guidance and detection patterns for magnetic domain memories, liquid crystal display (LCD) panels, thin film magnetic heads, etc. The skilled person will understand that in the context of such alternative applications, any use of the terms "reticle", "wafer" or "die" herein should be considered interchangeable with the more general terms "mask", "substrate" and "target portion", respectively.

[0066] In this document, the terms "radiation" and "beam" are used to cover all types of electromagnetic radiation, including UV radiation (e.g., having a wavelength of 365nm, 248nm, 193nm, 157nm or 126nm) and EUV (extreme ultraviolet radiation, e.g., having a wavelength range of about 5-100nm).

[0067] A pattern forming device (e.g., a semiconductor) may include or may form one or more design layouts. The design layout may be generated using a CAD (computer-aided design) program, a process commonly referred to as EDA (electronic design automation). Most CAD programs follow a set of predetermined design rules to create a functional design layout / pattern forming device. These rules are set by process and design constraints. For example, design rules define spatial tolerances between devices (such as gates, capacitors, etc.) or interconnects to ensure that the devices or lines do not interact in an undesirable manner. Design rules may include or specify specific parameters, parameter range limits, or other information. One or more design rule limits or parameters may be referred to as "critical dimensions" (CD). The critical dimension of a device may be defined as the minimum width of a line or hole or the minimum space or other feature between two lines or two holes. Therefore, CD determines the overall size and density of the designed device. One of the goals of device manufacturing is to faithfully reproduce the original design intent on the substrate (via a pattern forming device).

[0068] The terms "mask" or "patterning device" as used herein may be broadly interpreted as referring to a general semiconductor patterning device that can be used to impart an incident radiation beam with a patterned cross-section that corresponds to the pattern to be created in a target portion of the substrate. In addition to classical masks (transmissive or reflective; binary, phase-shifting, hybrid, etc.), other examples of such patterning devices include programmable mirror arrays and programmable LCD arrays.

[0069] The term "patterning process" as used herein refers to a process for creating an etched substrate by applying a specific light pattern as part of a photolithography process. However, a "patterning process" may also include (e.g., plasma) etching, as many of the features described herein may provide benefits for forming printed patterns using an etching (e.g., plasma) process.

[0070] As used herein, the term "pattern" refers to a desired pattern to be etched on a substrate (eg, a wafer).

[0071] As used herein, "printed pattern" (or pattern on substrate) refers to the physical pattern on the substrate that is etched based on the target pattern. The printed pattern may include grooves, channels, recesses, edges or other two-dimensional and three-dimensional features produced by photolithography processes, for example.

[0072] As used herein, the term "calibrate" means to modify (eg, improve or adjust) or validate something, such as a model.

[0073] A patterning system may be a system including any or all of the components described herein and other components configured to perform any or all operations associated with these components. A patterning system may include, for example, a lithographic projection device, a scanner, a system configured to apply or remove resist, an etching system, or other systems.

[0074] As an introduction, Figure 1 is a schematic diagram of a lithographic projection apparatus LA according to an embodiment. The LA may be used to produce a patterned substrate (eg, wafer) as described. For example, as part of a semiconductor manufacturing process, the patterned substrate may be inspected / measured by a SEM according to a FOV list.

[0075] The lithographic projection apparatus LA may include an illumination system IL, a first object table MT, a second object table WT, and a projection system PS. The illumination system IL may adjust a radiation beam B. In this example, the illumination system further includes a radiation source SO. The first object table (e.g., a patterning device table) MT may be provided with a patterning device holder to hold a patterning device MA (e.g., a mask), and connected to a first positioner to accurately position the patterning device relative to the article PS. The second object table (e.g., a substrate table) WT may be provided with a substrate holder to hold a substrate W (e.g., a resist-coated silicon wafer), and connected to a second positioner to accurately position the substrate relative to the article PS. The projection system (e.g., including a lens) PS (e.g., a refractive, reflective, or refractive-reflective optical system) may image an illuminated portion of the patterning device MA onto a target portion C (e.g., including one or more dies) of the substrate W. For example, the patterning device MA and the substrate W may be aligned using patterning device alignment marks M1, M2 and substrate alignment marks P1, P2.

[0076] The LA can be transmissive (i.e., have a transmissive patterning device). However, in general, it can also be reflective, for example (with a reflective patterning device). The device can employ different types of patterning devices for classical masks; examples include programmable mirror arrays or LCD matrices.

[0077] A source SO (e.g. a mercury lamp or an excimer laser, a LPP (laser produced plasma) EUV source) generates a radiation beam. The beam is fed to an illumination system (illuminator) IL, for example directly, or after passing through an adjustment device (such as a beam expander) or a beam delivery system BD (including guide mirrors, beam expanders, etc.). The illuminator IL may include an adjustment device AD ​​for setting the outer or inner radial extent of the intensity distribution (usually referred to as σ-outer and σ-inner, respectively). In addition, it will include various other components, such as an integrator IN and a converger CO. In this way, the beam B impinging on the pattern forming device MA has a desired uniformity and intensity distribution in its cross section.

[0078] In some embodiments, the source SO may be located within the housing of the lithographic projection apparatus (e.g., this is typically the case when the source SO is a mercury lamp), but it may also be remote from the lithographic projection apparatus. For example, the radiation beam it generates may be directed into the apparatus (e.g., with the aid of suitable guide mirrors). The latter case may occur, for example, when the source SO is an excimer laser (e.g., based on KrF, ArF or F2 laser emission).

[0079] The beam B may then intercept the patterning device MA held on the patterning device table MT. After passing the patterning device MA, the beam B may pass through a lens which focuses the beam B onto a target portion C of the substrate W. With the aid of the second positioning device (and the interferometric measurement device IF), the substrate table WT may be precisely moved, for example, to position a different target portion C in the path of the beam B. Similarly, the first positioning device may be used to precisely position the patterning device MA relative to the path of the beam B, for example, after mechanical retrieval of the patterning device MA from a patterning device library, or during scanning. Typically, movement of the tables MT, WT may be achieved with the aid of a long-stroke module (coarse positioning) and a short-stroke module (fine positioning). However, in the case of a stepper (as opposed to a step-and-scan tool), the patterning device table MT may be connected to a short-stroke actuator, or may be fixed.

[0080] The tool shown can be used in two different modes: a step mode and a scan mode. In the step mode, the patterning device table MT remains stationary and the entire patterning device image is projected onto a target portion C in one operation (i.e. a single "flash"). The substrate table WT can be moved in the x or y direction so that the beam B can illuminate different target portions C. In the scan mode, the same situation applies, except that a given target portion C is not exposed in a single "flash". Instead, the patterning device table MT can be moved in a given direction (e.g., the "scanning direction" or "y" direction) at a speed v so that the projection beam B scans the patterning device image. At the same time, the substrate table WT moves simultaneously in the same or opposite direction at a speed V=Mv, where M is the magnification of the lens (typically M=1 / 4 or 1 / 5). In this way, a larger target portion C can be exposed without compromising on resolution.

[0081] Figure 2 Schematic diagram of the lithography unit LC is shown. Figure 2 As shown, the lithography projection equipment (such as Figure 1 As shown, in Figure 2 The lithographic apparatus LA shown in FIG. 1 may form part of a lithographic cell LC, sometimes also referred to as a lithographic cell or (lithographic) cluster, which typically also includes a lithographic apparatus for lithographically processing a substrate W ( Figure 1 ) Equipment that performs pre-exposure and post-exposure processes. Traditionally, these equipment include a spin coater SC configured to deposit a resist layer, a developer for developing the exposed resist, a cooling plate CH and a baking plate BK, for example for adjusting the temperature of the substrate W, for example for adjusting the solvent in the resist layer. The substrate handler or robot RO picks up the substrates W from the input / output ports I / O1, I / O2, moves them between different process equipment, and transfers the substrates W to the loading station LB of the lithography equipment LA. The equipment in the lithography unit (also commonly referred to as tracks) is typically controlled by a track control unit TCU, and the track control unit TCU itself can be controlled by a monitoring system SCS, which can also control the lithography equipment LA (for example, via the lithography control unit LACU).

[0082] In order to expose the substrate W ( Figure 1 ) can be correctly and consistently exposed, the substrate needs to be inspected to measure characteristics of the patterned structure, such as feature edge placement, overlay error between subsequent layers, line thickness, critical dimensions (CD), etc. To this end, an inspection tool (not shown) may be included in the lithography cell LC. If an error is detected, adjustments may be made to the exposure of subsequent substrates or to other processing steps to be performed on the substrate W, particularly if the inspection is performed before other substrates W of the same batch or lot are still to be exposed or processed.

[0083] The inspection device (which may also be referred to as a metrology device) is used to determine properties of the substrate W, in particular how properties of different substrates W vary or how the relevant properties of different layers of the same substrate W vary from layer to layer. The inspection device may alternatively be configured to identify defects on the substrate W and may, for example, be part of a lithography cell LC, or may be integrated into a lithography apparatus LA, or may even be a stand-alone device. The inspection device may measure the properties using the actual substrate (e.g. an image of a charged particle—SEM—wafer pattern) or an image of the actual substrate, a latent image (an image in the resist layer after exposure), a semi-latent image (an image in the resist layer after a post-exposure bake step PEB), a developed resist image (wherein exposed or unexposed parts of the resist have been removed), an etched image (after a pattern transfer step such as etching), or other means.

[0084] Figure 3 FIG. 1 shows a schematic diagram of overall lithography, representing the collaboration between the three technologies to optimize semiconductor manufacturing. Typically, the patterning process in the lithography apparatus LA is one of the most critical steps in the process, which requires the substrate W( Figure 1 ) to dimension and place structures with high accuracy. To ensure this high accuracy, the three systems (in this example) can be combined in a so-called “holistic” control environment, such as Figure 3 As shown. One of these systems is a lithography apparatus LA, which is (virtually) connected to a metrology device (e.g., a metrology tool) MT (a second system) and a computer system CS (a third system). The "holistic" environment can be configured to optimize the cooperation between these three systems to enhance the overall process window and provide a tight control loop to ensure that the patterning performed by the lithography apparatus LA remains within the process window. The process window defines the range of process parameters (e.g., dose, focus, overlay) within which a particular manufacturing process can produce a defined result (e.g., a functional semiconductor device), and generally within which process parameter variations in the lithography process or patterning process are allowed.

[0085] The computer system CS may use the (portion of) the design layout to be patterned to predict which resolution enhancement technique should be used and perform computational lithography simulations and calculations to determine which mask layout and lithography equipment settings achieve the maximum overall process window for the patterning process (given by Figure 3 Typically, the resolution enhancement techniques are arranged to match the patterning possibilities of the lithographic apparatus LA. The computer system CS may also be used to detect where within the process window the lithographic apparatus LA is currently operating (e.g. using input from a metrology tool MT) to predict whether defects may exist (e.g. due to suboptimal processing) Figure 3 (indicated by the arrow pointing to “0” in the second scale SC2).

[0086] The metrology device (tool) MT may provide input to the computer system CS to enable accurate simulations and predictions, and may provide feedback to the lithographic apparatus LA to identify possible drift, e.g. in the calibration state of the lithographic apparatus LA (in Figure 3 (indicated by multiple arrows in the third scale SC3).

[0087] In the lithography process, it is necessary to frequently measure the created structures, for example for process control and verification. The tool for performing such measurements comprises a measurement tool (device) MT. Different types of measurement tools MT are known for performing such measurements, including scanning electron microscopes (SEM) or various forms of scatterometer measurement tools MT. In some embodiments, the measurement tool MT is or comprises a SEM.

[0088] In some embodiments, the measurement tool MT is or includes a spectroscopic scatterometer, an ellipsometer, or other light-based tool. The spectroscopic scatterometer can be configured so that the radiation emitted by the radiation source is directed onto a target feature of the substrate, and the reflected or scattered radiation from the target is directed to a spectrometer detector, which measures the spectrum of the specularly reflected radiation (i.e., a measurement of the intensity as a function of wavelength). From this data, the structure or profile of the target that produced the detected spectrum can be reconstructed, for example, by rigorous coupled wave analysis and nonlinear regression or by comparison with a library of simulated spectra. Ellipsometers allow the parameters of a lithography process to be determined by measuring the scattered radiation in each polarization state. Such a measurement tool (MT) emits polarized light (such as linear, circular or elliptical) by using, for example, an appropriate polarization filter in the illumination portion of the measurement device. A light source suitable for the measurement device can also provide polarized radiation.

[0089] It is often desirable to be able to determine computationally how a patterning process produces a desired pattern on a substrate. Thus, a simulation may be provided to simulate one or more portions of the process. For example, it may be desirable to be able to simulate a photolithography process that transfers a patterning device pattern to a resist layer of a substrate and the pattern produced in the resist layer after the resist is developed.

[0090] Figure 4An exemplary flow chart for simulating lithography in a lithographic projection apparatus is shown. An illumination model 431 represents the optical properties of illumination. A projection optics model 432 represents the optical properties of projection optics. A design layout model 435 represents the optical properties of a design layout (including changes in radiation intensity distribution and / or phase distribution caused by a given design layout), which is a representation of the arrangement of features on or formed by a pattern forming device. An aerial image 436 can be simulated using the illumination model 431, the projection optics model 432, and the design layout model 435. A resist image 438 can be simulated from the aerial image 436 using a resist model 437. For example, a mask image, such as a CTM mask and / or other masks (e.g., by a design layout model 435 and / or other models) can also be simulated. The simulation of lithography can, for example, predict the profile and / or CD in the resist image.

[0091] More specifically, the illumination model 431 can represent the optical properties of the illumination, including but not limited to the NA-sigma (σ) setting and any specific illumination shape (e.g., off-axis illumination, such as toroidal, quadrupole, dipole, etc.). The projection optics model 432 can represent the optical properties of the projection optics, including, for example, aberrations, distortions, refractive index, physical size or dimensions, etc. The design layout model 435 can also represent one or more physical properties of the physical pattern forming device. The optical properties associated with the lithographic projection apparatus (e.g., properties of the illumination, pattern forming device, and projection optics) determine the aerial image. Because the pattern forming device used in the lithographic projection apparatus can be changed, it is desirable to separate the optical properties of the pattern forming device from the optical properties of the rest of the lithographic projection apparatus, which includes at least the illumination and projection optics (hence the design layout model 435).

[0092] The resist model 437 may be used to calculate a resist image from the aerial image. The resist model is typically related to the properties of the resist layer (eg, the effects of chemical processes occurring during exposure, post-exposure baking, and / or development).

[0093] The model can be used to accurately predict edge placement, spatial image intensity slope, sub-resolution assist features (SRAF) and / or CD, etc., which can then be compared to the expected or target design. The expected design is defined as the pre-OPC design layout, which can be provided in a standardized digital file format (such as GDSII, OASIS or other file formats).

[0094] For example, simulation and modeling can be used to configure one or more features of the pattern forming device pattern (e.g., perform optical proximity correction), one or more features of the illumination (e.g., change one or more characteristics of the spatial / angular intensity distribution of the illumination, such as changing the shape), and / or one or more features of the projection optical device (e.g., numerical aperture, etc.). Such configurations may be referred to as mask optimization, source optimization, and projection optimization, respectively. Such optimizations may be performed separately or combined in different combinations. One such example is source mask optimization (SMO), which involves configuring one or more features of the pattern forming device pattern and one or more features of the illumination. Optimization techniques may focus on one or more clips. Optimization may use the machine learning models described herein to predict the values ​​of various parameters (including images, etc.).

[0095] In some embodiments, the optimization process of the system can use a cost function. The optimization process can include finding a set of system parameters (design variables, process variables, etc.) that minimize the cost function. The cost function can have any suitable form, depending on the goal of the optimization. For example, the cost function can be the weighted root mean square (RMS) of the deviation of certain characteristics (evaluation points) of the system relative to the expected values ​​(e.g., ideal values) of these characteristics. The cost function can also be the maximum value (i.e., the worst deviation) of these deviations. The term "evaluation point" should be widely interpreted as including any characteristic of the system or manufacturing method. Due to the practicality of the implementation of the system and / or method, the design and / or process variables of the system can be limited to a limited range and / or interdependent. In the case of a lithographic projection device, constraints are generally associated with physical properties and characteristics of the hardware (such as an adjustable range) and / or pattern forming device manufacturability design rules. The evaluation point can include a physical point on the resist image on the substrate, as well as non-physical characteristics, such as dose and focus.

[0096] In some embodiments, illumination model 431, projection optics model 432, design layout model 435, resist model 437, and / or other models related to and / or included in the integrated circuit manufacturing process may be empirical models for performing the operations of the methods described herein. The empirical models may predict outputs based on correlations between various inputs (e.g., one or more characteristics of a mask or wafer image, one or more characteristics of a design layout, one or more characteristics of a pattern forming device, one or more characteristics of an illumination used in a lithography process (such as wavelength, etc.), etc.).

[0097] As an example, the empirical model can be a machine learning model and / or any other parameterized model. In the above paragraphs, certain non-machine learning models are described that compute lithography physics models. Machine learning models differ in that they bypass all or part of the physical model (e.g., the optical model described above). In some embodiments, the machine learning model (for example) can be and / or include mathematical equations, algorithms, graphs, charts, networks (e.g., neural networks), and / or other tools and machine learning model components. For example, a machine learning model can be and / or include one or more neural networks having an input layer, an output layer, and one or more intermediate layers or hidden layers. In some embodiments, one or more neural networks can be and / or include a deep neural network (e.g., a neural network having one or more intermediate layers or hidden layers between the input layer and the output layer).

[0098] One or more neural networks can be based on a large number of neural units (or artificial neurons). One or more neural networks can loosely imitate the way biological brains work (for example, through a large number of biological neuron clusters connected by axons). Each neural unit of a neural network can be connected to many other neural units of the neural network. Such connections can be reinforcing or inhibiting in their influence on the activation state of the connected neural units. In some embodiments, each individual neural unit can have a summation function that combines the values ​​of all its inputs. In some embodiments, each connection (or neural unit itself) can have a threshold function so that the signal must exceed the threshold to propagate to other neural units. These neural network systems can be self-learning and training, rather than explicitly programmed, and can perform better in certain problem-solving areas compared to traditional computer programs. In some embodiments, one or more neural networks can include multiple layers (for example, the signal path traverses from the front layer to the back layer). In some embodiments, the neural network can use backpropagation techniques, in which forward stimulation is used to reset the weights of the "front end" neural units. In some embodiments, the stimulation and inhibition of one or more neural networks can flow more freely, and the connections interact in a more chaotic and complex way. In some embodiments, the intermediate layers of one or more neural networks include one or more convolutional layers, one or more recurrent layers, and / or other layers.

[0099] One or more neural networks can be trained (i.e., their parameters determined) using a set of training data (e.g., ground truth). The training data can include a set of training samples. Each sample can be a pair, including an input object (usually a vector, which can be called a feature vector) and a desired output value (also called a supervisory signal). The training algorithm analyzes the training data and adjusts the behavior of the neural network by adjusting the parameters of the neural network (e.g., the weights of one or more layers) based on the training data. For example, given a set of N training samples of the form {(x1,y1),(x2,y2),...,(x N ,y N )}, so that x i is the feature vector of the ith example, y i is its supervisory signal, and the training algorithm looks for a neural network g:X→Y, where X is the input space and Y is the output space. A feature vector is an n-dimensional vector that represents the numerical features of an object (e.g., a wafer design, a clip, etc. in the above example). The vector space associated with these vectors is often called a feature or latent space. After training, the neural network can be used to make predictions using new samples.

[0100] As described herein, embodiments of the present disclosure include (multiple) models that include one or more parameterized models (e.g., machine learning models, such as neural networks) and / or other models using an encoder-decoder architecture. In the middle (e.g., an intermediate layer) of the model (e.g., a neural network), the present model formulates a low-dimensional encoding (e.g., in a latent space) that encapsulates information in the input to the model. The (multiple) models herein exploit the low dimensionality and compactness of the latent space for parameter estimation and / or prediction.

[0101] Figure 5 A general encoder-decoder architecture 50 is shown by way of non-limiting example. The encoder-decoder architecture 50 has an encoding portion 52 (encoder) and a decoding portion 54 (decoder). Figure 5 In the example shown, the encoder-decoder architecture 50 may output a predicted image 56 and / or other outputs.

[0102] As another non-limiting example, Figure 6 An encoder-decoder architecture 50 is shown within a neural network 62. The encoder-decoder architecture 50 includes an encoding portion 52 and a decoding portion 54. Figure 6 In , x represents the encoder input (e.g., input image or other data), and x' represents the decoder output (e.g., predicted output image and / or other data). Figure 6 In , z represents the latent space 64 and / or the low-dimensional encoding (tensor / vector). In some embodiments, z is a latent variable or is related to a latent variable.

[0103] In some embodiments, the low-dimensional encoding z represents one or more features of the input. One or more encoded features of the input may be considered key or key features of the input. Encoded features may be considered key or key features of the input because they are more predictive and / or have other characteristics than other features of the desired output. The one or more encoded features (dimensions) represented in the low-dimensional encoding may be predetermined (e.g., determined by a programmer when creating the present modular autoencoder model), determined by a previous layer of the neural network, adjusted by a user via a user interface associated with the system described herein, and / or may be determined by other methods. In some embodiments, the number of encoded features (dimensions) represented by the low-dimensional encoding may be predetermined (e.g., determined by a programmer when creating the present modular autoencoder model), determined based on outputs from previous layers of the neural network, adjusted by a user via a user interface associated with the system described herein, and / or determined by other methods.

[0104] It should be noted that although machine learning models, neural networks and / or encoder-decoder architectures are mentioned in this specification, the machine learning models, neural networks and encoder-decoder architectures are merely examples, and the operations described herein can be applied to different parameterized models.

[0105] As described above, process information (e.g., images, measurements, process parameters, metrology indicators, etc.) can be used to guide various manufacturing operations. Utilizing the lower dimensionality of the latent space to predict and / or otherwise determine process information may be faster, more efficient, require less computing resources, and / or have other advantages over previous methods of determining process information.

[0106] Figure 7 An overview 700 of operations of one embodiment of the present method for determining a mask design is shown. Overview 700 is an overview of the training and / or inference operations described herein. In operation 702, a continuous multimodal representation of a probability distribution of a target design is generated in at least a portion of a latent space. In operation 704, feature variables are selected from the continuous multimodal representation in the latent space. In operation 706, a mask design is determined based on the target design and the variables and / or other information. Operation 708 may include one or more steps that are performed to enhance the mask design determination. These operations are briefly summarized in the paragraphs immediately following and then in the following Figures 8 to 17 Each operation is explained in depth in the discussion.

[0107] In some embodiments, one or more of the operations described in overview 700 may be performed simultaneously and / or sequentially. For example, during training, operations 704, 706, and 708 may be applied together (or partially applied, as some elements of 708 may be omitted). In some embodiments, one or more of these operations may be performed iteratively during training and / or inference. The following description is an example of a sequence of joint operational steps, such as one training iteration or one inference step. For example, for training, operations 702 and 708 are interrelated and performed iteratively. However, at inference time, the regularization from operation 708 (described below) may not be used, but the forward model (described below) may still be used.

[0108] In some embodiments, a non-transitory computer-readable medium stores instructions that, when executed by a computer, cause the computer to perform one or more of operations 702-708, or other operations. Operations 702-708 are intended to be illustrative. In some embodiments, these operations may be accomplished by one or more additional operations not described, or may be accomplished without one or more of the operations discussed. For example, in some embodiments, operation 708 may be eliminated. Additionally, Figure 7 The order of operations 702-708 shown in and described herein is not intended to be limiting. For example, some or all of operations 702-708 may be performed simultaneously.

[0109] The generation, selection, and determination (e.g., operations 702, 704, and 706) are performed by an electronic model that includes an encoder structure and a generation structure (e.g., a decoder) having a conditional mapping submodel. In some embodiments, as described herein, the model is a machine learning model. In some embodiments, the model includes an encoder-decoder architecture. In one embodiment, the encoder-decoder architecture includes a variational encoder-decoder architecture, and operation 702 and / or other operations include training a variational encoder-decoder architecture using a probabilistic latent space that generates realizations in an output space. The latent space includes low-dimensional encodings and / or other information (as described herein). If the latent space is formed by sampling from a distribution (such as a Gaussian) given distribution parameters (such as mu and sigma) calculated by the encoder, the latent space is probabilistic.

[0110] In some embodiments, the encoder structure and the generating structure form a U-net type deep learning model. The U-net type deep learning model with a conditional mapping submodel includes a first neural network block configured to generate a continuous multimodal representation of a probability distribution of a target design in a latent space, a second neural network block configured to select variables during training, and a third neural network block configured to determine a mask design based on the target design and the variables. In some embodiments, the latent space models the distribution of feature variables, which can be used to generate mask designs via, for example, variational Bayesian inference techniques.

[0111] As described above, in operation 702, a continuous multimodal representation of a probability distribution of a target design is generated in at least a portion of a latent space. In some embodiments, the target design is a target substrate design for a semiconductor wafer. In some embodiments, the target design includes an expected wafer pattern, a GDS file, a target layout, and / or intermediate data associated with the expected wafer pattern. In some embodiments, the target design may be associated with other data, including continuous transmission mask (CTM) data (the CTM includes a desired mask image), a CTM image, an intermediate mask design, and / or other data. The latent space includes a distribution of feature variables that can be used to generate a mask design based on the target design. For example, a mask feature (as opposed to the above-described coding feature) may include a shape or structure associated with a target and / or mask design for a semiconductor device.

[0112] In some embodiments, operation 702 includes jointly training a first neural network block, a second neural network block, and a third neural network block (e.g., before using the model for inference operations). In some embodiments, the second neural network block is trained to generate a distribution of feature variables present in the input sub-resolution assist feature (SRAF) and / or optical proximity correction (OPC) data. During training, the selected variables are used as ground truth values ​​to train the third neural network block to generate a mask design based on the input target design and pattern selection options given the selected variables.

[0113] At operation 704, variables are selected from the continuous multimodal representation in the latent space. The variables include a latent space representation for determining one or more features of the mask design. Selecting the variables includes selecting a pattern from the multimodal representation of the probability distribution and sampling the variables from the selected pattern. In some embodiments, the variables include information content from the mask domain or the propagation of that information from the second neural network block to the latent space. In some embodiments, the variables include information content from the OPC and / or SRAF domains or the propagation of that information from the second neural network block to the latent space. The mask, OPC, and / or SRAF domains may be and / or include data, calculations, manufacturing operations, and / or other information associated with the mask, OPC, and / or SRAF.

[0114] At operation 706, a mask design is determined based on the target design and the variables and / or other information. In some embodiments, the determined mask design includes an image. In some embodiments, determining the mask design based on the target design and the variables includes (1) mapping the target design, CTM data, and / or CTM image to the mask design, and / or (2) mapping the target design to the CTM data and / or CTM image. In some embodiments, determining the mask design includes determining sub-resolution assist feature (SRAF) and / or optical proximity correction (OPC) data for the mask design. In some embodiments, the SRAF data and the OPC data are determined as separate contributions.

[0115] Operation 708 may include one or more steps that are performed to enhance the mask design determination. In some embodiments, operation 708 includes training (or retraining) the first, second, and third neural network blocks by classifying the output mask design as fake or real using an adversarial training sub-model, such that after training, the adversarial sub-model cannot distinguish the output of the third neural network block from the real reference data. In some embodiments, operation 708 includes applying additional regularization / loss costs during training of the first, second, and third neural network blocks. Applying the regularization / loss cost includes applying a cost term that penalizes the amount of jagged edges in the determined mask design, re-weighting the cost term that penalizes the amount of jagged edges, applying a cost term that prioritizes binary pixel values ​​in the image associated with the determined mask design, applying a fixed selection option for selecting the best mask design, and / or applying regularization to the difference between two versions of the mask design.

[0116] In some embodiments, operation 708 includes performing forward consistency sub-modeling configured to ensure that the determined mask design will create a desired semiconductor wafer structure corresponding to the target design. In some embodiments, the forward consistency sub-modeling is performed by a fixed physical model and / or a parametric model that approximates the physics of the semiconductor manufacturing process. In some embodiments, operation 708 includes sampling the resulting conditional latent space by generating a plurality of selection options; and evaluating a process window key performance indicator of the resulting mask design to determine the most robust mask that the pre-trained model can produce.

[0117] By way of non-limiting example, Figure 8 A generalized high-level representation of a model 800 associated with some of the ideas described herein is shown, including an encoder structure, a generative structure, and a conditional mapping sub-model. Figure 8 A generalized representation of model 800 is shown, including an encoder structure 802, a generator structure 804, and a conditional mapping submodel 806. Model 800 generates a unimodal distribution. Model 800 assumes that both target data and mask data can be mapped to a common distribution, but this embodiment is not configured to create a multimodal distribution (the resulting distribution will be a common distribution of features associated with the target and mask, rather than a distribution of mask features). Figure 8 More details on the high-level concepts presented in Fig. 9 + Display and description.

[0118] The encoder structure 802 and the generator structure 804 form a U-net type deep learning model 810. A true and continuous variational low-dimensional latent space 812 is included as part of the model 800. During inference, the input image 814 (target design) is simultaneously projected to a CTM-like image 816 and encoded into a latent space that models (via variational Bayes) the distribution of mask variables that can be generated. Given an input image 814, samples from the latent space probability density function will each generate their own mask variable 820. Since the latent space 812 is variational, a function such as σ 先验 Parameters such as provide information about how the output (SRAF+OPC) image 830 changes for that particular input image 814.

[0119] The minimal training set for model 800 may include a target design and a corresponding mask design (e.g., more than one mask design per target design). However, if a CTM is also available, the CTM may be used as a secondary ground truth during training to enhance the physical interpretability of the U-net output. (Note that the generalized model 800 can also be applied to free-form SRAF+OPC.)

[0120] Fig. 9A more specific representation of model 800 is shown, including an encoder structure 802, a generation structure 804, and a conditional mapping submodel 806. The conditional mapping submodel 806 depends on existing features in a given OPC / SRAF data 920, and infers OPC / SRAF features from the input CTM or, alternatively, directly from the input target information. In this example, the encoding features are learned from the OPC / SRAF data. The data 920 and data associated with the mask design 904 (described below) are associated with the same sample to produce a consistent mapping (one is the true sample, the other is the output approximation).

[0121] The conditional mapping submodel 806 is used to generate a latent space 812 so that it models a discrete categorical distribution. Here, "s" is the probability distribution of a given feature / variable that is sampled to generate categorical samples with 1-hot encodings represented by "d". This can be understood and made tractable, for example, using the GumbelSoftmax method. Fig. 9 The conditional construction shown in helps train the model 800 so that it knows the choice to make for a given feature in the reference data. The choice can be encoded via a discrete categorical 1-hot encoding d, such as Fig. 9 As shown. Fig. 9 , d conditions or otherwise selects a version of the features to be inferred in the output image (e.g., mask design 904).

[0122] like Fig. 9 As shown, the encoder structure 802 and the generating structure 804 again form a U-net type deep learning model. The U-net type deep learning model with the conditional mapping submodel 806 includes a first neural network block (encoder structure 802), a second neural network block (conditional mapping submodel 806), and a third neural network block (generating structure 804), wherein the first neural network block is configured to generate a continuous multimodal representation of the probability distribution of the mask design in the latent space 812, the second neural network block is configured to select variables 902 during training, and the third neural network block is configured to determine the mask design 904 based on the target design 900 (e.g., represented by CTM in this example) and the variables 902.

[0123] As mentioned above (for example, in Figure 7), a continuous multimodal representation of a probability distribution of a target design 900 is generated in at least a portion of a latent space 812. In some embodiments, the target design 900 is a target substrate design for a semiconductor wafer. In some embodiments, the target design 900 includes an expected wafer pattern and / or intermediate data associated with the expected wafer pattern, the intermediate data including continuous transmission mask (CTM) data, CTM images, intermediate mask designs, and / or other target designs. The latent space 812 includes a distribution of feature variables 910, which can be used to generate a mask design based on the target design 900. For example, the feature can include a shape or structure associated with a target and / or mask design for a semiconductor device.

[0124] Variables 902 are selected from a continuous multimodal representation in latent space 812. Variables 902 include a latent space representation for determining one or more features of mask design 904. Selecting variables 902 includes selecting a pattern from the multimodal representation of the probability distribution and sampling variables from the selected pattern. In some embodiments, variables 902 include information content from the OPC domain and / or the SRAF domain (OPC / SRAF data 920), or propagates this information from the second neural network block (conditional mapping submodel 806) to the latent space 812.

[0125] The mask design 904 is determined based on the target design 900 and the variables 902 and / or other information. In some embodiments, the determined mask design 904 includes an image. In some embodiments, determining the mask design 904 based on the target design 900 and the variables 902 includes (1) mapping the target design 900, CTM data, and / or CTM image to the mask design 904, and / or (2) mapping the target design 900 to the CTM data and / or CTM image. In some embodiments, 920 is also changed to match 904. (So there may actually be three options, for example: 1.900=target, 920, 904=OPC / SRAF; 2.900=CTM, 920, 904=OPC / SRAF; and 3.900=target, 920, 904=CTM). In some embodiments, determining the mask design 904 includes determining sub-resolution assist features (SRAF) and / or optical proximity correction (OPC) data for the mask design 904. In some embodiments, SRAF data and OPC data are determined as separate contributions.

[0126] Fig. 9 The mathematics associated with the above operations are also shown. Fig. 9 In the above equation, o represents OPC / SRAF data, c represents CTM data, s represents the posterior mode selection probability, d represents the discrete classification 1-hot encoding and / or the classification sample output of s, and represents the inferred OPC / SRAF data. (In this example, h=conditional encoder model (as part of the conditional mapping submodel), l=latent variables, k=sum index, g=generative model (as part of U-net), N=normal distribution (mean=k / n and variance=1 / n), and n=latent space dimension. In some embodiments, h=conditional encoder model, ~ indicates that the variables are distributed according to a given distribution, and since d is a 1-hot encoding, the summation is a selection of variables. In addition, in the context of other terms used herein, f=first model or block, h=second model or block, and g=third model or block.)

[0127] Fig.10 An adversarial submodel 1000 is shown, which may be included in model 800 and / or used to train model 800. As described above (see Figure 7 ), the model 800 may be trained by classifying the output mask design 904 as false or true using the adversarial training sub-model 1000, such that after training, the adversarial sub-model cannot distinguish the output from the true reference data (e.g., OPC / SRAF data 920 in this example). In this case, the task of the model 800 is to trick the adversarial sub-model 1000 into classifying its output (e.g., 904) as true. This is configured to ensure that the output of the model 800 is indistinguishable from the true reference data and does not include spurious features. Note that in this example, d (as described above - see also the selected variable 902) is randomly chosen (1-hot encoded).

[0128] Fig.11 A forward consistency submodel 1100 is shown, which may be included in model 800 and / or used to train model 800. As described above (see Figure 7 ), forward consistency sub-modeling may be performed to ensure that the determined mask design 904 will create a desired semiconductor wafer structure corresponding to the target design. In some embodiments, the forward consistency sub-modeling is performed by a fixed physics model and / or a parametric model that approximates the physics of the semiconductor manufacturing process.

[0129] The forward consistency submodel is configured to ensure that the inferred OPC / SRAF features (for example) are suitable for creating the desired pattern on the wafer. The forward consistency submodel 1100 can be derived from the optical elements of the lithographic equipment (e.g., as described above) and the physical properties of the resist (e.g., such that the forward consistency submodel 1100 is a fixed physical model). Alternatively, the forward consistency submodel 1100 can be a parametric model that approximates the physics of one or more manufacturing processes, where the model parameters are based on experimental and / or empirical data. For example, the forward consistency submodel 1100 can be pre-trained (and / or built based on physical properties). The forward consistency submodel 1100 is configured to ensure that any sampling choice applied by d will result in a valid mask design, i.e., if d changes, then if the resulting (where f(c)→l k ) through the forward consistency sub-model 1100, a similar target design approximation t is still output, and remains unchanged (where m is the forward consistency submodel as described in this article).

[0130] In addition to the above sub-models, additional regularization / loss costs may be applied during training of model 800. As described above ( Figure 7 ), additional regularization / loss cost ( Figure 8 , Fig. 9 ). Applying a regularization / loss cost includes applying a cost term that penalizes the amount of jagged edges in a determined mask design, re-weighting a cost term that penalizes the amount of jagged edges, applying a cost term that prioritizes binary pixel values ​​in an image associated with a determined mask design, applying a fixed selection option for selecting the best mask design, and / or applying regularization to the difference between two versions of a mask design. This can ensure that the resulting inferred image output by the model 800 has features with sharp edges and is approximately rectangular, for example. This can be achieved by placing a penalty term on the inferred image gradient, similar to a total variation method. Additionally, a penalty can be added to second-order image gradients (e.g., cross terms in the XY direction). This is to ensure that the model 800 will prefer results with a small number of corners / non-jagged edges for the output OPC / SRAF data.

[0131] Note that the output OPC and SRAF data can be processed together or separately. By way of non-limiting example, Fig.12One embodiment of model 800 is shown in which OPC / SRAF contributions 1202 and 1204 are processed separately and then combined 1206 to generate mask design 904. From a regularization perspective, it is possible to make separate models for OPC and SRAF at a time (e.g., they may have different requirements in terms of rectangularity). However, another reason to separate them is that they may require different computational resolutions. For example, a SRAF model (which includes only rectangles) can use a coarser pixel resolution than an OPC model (which may have more complex polygons and finer details).

[0132] Training model 800 (including the first, second and third neural network blocks described above, such as Fig. 9 and Fig.12 ) involves using training data generated using the current method (described above). The first, second and third neural network blocks are trained together by optimizing the cost(s) in an adversarial manner, for example, by alternating two sub-optimization tasks in a scheme similar to the expectation-maximization method.

[0133] For example, Fig.13 The iterations involved in finding the joint solution for the above training optimization are schematically shown (as described above with respect to Figure 7 As described above, various operations may be performed simultaneously, sequentially, iteratively, etc.). Fig.13 As shown, multiple iterations of the optimization solver (e.g., multiple stochastic gradient descent steps) (operation 1302) can be used to partially solve the first optimization (keeping a (which represents the adversarial submodel described herein) and m (the forward consistency submodel) of model 800 fixed); and multiple (e.g., one or more) stochastic gradient descent steps (operation 1304) can be used to partially solve the second optimization (keeping f, g, h, and m of model 800 fixed). The first and second optimizations are described below. Fig.14 Operations 1302 and 1304 may be repeated 1306 until convergence and / or other stopping criteria are met.

[0134] According to and / or based on Fig.14 Formula 1 and Formula 2 shown are used to train the model 800 ( Figure 8 , Fig. 9 , Fig.12 etc.). Fig.14 As shown, formula (1) includes a fidelity term 1402 associated with the reference OPC / SRAF features; a variation term 1404, which is configured to ensure that the latent space (e.g., Figure 8 , Fig. 9 , Fig.12=The latent space 812 shown in FIG. 8 follows an appropriate (e.g., continuous multimodal) distribution; a target matching term 1406, where the function m is a known physical model (e.g., the forward submodel described herein), which is configured to map the mask design to the target design; an adversarial term 1408, which is configured to train the model 800 to create outputs that deceive the adversarial submodel. Formula (2) includes a discriminator training term 1410, which is configured to classify the training or reference samples and the outputs generated by the model 800.

[0135] In equations (1) and (2), B represents the predicted OPC / SRAF image, for example The binary cross entropy loss between each pixel of and the true value o; the latent variable l used in the variational prior k The average value μ k and variance σ k The KL divergence is respectively defined as, for example, and KL(s,n 分类 ) represents the KL divergence about a fixed n-class probability distribution, i.e., with equal probability for each class; represents an additional cost term that can constrain the solution properties for a given choice d. Note that for simplicity only one latent element and a single vector d are described. For more latent elements, the same process applies, with a choice applied at each latent position. The notation represents the conditional output of model 800 when d is generated based on known samples o, and the sampled output of model 800 when d is randomly generated (d is random but still 1-hot encoded) or uses a specific choice that does not depend on known samples o. This is an important distinction because o is only known during model training. During the application of model 800, a predefined vector d is provided that is configured to produce an appropriate mask design And the model 800 is trained in such a way that any sample d produces an appropriate target design.

[0136] Note that Fig. 9 , Fig.10 , Fig.11 , Fig.12 In the figure, a block represents a potential element or a potential element with a single element. This is to simplify the drawings, and adding more potential elements means repeating the same scheme.

[0137] The additional regularization cost term is further described below Note that not all of the following options need to be or should be applied simultaneously, i.e., multiple parameters β i can be set to 0. Similarly, for the parts in the cost function described here, the multiple parameters αi can also be set to zero. A first option for regularizing the cost term consists in regularizing the resulting image given different samples d The cost term that penalizes the amount of jagged edges in the image. Image gradient (d x ,d y : Use the difference between adjacent pixels on the x or y axis) or the second-order cross term d xy (applied on d y on x ). A penalty is imposed on the magnitude in the l1-norm (note that other norms can be used), resulting in the cost:

[0138]

[0139] These scaling factors β are used as configuration parameters. To ensure that this is valid for each condition / choice of a feature variable defined by a choice d, a penalty is imposed on samples drawn from the possible distribution of potential choices d. x ,d y The term is configured to ensure that the resulting mask design is piecewise constant / flat, while d xy The terms are configured to ensure that the mask features have a small number of corners.

[0140] A second option for regularizing the cost term includes reweighting the cost term, which penalizes the amount of jagged edges. Since a binary mask is sought, the cost can be reweighted so that it does not penalize values ​​of 0 or 1 (e.g., looking for low gradients or very steep gradients). The cost term from the following formula (4) becomes non-convex. However, given the impact of the use of a neural network model (described above), this does not cause a significant problem due to the inherent non-convexity. The term T2 describes such a cost:

[0141]

[0142] Note that the domain of the mapping is [0, 1], so Equation (4) is well behaved. Well behaved can mean that the terms in the cost function are non-negative and bounded (by 1). The upper limit of 1 is somewhat arbitrary because in the coefficients β i Any positive maximum value can be absorbed in. By using an appropriate activation function, the output of the model can be constrained between 0 and 1. Therefore, (4) will only evaluate outputs that belong to this interval. If the model does not have these limits, the minimum value of (4) is located at - infinity, which is not the intended target of T2. In fact, T2 is only used when the model (the third model) can only output values ​​between 0 and 1. The same is true for the above formula (3). Traditional iterative reweighting schemes can also be used.

[0143] A third option for regularizing the cost term consists in placing priority on cost terms with binary pixel values ​​(values ​​of 0 or 1 rather than anything in between), yielding

[0144]

[0145] A fourth option for regularizing the cost term includes a fixed selection option for the best choice, d best for all targets, resulting in a cost of:

[0146]

[0147] is minimized. This allows for the selection of a "best" or otherwise optimized result.

[0148] Alternatively, to reduce the bias introduced by regularization, regularization can be applied to the difference between two versions of the mask design, e.g. or This ensures that two possible mask designs associated with the same CTM (eg, target design) differ only by a small amount and that these differences do not have many (jagged) edges.

[0149] The choice of cost used may vary in cases where certain terms are dropped or all described sub-models are not included (e.g., the adversarial sub-model is not included). Also, to simplify notation, the cost used to generate the output image is not explicitly described. The sampling process, although in the above Fig. 9 , Fig.10 and Fig.11 The sampling process is shown in FIG.

[0150] In training model 800 (in Figure 8 , Fig. 9 , Fig.12 After that, by providing a predefined selection d (for example, via the above-mentioned forward submodel m (for example, Fig.11 The sub-model 1100 in (a) is constrained to be the selection with the "best" performance in terms of the resulting target design) to determine (e.g., perform inference) a mask design (e.g., including OPC / SRAF data).

[0151] Fig.15 An example of using the trained model 800 to infer (or otherwise determine) a mask design 904 (with OPC / SRAF features) is shown. Fig.15, the variable 902 selection d is determined during training to produce the "best" target image (among other possible variables that may have been selected). The resulting mask design 904 may be further processed 1502 via conventional methods (e.g., to produce mask design 904a) to correct for any possible small details that, for example, cannot be manufactured. For example, in some embodiments, the sub-model 1100 may be used to evaluate performance.

[0152] Fig.16 A schematic diagram of the model 800 shown in the previous figure is provided, which is reconfigured to train a fixed latent selection d k , to achieve target performance levels for predefined key performance indicators (KPIs) regarding semiconductor manufacturing processes. k represents the variable selection associated with the key performance indicator k. Such a model can be used to quantify different perturbations in the manufacturing process to ensure that the mask prints an effective target design for a wide manufacturing process window. In some embodiments, the target design (e.g., input CTM image and / or data) and the fixed selection option d k (For example, for each potential position, one of the middle elements is selected, so that the fixed selection d k The model 800 is trained using the perturbation parameters (the perturbation is optimal relative to a given process window).

[0153] Fig.16 Two different possible example options (option 1602 and option 1604) are shown for model 800, which is reconfigured for training with a fixed potential choice d k, to achieve a target performance level for a predefined key performance indicator (KPI) for a semiconductor manufacturing process. Option 1602 is associated with a process window metric. In option 1602, an output mask design 904 (e.g., a resulting OPC / SRAF mask) is passed through a process window model 1606 that models a (random, pseudo-random, or predefined) perturbation 1608 associated with a process window variation. The perturbation 1608 may be, for example, a sample of a statistical pattern of physical variations that may occur during a manufacturing process. The output of the model 1606 is a target design 1610 determined based on the process perturbation 1608 and the input mask design 904. For example, the model 1606 may be a forward model that includes a resist model that is perturbed to model small variations in the resist (other examples are contemplated). In some embodiments, a cost configured to bring the same performance about the resulting target design over a range of perturbations may be added. Here e is a given resulting target for a given perturbation 1608. The system is configured so that for any perturbation, e should be close to the desired target t of the design, and so this can be added as a cost term during training. The symbol p represents the process window model. It can be similar to the forward consistency sub-model, but extended with additional process variation parameters or other scanner perturbations such as focus and dose variations.

[0154] Option 1604 is associated with the OPC / SRAF mask design feature. Option 1604 can be used to place a penalty term directly on the OPC / SRAF feature. An example of such a penalty term is the regularization option described above, which is configured to produce an OPC / SRAF feature with a (or close to) rectangular shape. The above operation is used to randomly select d, while in option 1604, there is a penalty term that is calculated via d. k The specific determinations made about the variables. This can be expanded depending on what criteria are important for the mask design (e.g., option 1604 can be configured to restrict multiple features to reduce the cost of manufacturing the mask, increase the cost of designing a free-form OPC mask instead of a rectangular cost term, etc.). In option 1604, the output mask design 904 (e.g., the resulting OPC / SRAF mask) is passed through a mask characteristic model 1620. The mask characteristic model h effectively converts these mask constraints into some (single or multiple) numbers. On these numbers, the model imposes a penalty on the cost during model training, for example, to ensure that the expected value is close.

[0155] Fig.17 An example embodiment of a model 800 is shown that is configured to account for or otherwise incorporate lithography scanner focus disturbances (eg, using the process window model 1606 described above). Fig.17 It is shown how to augment the above training cost option with an additional term that encodes the target approximation of the perturbation distribution. This cost term can be, for example, the perturbation target design The mean square error between the expected target design t Note that during training, samples for perturbations may be drawn from a distribution of possible perturbations (shown by 1608), and thus the model 800 is trained for optimization across the entire perturbation distribution, which is defined a priori. Fig.17 A sampling 1700 of the distribution of possible focus disturbances in a scanner, a focus model 1702 of an aerial image 1704, a forward model 1706 associated with the resist, and / or other models are shown.

[0156] Note that adjustments can be made to the semiconductor manufacturing process based on the model output and / or other information. For example, the adjustment can include changing one or more semiconductor manufacturing process parameters. The adjustment can include pattern parameter changes (e.g., size, position, and / or other design variables), and / or any adjustable parameters, such as adjustable parameters of the etching system, source, pattern forming device, projection optics, dose, focus, etc. The parameters can be automatically or otherwise electrically adjusted by a processor (e.g., a computer controller), manually adjusted by a user, or otherwise adjusted. In some embodiments, a parameter adjustment can be determined (e.g., the amount by which a given parameter should be changed), and the parameter can be adjusted from a priori parameter set point to a new parameter set point, for example.

[0157] Fig.18 is an example computer system CS (which may be similar to Figure 3 CS shown in or with Figure 3 The computer system CS is the same as the CS shown in FIG. 1 . The computer system CS includes a bus BS or other communication mechanism for transmitting information, and a processor PRO (or multiple processors) coupled to the bus BS for processing information. The computer system CS also includes a main memory MM coupled to the bus BS, such as a random access memory (RAM) or other dynamic storage device, for storing information and instructions to be executed by the processor PRO. The main memory MM can also be used to store temporary variables or other intermediate information during the execution of instructions by the processor PRO. The computer system CS also includes a read-only memory (ROM) ROM or other static storage device coupled to the bus BS for storing static information and instructions for the processor PRO. A storage device SD such as a magnetic disk or optical disk is provided and coupled to the bus BS for storing information and instructions.

[0158] The computer system CS can be coupled to a display DS via a bus BS, such as a cathode ray tube (CRT) or a flat panel or touch panel display, for displaying information to a computer user. An input device ID including alphanumeric keys and other keys is coupled to the bus BS for transmitting information and command selections to the processor PRO. Another type of user input device is a cursor control CC, such as a mouse, a trackball or cursor direction keys, for transmitting direction information and command selections to the processor PRO, and for controlling cursor movement on the display DS. The input device typically has two degrees of freedom in two axes, a first axis (e.g., x) and a second axis (e.g., y), which allows the device to specify a position in a plane. A touch panel (screen) display can also be used as an input device.

[0159] In some embodiments, the part of one or more operations described herein can be executed by the computer system CS in response to the processor PRO executing one or more sequences of one or more instructions contained in the main memory MM. Such instructions can be read into the main memory MM from another computer-readable medium such as a storage device SD. The execution of the instruction sequence included in the main memory MM makes the processor PRO perform the processing steps (operations) described herein. One or more processors in a multiprocessing device can also be used to execute the instruction sequence contained in the main memory MM. In some embodiments, a hard-wired circuit system can be used to replace a software instruction or be combined with a software instruction. Therefore, the description herein is not limited to any specific combination of hardware circuit system and software.

[0160] The term "computer-readable medium" as used herein refers to any medium that participates in providing instructions to the processor PRO for execution. Such a medium can take many forms, including but not limited to non-volatile media, volatile media and transmission media. Non-volatile media include, for example, optical or magnetic disks, such as storage devices SD. Volatile media include dynamic memory, such as main memory MM. Transmission media include coaxial cables, copper wires and optical fibers, including wires, which include bus BS. Transmission media can also take the form of sound waves or light waves, such as sound waves or light waves generated during radio frequency (RF) and infrared (IR) data communications. Computer-readable media can be non-transient, such as floppy disks, flexible disks, hard disks, magnetic tapes, any other magnetic media, CD-ROMs, DVDs, any other optical media, punch cards, paper tapes, any other physical media with hole patterns, RAMs, PROMs, EPROMs, FLASH-EPROMs, any other memory chips or cassette memories. Non-transient computer-readable media may have (machine-readable) instructions recorded thereon. When executed by a computer, these instructions can implement any operation described herein. For example, transitory computer readable media may include carrier waves or other propagated electromagnetic signals.

[0161] Various forms of computer-readable media may be involved in carrying one or more sequences of one or more machine-readable instructions to the processor PRO for execution. For example, the instructions may initially be carried on a disk of a remote computer. The remote computer may load the instructions into its dynamic memory and send the instructions over a telephone line using a modem. The modem local to the computer system CS may receive the data on the telephone line and convert the data into an infrared signal using an infrared transmitter. An infrared detector coupled to the bus BS may receive the data carried in the infrared signal and place the data on the bus BS. The bus BS carries the data to the main memory MM, from which the processor PRO retrieves and executes the instructions. The instructions received by the main memory MM may optionally be stored on a storage device SD before or after being executed by the processor PRO.

[0162] The computer system CS may also include a communication interface CI coupled to the bus BS. The communication interface CI provides a bidirectional data communication coupled to a network link NDL, which is connected to a local area network LAN. For example, the communication interface CI may be an integrated services digital network (ISDN) card or a modem to provide a data communication connection to a corresponding type of telephone line. As another example, the communication interface CI may be a local area network (LAN) card to provide a data communication connection to a compatible LAN. A wireless link may also be implemented. In any such implementation, the communication interface CI sends and receives electrical, electromagnetic or optical signals carrying digital data streams representing various types of information.

[0163] The network link NDL typically provides data communications to other data devices through one or more networks. For example, the network link NDL can provide a connection to the host computer HC through a local area network LAN. This can include data communication services provided by the global packet data communication network (now generally referred to as the "Internet" INT). The local area network LAN (Internet) can use electrical, electromagnetic or optical signals that carry digital data streams. The signals through the various networks and the signals on the network data link NDL and the signals through the communication interface CI (which carries digital data to and from the computer system CS) are exemplary forms of carrier waves that transmit information.

[0164] The computer system CS can send messages and receive data, including program code, through (multiple) networks, network data links NDL and communication interfaces CI. In the Internet example, the host HC can send the request code of the application program through the Internet INT, the network data link NDL, the local area network LAN and the communication interface CI. For example, such a downloaded application program can provide all or part of the method described herein. The received code can be executed by the processor PRO when it is received, or stored in the storage device SD or other non-volatile memory for later execution. In this way, the computer system CS can obtain the application program code in the form of a carrier wave.

[0165] The concepts disclosed herein can be used with any imaging, etching, polishing, inspection, etc. system for sub-wavelength features, and can be used with emerging imaging technologies that can produce shorter and shorter wavelengths. Emerging technologies include EUV (extreme ultraviolet), DUV lithography that can produce 193nm wavelengths using ArF lasers, and even 157nm wavelengths using fluorine lasers. In addition, EUV lithography can produce wavelengths in the range of 20-50nm by using synchrotrons or by bombarding materials (solid or plasma) with high energy electrons to produce photons in this range.

[0166] Embodiments of the present disclosure may be further described by the following clauses.

[0167] 1. A non-transitory computer readable medium having instructions thereon, which, when executed by a computer, cause the computer to perform a method of determining a mask design, the method comprising:

[0168] generating a continuous multimodal representation of a probability distribution of a target design in at least a portion of a latent space, the latent space comprising a distribution of feature variables that can be used to generate a mask design based on the target design;

[0169] selecting variables from the continuous multimodal representation in the latent space, the variables comprising a latent space representation for determining one or more features of the mask design; and

[0170] The mask design is determined based on the target design and the variables.

[0171] 2. A method according to clause 1, wherein selecting the variable includes selecting a mode from the multimodal representation of the probability distribution and sampling the variable from the selected mode.

[0172] 3. A method according to any of the preceding clauses, wherein the generating, the selecting and the determining are performed by an encoder structure and a generating structure having a conditional mapping sub-model.

[0173] 4. A method according to clause 3, wherein the encoder structure and the generation structure form a U-net type deep learning model.

[0174] 5. A method according to clause 4, wherein the U-net type deep learning model with the conditional mapping submodel includes a first neural network block, a second neural network block and a third neural network block, the first neural network block is configured to generate the continuous multimodal representation of the probability distribution of the target design in the portion of the latent space, the second neural network block is configured to select the variables during training, and the third neural network block is configured to determine the mask design based on the target design and the variables.

[0175] 6. A method according to clause 5, wherein the first neural network block, the second neural network block and the third neural network block are trained jointly.

[0176] 7. A method according to clause 5 or 6, wherein the second neural network block is trained to generate a distribution of the feature variables present in the input sub-resolution assist feature SRAF and / or optical proximity correction OPC data.

[0177] 8. A method according to clause 6 or 7, wherein during training, the selected variables are used as ground truth values ​​to train the third neural network block to generate the mask design based on the input target design and the mode selection options given the selected variables.

[0178] 9. A method according to clause 8, wherein the variables include information content from the OPC and / or SRAF domains, or the propagation of said information from the second neural network block to the latent space.

[0179] 10. The method according to any one of clauses 5 to 9 further includes: training the first neural network block, the second neural network block and the third neural network block by classifying the output mask design as false or true using an adversarial training sub-model, so that after training, the adversarial sub-model cannot distinguish the output of the third neural network block from the real reference data.

[0180] 11. The method according to any one of clauses 5 to 10 further comprises: applying additional regularization / loss costs during the training of the first neural network block, the second neural network block and the third neural network block.

[0181] 12. A method according to claim 11, wherein applying the regularization / loss cost includes: applying a cost term that penalizes the number of jagged edges in the determined mask design, reweighting the cost term that penalizes the number of jagged edges, applying a cost term that prioritizes binary pixel values ​​in the image associated with the determined mask design, applying a fixed selection option for selecting the best mask design, and / or applying regularization on the difference between two versions of the mask design.

[0182] 13. A method according to any of the preceding clauses, wherein the target design comprises an expected wafer pattern and / or intermediate data associated with the expected wafer pattern, the intermediate data comprising continuous transmission mask (CTM) data, a CTM image and / or an intermediate mask design.

[0183] 14. A method according to clause 13, wherein determining the mask design based on the target design and the variables includes: (1) mapping the target design, the CTM data and / or the CTM image to the mask design, and / or (2) mapping the target design to the CTM data and / or the CTM image.

[0184] 15. A method according to any of the preceding clauses, wherein the latent space models the distribution of the feature variables, which can be used to generate mask designs via variational Bayesian inference techniques.

[0185] 16. A method according to any of the preceding clauses, wherein the feature comprises a shape or structure associated with a target and / or reticle design of a semiconductor device.

[0186] 17. The method of any of the preceding clauses, further comprising performing forward consistency sub-modeling configured to ensure that the determined mask design will create a desired semiconductor wafer structure corresponding to the target design.

[0187] 18. The method of clause 17, wherein the forward consistent sub-modeling is performed by a fixed physical model and / or a parametric model that approximates the physics of a semiconductor manufacturing process.

[0188] 19. The method of any of the preceding clauses, wherein determining the mask design comprises determining sub-resolution feature (SRAF) and / or optical proximity correction (OPC) data for the mask design.

[0189] 20. The method of clause 19, wherein the SRAF data and the OPC data are determined as separate contributions.

[0190] 21. The method of any of the preceding clauses, wherein the target design is a target substrate design for a semiconductor wafer.

[0191] 22. A method according to any of the preceding clauses, wherein the mask design determined comprises an image.

[0192] 23. The method according to any of the preceding clauses further includes sampling the resulting conditional latent space by generating multiple selection options; and evaluating the process window key performance indicators for the resulting mask design to determine the most robust mask that the pre-trained model can produce.

[0193] 24. The method according to any of the preceding clauses further includes constructing an optimization problem and evaluating the process window key performance indicators of the resulting mask design based on the output from the optimization problem, thereby determining the most robust mask that the pre-trained model can produce.

[0194] 25. The method according to any of the preceding clauses, further comprising fixing a given latent parameterization and training the model to optimize various process window key performance indicators given a perturbation of the process window.

[0195] 26. A non-transitory computer-readable medium having instructions thereon that, when executed by a computer, cause the computer to perform the method of any one of clauses 1 to 23.

[0196] 27. A method of determining a semiconductor mask design utilizing a model that learns a multimodal distribution of mask features and selects variables that result in efficient semiconductor wafer imaging, the method comprising:

[0197] generating, using a first neural network module of the model, a continuous multimodal representation of a probability distribution of a wafer target design in at least a portion of a latent space, the latent space comprising a distribution of feature variables that can be used to generate a mask design based on the target design;

[0198] selecting, using a second neural network block of the model and during training of the model, variables from the continuous multimodal representation in the latent space, the variables comprising a latent space representation to be used to determine one or more features of the mask design, wherein the selecting comprises selecting a mode from the multimodal representation of the probability distribution and sampling the variables from the selected mode; and

[0199] determining the mask design based on the target design and the variables using a third neural network block of the model,

[0200] The model is a U-net type deep learning model with a conditional mapping sub-model.

[0201] While the concepts disclosed herein may be used to fabricate with substrates such as silicon wafers, it should be appreciated that the disclosed concepts may be used with any type of fabrication system, such as those used to fabricate on substrates other than silicon wafers.

[0202] Furthermore, combinations and sub-combinations of disclosed elements may include separate embodiments. For example, one or more of the operations described above may be included in separate embodiments, or they may be included together in the same embodiment.

[0203] The above description is intended to be illustrative rather than limiting. It will therefore be apparent to those skilled in the art that modifications may be made as described without departing from the scope of the claims set forth below.

Claims

1. A non-transitory computer readable medium having instructions thereon, which, when executed by a computer, cause the computer to perform a method of determining a mask design, the method comprising: generating a continuous multimodal representation of a probability distribution of a target design in at least a portion of a latent space, the latent space comprising a distribution of feature variables that can be used to generate a mask design based on the target design; selecting variables from the continuous multimodal representation in the latent space, the variables comprising a latent space representation for determining one or more features of the mask design; as well as The mask design is determined based on the target design and the variables.

2. The medium of claim 1, wherein selecting the variable comprises selecting a mode from the multimodal representation of the probability distribution and sampling the variable from the selected mode.

3. The medium of claim 1, wherein the generating, the selecting, and the determining are performed by an encoder structure and a generating structure having a conditional mapping sub-model.

4. A medium according to claim 3, wherein the encoder structure and the generation structure form a deep learning model, wherein the deep learning model with the conditional mapping sub-model includes a first neural network block, a second neural network block and a third neural network block, the first neural network block is configured to generate the continuous multimodal representation of the probability distribution of the target design in the portion of the latent space, the second neural network block is configured to select the variables during training, and the third neural network block is configured to determine the mask design based on the target design and the variables.

5. A medium according to claim 4, wherein the first neural network block, the second neural network block and the third neural network block are trained jointly, and wherein the second neural network block is trained to generate the distribution of the feature variables present in the input sub-resolution auxiliary feature SRAF and / or optical proximity correction OPC data.

6. The medium according to claim 4, wherein During training, the selected variables are used as ground truth to train the third neural network block to generate the mask design based on the input target design and the pattern selection options given the selected variables.

7. A medium according to claim 6, wherein the variables include information content from the OPC and / or SRAF domains, or the propagation of said information from the second neural network block to the latent space.

8. A medium according to claim 5, wherein the method further comprises training the first neural network block, the second neural network block and the third neural network block by classifying the output mask design as fake or real using an adversarial training sub-model, so that after training, the adversarial sub-model cannot distinguish the output of the third neural network block from real reference data.

9. The medium of claim 5, wherein the method further comprises: During the training of the first neural network block, the second neural network block, and the third neural network block, additional regularization / loss costs are applied, wherein applying the regularization / loss costs comprises: applying a cost term that penalizes the number of jagged edges in the determined mask design, reweighting the cost term that penalizes the number of jagged edges, applying a cost term that prioritizes binary pixel values ​​in the image associated with the determined mask design, applying a fixed selection option for selecting an optimal mask design, and / or applying regularization on the difference between two versions of the mask design.

10. The medium of claim 1, wherein the target design comprises an expected wafer pattern and / or intermediate data associated with the expected wafer pattern, the intermediate data comprising continuous transmission mask (CTM) data, a CTM image, and / or an intermediate mask design, and wherein determining the mask design based on the target design and the variables comprises: (1) mapping the target design, the CTM data and / or the CTM image to the mask design, and / or (2) mapping the target design to the CTM data and / or the CTM image.

11. A medium according to claim 1, wherein the method further includes performing forward consistency sub-modeling, wherein the forward consistency sub-modeling is configured to ensure that the determined mask design will create an expected semiconductor wafer structure corresponding to the target design, wherein the forward consistency sub-modeling is performed by a fixed physical model and / or a parametric model that approximates the physics of the semiconductor manufacturing process.

12. The medium of claim 1, wherein determining the mask design comprises determining sub-resolution assist feature (SRAF) and / or optical proximity correction (OPC) data of the mask design, and wherein the SRAF data and the OPC data are determined as separate contributions.

13. The medium of claim 1, wherein the method further comprises sampling the resulting conditional latent space by generating a plurality of selection options; and evaluating a process window key performance indicator of the resulting mask design to determine the most robust mask that the pre-trained model is capable of producing.

14. The medium of claim 1, wherein the method further comprises constructing an optimization problem and evaluating a process window key performance indicator of a resulting mask design based on an output from the optimization problem to determine the most robust mask that the pre-trained model is capable of producing.

15. The medium of claim 1, wherein the method further comprises fixing a given latent parameterization and training the model to optimize for various process window key performance indicators for a given process window perturbation.

Images