Methods and information processing devices for determining the set of sample injection regions

Abstract

A method and information processing apparatus for determining a set of sample injection regions are disclosed. The method provides a method for determining a set of sample injection regions from multiple injection regions of a substrate, wherein the position of a mark in each sample injection region is actually measured. The method includes: setting an initial arrangement of the set of sample injection regions, and adding injection regions to the set of sample injection regions other than those in the initial arrangement, such that the uncertainty of the estimated value of the measured value of the position of the mark obtained using an estimation model exceeds a predetermined threshold in that injection region.

Description

Technical Field

[0001] This invention relates to a method for determining a set of sample injection regions, a method for obtaining measurement values, an information processing apparatus, a photolithography apparatus, a storage medium, and a method for manufacturing articles. Background Technology

[0002] With the micropatterning and higher integration of devices, the demand for improved device alignment accuracy is growing. To improve alignment accuracy, it is necessary to inspect alignment offsets (hereinafter referred to as overlays) with high precision and control the offset amount based on the alignment offsets. Therefore, the demand for increased overlay inspection accuracy is also increasing.

[0003] To perform alignment and overlay inspections with high precision, it is necessary to increase the number of measurement points on the target object (such as a wafer or panel, hereinafter referred to as the substrate). However, increasing the number of measurement points leads to increased measurement / inspection time and reduced productivity. To prevent this, International Publication No. 2018 / 133999 and U.S. Patent No. 10545412 each propose a Virtual Metrology system. This system uses statistical models based on data from various sensors during device operation, operation logs in the device manufacturing process, the type of device used to process the substrate, etc., to estimate the values ​​of points that do not actually undergo measurement / inspection. This achieves a virtual increase in the number of measurement points.

[0004] Virtual measurement systems enable a reduction in the number of measurement points. However, currently, the number of measurement points is reduced through trial and error, and no clear method for reduction has been established. Summary of the Invention

[0005] The present invention provides, for example, a technique for determining a measurement position on a substrate, which is advantageous in improving the alignment accuracy of the substrate.

[0006] In one aspect, the present invention provides a method for determining a set of sample injection regions from a plurality of injection regions of a substrate, wherein the position of a mark in each sample injection region is to be actually measured, the method comprising the steps of: setting an initial arrangement of the set of sample injection regions, and adding to the set of sample injection regions, in addition to the sample injection regions in the initial arrangement, an injection region indicating that the uncertainty of the estimated value of the measured value of the position of the mark obtained using an estimation model exceeds a predetermined threshold.

[0007] The present invention provides, in a second aspect, a method for obtaining a measurement of the position of a mark in each of a plurality of ejection regions of a substrate, comprising: measuring the position of a mark in each sample ejection region included in a set of sample ejection regions determined using the method according to the first aspect; and estimating a measurement of the position of a mark in each ejection region not included in a set of sample ejection regions by using an estimation model and based on the measurement results of the marks in each sample ejection region included in the set of sample ejection regions.

[0008] In a third aspect, the present invention provides an information processing apparatus for determining a set of sample injection regions from a plurality of injection regions of a substrate, wherein the position of a mark in each sample injection region is to be actually measured, comprising: a processing unit, wherein the processing unit sets an initial arrangement of the set of sample injection regions and adds to the set of sample injection regions an injection region, in addition to the sample injection regions in the initial arrangement, an injection region indicating that the uncertainty of the estimated value of the measured value of the position of the mark obtained using an estimation model exceeds a predetermined threshold.

[0009] The present invention provides a photolithography apparatus in a fourth aspect, comprising: a substrate stage configured to position a substrate based on measurements of the position of a mark obtained using the method according to the second aspect.

[0010] In its fifth aspect, the present invention provides a computer-readable storage medium storing a program for causing a computer to perform the steps of the method according to the first aspect.

[0011] The present invention provides, in its sixth aspect, a method for manufacturing an article, comprising: forming a pattern on a substrate by using a photolithography apparatus according to the fourth aspect; and processing the substrate on which the pattern has been formed, wherein an article is manufactured from the processed substrate.

[0012] Further features of the present invention will become clear from the following description of exemplary embodiments (with reference to the accompanying drawings). Attached Figure Description

[0013] Figure 1 This is a view showing the arrangement of the exposure apparatus;

[0014] Figure 2 This is a view showing the arrangement of the alignment detection optical system;

[0015] Figure 3 This is a flowchart of the exposure process;

[0016] Figure 4 This is a view showing an example of the layout of the sample ejection region on the substrate;

[0017] Figure 5 It is a view in which the alignment measurements in each sample injection region are represented by vectors;

[0018] Figure 6 It is a view in which the virtual alignment measurements in each injection region are represented by vectors;

[0019] Figure 7 This is a view showing an example configuration of a virtual measurement system for estimating alignment measurements;

[0020] Figure 8 This is a view showing the data flow during the learning process of the alignment measurement estimation model;

[0021] Figure 9 This is an exemplary view showing the variance of the probability distribution of the estimated error of the alignment measurements in each injection zone;

[0022] Figure 10 This is a view showing an example of a sample injection region selected based on variance;

[0023] Figure 11 This is a flowchart illustrating the process of determining the sample injection region;

[0024] Figure 12 This is a view showing an example of the location and variance of the initial sample injection region;

[0025] Figure 13 This is a view showing an example of the location and variance of the injection region of the sample after treatment;

[0026] Figure 14 This is a view showing an example configuration of a virtual measurement system for estimating coverage measurements;

[0027] Figure 15 This is a view showing the data flow during the learning process of the model for estimating the covered measurements; and

[0028] Figure 16 This is a flowchart illustrating the process of determining the sample injection region. Detailed Implementation

[0029] In the following description, embodiments will be illustrated in detail with reference to the accompanying drawings. Note that the following embodiments are not intended to limit the scope of the claimed invention. Several features are described in the embodiments, but this does not limit the invention to claiming all such features, and multiple such features may be suitably combined. Furthermore, in the drawings, the same or similar configurations are given the same reference numerals, and redundant descriptions thereof are omitted.

[0030] This invention provides a technique for determining the position of measurement points on a substrate, which is advantageous in improving the alignment accuracy of the substrate, and a photolithography apparatus incorporating such a technique will be described below. As a specific example, an example in which the invention is applied to an exposure apparatus, which is an example of a photolithography apparatus, will be described below. However, the photolithography apparatus is not limited to an exposure apparatus but can be other photolithography apparatuses. For example, the photolithography apparatus can be a drawing apparatus that performs drawing on a substrate (with a photosensitizer thereon) using a beam of charged particles. Alternatively, the photolithography apparatus can be an imprinting apparatus that forms a pattern on a substrate by molding an imprinting material onto the substrate using a mold.

[0031] <First Embodiment>

[0032] Figure 1 This is a view illustrating the arrangement of the exposure apparatus 1 according to an embodiment. The exposure apparatus 1 includes a projection optics system 3, a substrate chuck 5, a substrate stage 6, and an alignment detection optics system 7. A pattern (e.g., a circuit pattern) to be transferred to a substrate 4 is drawn on a master plate 2, also referred to as an intermediate mask, and the master plate 2 is supported by a master plate stage (not shown). The projection optics system 3 projects the master plate 2, on which the pattern (e.g., a circuit pattern) is drawn, onto the substrate 4. The substrate chuck 5 holds the substrate 4, in which the underlying pattern and alignment marks have already been formed in a previous step.

[0033] like Figure 4 As shown, multiple injection areas 41 are arranged in a grid pattern on the substrate 4, each injection area being a transfer area of ​​the pattern of the original 2. In each injection area, a lower-layer pattern transferred (formed) in a previous step and alignment marks are formed. Generally, the same pattern is formed in each injection area. In a predetermined injection area 43 among the multiple injection areas 41, pre-alignment marks 11 are formed for use in coarse alignment. Further, in a region among the multiple injection areas 41 predetermined as a sample injection area 42, fine alignment marks 12 are formed for use in fine alignment.

[0034] The substrate stage 6 positions the substrate 4 at a predetermined location. The alignment detection optical system 7 may include an imaging device that captures alignment marks on the substrate 4 and obtains an image of the marks. The controller C generally controls the various units used for exposure processing. The controller C may be implemented by an information processing device (computer) including a CPU and memory. Such a computer may include a display unit that displays various information. In this embodiment, the controller C may be used as a processing unit that performs processing to obtain the position of the marks by processing the image obtained by the alignment detection optical system 7.

[0035] In the example, controller C uses alignment detection optics 7 to detect the relative position between the original 2 and the substrate 4, and performs alignment by controlling the substrate stage 6 based on the detection result. After this, controller C causes illumination optics (not shown) to emit exposure light, and projection optics 3 uses the exposure light to project the pattern drawn in the original 2 onto the substrate 4.

[0036] Figure 2 The arrangement of the alignment detection optical system 7 is shown. Illumination light from the light source 8 is reflected by the beam splitter 9, passes through the lens 10, and illuminates the alignment mark 11 on the substrate 4. Diffracted light from the alignment mark passes through the lens 10, the beam splitter 9, and the lens 13, forming an image of the alignment mark 11 on the sensor 14. The sensor 14 photoelectrically converts the formed image into an image signal and sends the image signal to the controller C. The controller C performs position detection processing on the received image signal to measure the position of the alignment mark. The controller C performs such mark position measurements on measurement points on multiple sample injection areas in the substrate 4, which will be described later, and uses virtual measurement technology to obtain virtual measurement values ​​of the mark positions in the injection areas other than the sample injection areas. Based on the measurement values ​​obtained as described above in each injection area, the controller C aligns the position of the substrate stage 6 with the original 2.

[0037] The following will refer to Figure 3 The flowchart describes a series of steps in the exposure apparatus 1 from substrate alignment to exposure. In step S301, the substrate 4 is loaded into the exposure apparatus 1, and the substrate 4 is held by the substrate chuck 5.

[0038] In step S302, a pre-alignment measurement is performed. During the pre-alignment measurement, the position of the pre-alignment mark 11 on the substrate is detected by the alignment detection optical system 7. The alignment detection optical system 7 detects the pre-alignment mark 11 in multiple injection areas used for pre-alignment, forming the pre-alignment mark 11 in each injection area used for pre-alignment. Based on the detection results, the controller C calculates the offset and first-order linear components (magnification and rotation) of the entire substrate.

[0039] Next, in step S303, controller C sets the optimal arrangement of sample injection regions for fine alignment. The sample injection region is one of multiple injection regions on the substrate, and the marked location of that region should be actually measured. A method for determining the set of sample injection regions will be described later.

[0040] Next, in step S304, a fine alignment measurement is performed. In the fine alignment measurement, the controller C drives the substrate stage 6 based on the pre-alignment measurement results so that the alignment detection optical system 7 can observe the position of the fine alignment mark 12. Subsequently, the controller C uses the alignment detection optical system 7 to measure the position of the precise alignment mark 12 in the sample injection area.

[0041] In step S305, controller C estimates the alignment measurements in the injection regions other than the sample injection region using an estimation model of the measured values ​​of the marked positions (hereinafter referred to as "alignment measurements"). That is, it is not necessary to perform actual measurements in the injection regions other than the sample injection region. Based on the actual measurements of the sample injection region and the estimated alignment measurements in the injection regions other than the sample injection region, accurate alignment measurements for all injection regions on the substrate are obtained.

[0042] In step S306, for each injection area, the controller C drives the substrate stage 6 based on the fine alignment measurement results obtained in steps S304 and S305, projects the pattern of the original 2 onto the substrate via the projection optics system 3, and exposes the substrate. Then, in step S307, the exposed substrate is unloaded.

[0043] When distortion occurs in substrate 4, controller C has the function of correcting higher-order deformation components. This function will be described. Here, an example of a third-order polynomial model is shown, but the correction model is not limited to the correction model shown in this embodiment. Models of any order can be used, or other models besides polynomial models can be used.

[0044] When the deformation in the substrate is represented by a third-order polynomial model, the correction values ​​(offset X (ShiftX) and offset Y (ShiftY)) for each injection region are represented by the following equation (1):

[0045] ShiftX = k1 + k3x + k5y + k7x 2 +k9xy+k 11 y 2 +k 13 x 3 +k 15 x 2 y+k 17 xy 2 +k 19 y 3

[0046] ShiftY = k2 + k4y + k6x + k8y 2 +k 10 xy+k 12 x 2 +k14 y 3 +k 16 xy 2 +k 18 x 2 y+k 20 x 3 ...(1) where x and y indicate the location of the injection region on the substrate surface. Controller C obtains regression coefficients k1 to k based on the actual alignment measurements in each injection region. 20 To calculate the correction value.

[0047] For example, to obtain actual measurement data, alignment measurements can be performed on certain injection regions on the substrate. The injection regions used in this case are called sample injection regions. Figure 4 The example shown illustrates a set of 14 sample injection regions. A larger number of sample injection regions is required to correct for higher-order substrate deformation components. However, due to the trade-off between increasing the number of sample injection regions and alignment measurement time, the number of sample injection regions is determined while considering device productivity.

[0048] The virtual measurement system will be described next. It should be noted that the model structure shown here is merely an example, and the invention is not limited to this model.

[0049] Here, let's assume... Figure 4 The set of sample injection regions was defined as shown. Actual alignment measurements were performed only within these sample injection regions. Figure 5 In this model, the alignment measurements dx and dy obtained in each sample injection region are represented by vectors. Each vector corresponds to the alignment measurement value in each sample injection region. Controller C uses various data to estimate the alignment measurements in injection regions where no actual measurements are performed (i.e., injection regions other than the sample injection regions), and obtains... Figure 6 The virtual alignment measurement values ​​shown are illustrated. This is an overview of the virtual measurement system.

[0050] The estimation function of the system will be described below. Figure 7 This illustrates a configuration example of a virtual measurement system. In this system, the following data is used as input data for alignment measurement estimation model 200:

[0051] • Coverage measurement value 101 in the previous step (coverage measurement value between the upper and lower layers);

[0052] • Device manufacturing process parameter 102;

[0053] • Exposure device sensor data 103; and

[0054] • Sample alignment measurement value 104 is the alignment measurement value in the sample injection area obtained by actual measurement.

[0055] Device manufacturing process parameters 102 may include, for example, parameters related to device manufacturing, such as the ID of the device manufacturing apparatus that processed the substrate, the amount of substrate warpage occurring during device manufacturing, and apparatus parameters when applying resist to the substrate. Exposure apparatus sensor data 103 may include, for example, the pressure value when the substrate is fixed in place by suction on the substrate stage, the substrate temperature during substrate temperature control, the history of exposure heat load generated in the exposure apparatus, and marker image data during alignment measurement, etc. Sample alignment measurement values ​​104 correspond to, for example, in Figure 5 The alignment measurements are shown in the sample injection area.

[0056] Based on these input data, the alignment measurement estimation model 200 estimates the alignment measurements in the injection regions other than the sample injection region, and outputs an estimated value 300 for the alignment measurement in each injection region. The vector representation of the estimated value 300 is as follows: Figure 6 As shown in the figure, the estimated model 200 is a model in which the input / output relationship has been pre-learned through machine learning or similar methods.

[0057] Figure 8 The data flow during the learning process of estimation model 200 is shown. During the learning process, parameters within the estimation model are learned using actual measurements 400 as teacher data; these actual measurements 400 are alignment measurements in the injection region where alignment measurements are actually performed. The format of the actual measurements 400 used as teacher data is the same as the format of the estimated values ​​300—that is, as shown below. Figure 6 The format of the measurements shown is similar.

[0058] By using the estimation model 200 learned as described above, alignment measurement values ​​are estimated in the injection region (the injection region other than the sample injection region) where alignment measurements are not performed. Therefore, correction of higher-order deformation components of the substrate is achieved, and high-precision alignment can be performed.

[0059] Next, a method for determining the sample injection region according to an embodiment will be described. Controller C calculates the estimation error of the estimated value 300 for each injection region based on multiple substrate data used for learning the estimation model 200. The estimation error is the difference between the actual measured value and the estimated value. Then, controller C calculates the inter-data distribution (estimation error distribution) of the estimation error in each injection region. This distribution represents the distribution of the probability of the estimation error occurring and represents the uncertainty of the estimated value in the injection region. For example, when using the variance of the estimation error distribution, it can be as follows... Figure 9The variance is represented as shown. The horizontal diameter of the ellipse indicates the variance of the estimated error distribution of the alignment offset dx in the x-direction, and the vertical diameter of the ellipse indicates the variance of the estimated error distribution of the alignment offset dy in the y-direction. In this example, it can be seen that the variance of the estimated error distribution tends to be larger and the estimation uncertainty is higher closer to the periphery of the substrate. To increase the estimation accuracy of the estimation model, it is advantageous to select the injection region where the variance of the estimated error distribution (estimation uncertainty) is high as the sample injection region for actually performing alignment measurements. Therefore, in this embodiment, the injection region where the variance of the estimated error distribution exceeds a predetermined threshold is selected as the sample injection region.

[0060] Figure 10 An example of sample injection regions selected because the variance of the estimated error distribution exceeds a predetermined threshold is shown. Here, if two injection regions with high variances of the estimated error distribution (estimated uncertainty) are adjacent to each other, then these two injection regions are selected as sample injection regions. In the following description, a method for achieving the desired effect by selecting a smaller number of sample injection regions in the situation described above will be described.

[0061] Here, as an example, an example is shown where the estimation model is a multinomial regression model. When learning the estimation model from teacher data, this multinomial regression model learns the regression coefficients of the polynomial. For example, for a third-order polynomial, the multinomial regression model learns the individual regression coefficients k1 to k2 of equation (1) described above. 20 Note that the regression coefficients k1 to k6 are first-order linear components and can be corrected using actual measurements during the actual alignment measurements of the sample injection area. Therefore, in practice, the multinomial regression model learns the remaining regression coefficients k7 to k6. 20 At this point, controller C calculates the distribution of regression coefficients based on the regression coefficients in each teacher's data. Then, based on the distribution of regression coefficients, controller C calculates the distribution of estimated alignment measurements in each injection region on the substrate. This determines the distribution of estimated alignment measurements used as the output value of the multinomial regression model. Controller C uses this to change the sample injection region. The following will refer to... Figure 11 The flowchart in the document describes the process of determining the sample injection region.

[0062] In step S200, the controller C determines the minimum number of sample injection regions. For example, in the case of a third-order polynomial, there are 7 second-order and higher-order coefficients for each of x and y. More specifically, the second-order and higher-order coefficients for x are k7, k9, k... 11 k 13 k 15 k 17 and k19 The second- and higher-order coefficients used for y are k8, k... 10 k 12 k 14 k 16 k 18 and k 20 Therefore, in this case, the minimum number of sample injection regions is set to 7.

[0063] In step S201, controller C sets the initial arrangement of the set of sample injection regions. The positions of the currently used sample injection regions can be used as is, or unless the arrangement becomes unbalanced, injection regions can be selected sequentially starting from the injection region with the highest variance of the estimated alignment measurements calculated based on the teacher data. Figure 12 An example of the variance of the estimated values ​​of the alignment measurements is shown when seven sample injection regions are selected.

[0064] In step S202, controller C verifies whether there are any sample regions, other than the sample injection regions in the initial arrangement, where the variance of the estimated values ​​of the alignment measurements obtained using the estimation model exceeds a predetermined threshold. If there are no injection regions where the variance exceeds the threshold, the determination process terminates. If there are injection regions where the variance exceeds the threshold, then the injection regions where the variance exceeds the threshold are added to the set of sample injection regions. In an embodiment, the process described below is performed according to step S203.

[0065] In step S203, among the injection regions extracted as injection regions each having a variance exceeding a predetermined threshold, the injection region with the highest variance is added to the set of sample injection regions. Then, in step S204, the controller C obtains the probability distribution of the regression coefficients acquired during the learning period as prior distribution information. In step S205, assuming that the measurement in the currently selected sample injection region is the maximum likelihood of the estimated value of the aligned measurement, the controller C calculates the posterior distribution of the probability distribution of the regression coefficients (first calculation step). For example, Bayesian inference can be used to calculate the posterior distribution. A detailed description of Bayesian inference will be omitted here. To calculate the posterior distribution in the case where additional observation data is obtained for the prior distribution, the following equation (2) can be used:

[0066] p(θ|D)=p(D|θ)p(θ) / p(D)...(2)

[0067] Where θ is the model parameter, D is the observed data, p(θ|D) is the posterior distribution of the model parameter θ with the additional observed data, p(D|θ) is the likelihood function representing the probability of the observed data D occurring based on the model parameter θ, p(θ) is the prior distribution of the model parameter θ, and p(D) is the marginal likelihood of the observed data D.

[0068] Note that if the marginal likelihood p(D) cannot be calculated analytically, then methods such as MCMC (Markov Chain Monte Carlo) can be used as a sample method to calculate the posterior distribution p(θ|D). Alternatively, approximate inference, such as variational inference, can be used to calculate the posterior distribution p(θ|D), in which approximate probability distributions are used for analytical computation.

[0069] In step S206, based on the posterior distribution of the finally calculated regression coefficients, controller C calculates the variance of the estimated values ​​of alignment measurements (hereinafter also referred to as "alignment estimates") in each injection zone (second calculation step). Figure 13 An example of variance at this point is shown. Figure 13 In the process, sample injection region A is added to... Figure 12 The initial arrangement is shown in the figure. This can reduce the variance (uncertainty) in the vicinity of the sample injection region A.

[0070] The process then returns to step S202. In step S202, the controller C re-verifies whether there are any injection regions, other than the currently selected sample injection region, where the variance of the estimated value of the alignment measurement exceeds a predetermined threshold. Therefore, steps S202 to S206 are repeated until there are no injection regions with variance exceeding the predetermined threshold. If it is confirmed in step S202 that there are no injection regions with variance exceeding the predetermined threshold, then the process terminates. Thus, a model capable of performing estimation with the desired accuracy can be obtained.

[0071] Note that if the model includes input data other than the sample alignment measurement 104 (overlay measurement 101, device manufacturing process parameters 102, or exposure device sensor data 103), then each data point can be treated as the maximum likelihood value of the learning data. Alternatively, actual data can be used at regular intervals when the input data is obtained during the actual estimation process.

[0072] Note that although the alignment marks in the sample injection region are described above as being arranged at a single point within the sample injection region, the invention is not limited thereto. Multiple alignment marks may be included in the sample injection region, and a model can be used that also represents the shape of the sample injection region by measuring multiple alignment marks. Furthermore, although the measurements regarding the sample injection region have been described as alignment measurements representing the position (offset) of the mark in each of the x and y directions parallel to the substrate surface, the invention is not limited thereto. The offsets in the x and y directions can be replaced, for example, with focal point measurements representing the offset in the z direction, which is perpendicular to the substrate surface, in each injection region.

[0073] The estimation model can be formed, for example, by a neural network. Here, a neural network is a model with a multi-layered network structure, including input layers, intermediate layers, output layers, etc. During the learning of the estimation model, algorithms such as error backpropagation are used to optimize the coupling weighting coefficients in the neural network by using learning data that indicates the relationship between the input data and the teacher data. The error backpropagation method is a method of adjusting the coupling weighting coefficients between the nodes of the neural network to reduce the error between the output data and the teacher data.

[0074] Information about the set of sample injection regions determined by the method described above can be communicated to the user, for example, by displaying information on a display unit.

[0075] <Second Embodiment>

[0076] In a second embodiment, a virtual measurement system for estimating coverage measurements, which are the relative positions between coverage marks in the upper and lower layers of a substrate, will be described below.

[0077] Figure 14 An example configuration of a virtual measurement system for estimating coverage measurements according to a second embodiment is shown. In this system, the following data is used as input data for the coverage measurement estimation model 210:

[0078] • Alignment estimate 111;

[0079] • Device manufacturing process parameters 112;

[0080] • Exposure device sensor data 113; and

[0081] • Coverage check parameter 114.

[0082] Alignment estimate 111 can be estimate 300, which is the output data of estimation model 200 according to the first embodiment. Alternatively, with respect to alignment estimate 111, the input data of estimation model 200 according to the first embodiment, which serves as the basis for estimate 300, can be used as the input to estimation model 210. Device manufacturing process parameters 112 may include, for example, the ID of the developing apparatus that processes the substrate that has already been exposed in the exposure apparatus, measurement parameters, pattern information of the original plate to be used, etc. Exposure apparatus sensor data 113 may include, for example, the control deviation between the substrate stage and the original plate stage, the internal temperature of the apparatus during temperature control, the measurement value of the accelerometer inside the apparatus, etc. Coverage inspection parameters 114 may include, for example, the measurement signals of the coverage inspection apparatus, apparatus logs, etc.

[0083] Based on the input data as described above, the coverage measurement estimation model 210 estimates the coverage measurement in each injection region and outputs an estimated value 310 of the coverage measurement in each injection region. The estimation model 210 is a model in which the input / output relationship has been pre-learned through machine learning or the like.

[0084] Figure 15 The data flow during the learning of the estimation model 210 is shown. In the second embodiment, the injection region where the coverage measurement is performed is referred to as the sample injection region. During the learning period, parameters within the estimation model are learned using actual measurement values ​​410 as teacher data; the actual measurement values ​​410 are the coverage measurement values ​​in the injection region where the coverage measurement is actually performed.

[0085] By using the estimation model 210 learned as described above, coverage measurement values ​​are estimated in the injection areas where coverage measurement was not performed (injection areas other than the sample injection area). This reduces the frequency of coverage checks and improves productivity.

[0086] Note that since the method is implemented simply by replacing the alignment measurements in the first embodiment with coverage measurements, a detailed description of the method for determining the sample injection region will be omitted here. Also in this embodiment, the sample injection region can be optimized through a process similar to that in the first embodiment.

[0087] <Third Embodiment>

[0088] In the third embodiment, the exposure device 1 has the function of notifying the user when the performance of the estimation model is low. This function can be applied to either the alignment measurement estimation model according to the first embodiment or the coverage measurement estimation model according to the second embodiment. A description will be given below using the alignment measurement estimation model according to the first embodiment as an example.

[0089] refer to Figure 16 The flowchart below will describe the method for determining the sample injection region according to this embodiment. For example, in Figure 11 In the flowchart, the same step numbers indicate the same processing step, and their descriptions will be omitted. Figure 16 In the flowchart, step S207 is added between steps S202 and S203. In step S207, controller C calculates the average value of the estimation error for each injection region and determines whether there are injection regions where the average value exceeds a predetermined value. The estimation error is the error between the estimated value and the measured value obtained through actual measurement. If there are injection regions where the average estimation error exceeds the predetermined value, it means that an error has occurred in the estimation of the alignment measurement value on the substrate surface that cannot be represented by the currently used estimation model. Therefore, it is determined that the model's expressiveness is insufficient. In this case, in step S208, controller C notifies the user that the model's expressiveness is insufficient. For example, the notification is given by displaying a message indicating that the model's expressiveness is insufficient on the display unit. The notification can also be given by sound or the like. In this way, the user can know the timing when the estimation model needs to be modified.

[0090] <Fourth Embodiment>

[0091] In the fourth embodiment, the exposure apparatus has the function of detecting anomalies in the measured values. The range within which the alignment measured values ​​normally fall in each injection region can be understood empirically based on the distribution of the alignment measured value estimation model. Therefore, the controller C determines whether the measured value obtained when performing an alignment measurement (step S304) in each sample injection region falls within a predetermined range set based on the variance of the distribution of the estimated alignment measured values ​​calculated according to the estimation model. If the measured value falls outside this range, then the controller C determines that the measured value is an anomaly.

[0092] If an anomaly is detected in the measurement, controller C can notify the user. This notification can be given, for example, by displaying a message indicating the anomaly on the display unit. Notification can also be given via sound. In this way, the user can determine when it is necessary to modify the estimation model.

[0093] <Fifth Embodiment>

[0094] In the fifth embodiment, the probability distribution of regression coefficients at the location of each injection region on the substrate surface is displayed as a graph on the GUI. This probability distribution is obtained using the alignment measurement estimation model according to the first embodiment or the coverage measurement estimation model according to the second embodiment. The representation of this graph when displayed can be, for example, as shown below. Figure 9 The diagram illustrates this. Users can control the timing of the execution steps to improve performance, focusing on injection areas with high uncertainty.

[0095] <Example of Article Manufacturing Method>

[0096] The article manufacturing method according to an embodiment of the present invention is applicable to the manufacture of articles such as microdevices, for example semiconductor devices or components having microstructures. The article manufacturing method according to this embodiment includes the steps of transferring a pattern of a master template onto a substrate using a photolithography apparatus (exposure apparatus, imprinting apparatus, drawing apparatus, etc.), and processing the substrate onto which the pattern was transferred in the former step. The manufacturing method also includes other known steps (oxidation, thin film formation, deposition, doping, planarization, etching, resist removal, dicing, bonding, encapsulation, etc.). The article manufacturing method of this embodiment is more advantageous than conventional methods in at least one of the following aspects: article performance, quality, productivity, and production cost.

[0097] <Other Embodiments>

[0098] One or more embodiments of the present invention can also be implemented by a computer that reads and executes computer-executable instructions (e.g., one or more programs) recorded on a storage medium (which may also be more fully referred to as a "non-transient computer-readable storage medium") to perform the functions of one or more embodiments described above and / or includes one or more circuits (e.g., application-specific integrated circuits (ASICs)) for performing the functions of one or more embodiments described above, and by a method executed by a computer of a system or device, for example, by reading and executing computer-executable instructions from a storage medium to perform the functions of one or more embodiments described above and / or controlling one or more circuits to perform the functions of one or more embodiments described above. The computer may include one or more processors (e.g., a central processing unit (CPU), a microprocessor unit (MPU)) and may include a network of individual computers or individual processors to read and execute computer-executable instructions. The computer-executable instructions may be provided to the computer, for example, from a network or a storage medium. The storage medium may include, for example, a hard disk, random access memory (RAM), read-only memory (ROM), a storage device for a distributed computing system, an optical disc (such as a CD, DVD, or Blu-ray disc), or a digital versatile disc (BD). TM ), flash memory devices, memory cards, etc.

[0099] Other embodiments

[0100] The embodiments of the present invention can also be implemented by providing software (programs) that perform the functions of the above embodiments to a system or device via a network or various storage media, and the computer or central processing unit (CPU) or microprocessor unit (MPU) of the system or device reads out and executes the program.

[0101] While the invention has been described with reference to exemplary embodiments, it should be understood that the invention is not limited to the disclosed exemplary embodiments. The scope of the following claims will be given the broadest interpretation to include all such modifications and equivalent structures and functions.

Claims

1. A method for determining a set of sample injection regions from a plurality of injection regions of a substrate, wherein the marked positions in each sample injection region are to be actually measured, the method comprising: Set the initial arrangement of the sample injection regions; as well as The following injection region is added to the set of sample injection regions in addition to the sample injection regions in the initial arrangement, where the value of the uncertainty of the estimated value of the mark position estimated by using the estimation model exceeds a predetermined threshold, and the value of the uncertainty of the estimated value indicates the probability of the error in the estimated mark position.

2. The method of claim 1, further comprising, for a plurality of injection regions of the substrate, calculating the error based on actual measured values ​​obtained by actually measuring the positions of the marks and estimated values ​​of the positions of the marks.

3. The method according to claim 2, wherein, The calculation also includes calculating the probability of the error occurring in each of the plurality of injection regions based on the error.

4. The method according to claim 1, wherein, The estimation model receives input data, including overlay measurements between the upper and lower layers, device fabrication process parameters, sensor parameters used in the photolithography step, and measurements obtained by actually measuring the position of the marker in the sample injection region. It outputs an estimate of the measured position of the marker in the injection region other than the sample injection region.

5. The method according to claim 4, wherein, The estimation model is one in which the input / output relationship is learned when measurements obtained by actually measuring the position of the markers in the sample injection region are used as teacher data.

6. The method according to claim 5, wherein, The estimation model is a multinomial regression model that learns the regression coefficients of a multinomial based on teacher data.

7. The method according to claim 6, wherein, In the addition process, injection regions whose uncertainty values ​​exceed a threshold are extracted, and the injection region with the largest uncertainty value among the extracted injection regions is added to the set of sample injection regions.

8. The method of claim 7, further comprising: After adding new sample injection regions, calculate the posterior distribution of the probability distribution of the regression coefficients; as well as Based on the calculated posterior distribution, the uncertainty of the estimated value of the measurement indicating the location of the mark in each injection region is calculated, and The process of adding, calculating the posterior distribution, and calculating the values ​​is repeated until there are no more injection regions whose calculated uncertainty values ​​exceed a threshold.

9. The method according to claim 1, wherein, The mark is located in a direction parallel to the surface of the substrate.

10. The method according to claim 1, wherein, The mark is located in a direction perpendicular to the surface of the substrate.

11. The method according to claim 1, wherein, The position of the marker is the relative position between the overriding marker in the upper layer and the overriding marker in the lower layer.

12. The method according to claim 1, further comprising: Inform the user of the set of determined sample injection regions.

13. The method of claim 1, further comprising: Calculate the average of the estimation error, which is the error of the estimated value relative to the measured value obtained through actual measurement; as well as Notify the user if there are any injection zones where the calculated average value exceeds the predetermined value.

14. The method according to claim 1, further comprising: Inform the user whether the measurement obtained by actually measuring the position of the mark in the defined sample injection area falls outside the predetermined range set based on the value indicating uncertainty.

15. The method according to claim 1, wherein, The value indicating the uncertainty of the estimate includes the variance of the estimate.

16. The method of claim 6, further comprising: The display unit shows the probability distribution information of the regression coefficients.

17. A method for obtaining a measurement of the position of a mark in each of a plurality of ejection regions of a substrate, comprising: The location of the marker in each sample injection region included in the set of sample injection regions determined using the method according to claim 1 is measured, and By using an estimation model and based on measurements of the marks in each of the sample injection regions included in the set of sample injection regions, the measured locations of the marks in each of the injection regions not included in the set of sample injection regions are estimated.

18. An information processing apparatus for determining a set of sample injection regions from a plurality of injection regions of a substrate, wherein the position marked in each sample injection region is to be actually measured, the information processing apparatus comprising: Processing unit The processing unit sets the initial arrangement of the sample injection regions, and The following injection region is added to the set of sample injection regions in addition to the sample injection regions in the initial arrangement, where the value of the uncertainty of the estimated value of the mark position estimated by using the estimation model exceeds a predetermined threshold, and the value of the uncertainty of the estimated value indicates the probability of the error in the estimated mark position.

19. A photolithography apparatus, comprising: A substrate stage configured to position the substrate based on measurements of the position of a marker obtained using the method of claim 17.

20. A computer-readable storage medium storing a program for causing a computer to perform the steps of the method according to claim 1.

21. A method for manufacturing an article, comprising: Patterns are formed on a substrate by using the photolithography apparatus according to claim 19; as well as The substrate on which the pattern is formed is processed. Among them, articles are manufactured from processed substrates.

Images