Method and system for targeted monitoring of semiconductor metrology quality

Abstract

A method and system for monitoring the quality of semiconductor measurements in a targeted manner is presented. Rather than relying on one or more general indicators to determine the overall measurement quality, one or more target measurement quality indicators are determined. Each target measurement quality indicator provides insight into whether a specific operational issue is adversely affecting the measurement quality. In this manner, the one or more target measurement quality indicators not only highlight measurement defects, but also provide insight into the specific operational issues causing the measurement defects. In some embodiments, the values ​​of the one or more target measurement quality indicators are determined based on features extracted from the measurement data. In some embodiments, the values ​​of the one or more target measurement quality indicators are determined based on features extracted from one or more indicators of comparison between the measurement data and corresponding measurement data simulated by a trained measurement model.

Description

[Technical Field]

[0001] The described embodiments relate to measurement systems and methods, and more particularly to methods and systems for improving measurements of semiconductor structures. [Background technology]

[0002] Semiconductor devices, such as logic and memory devices, are typically fabricated by a series of processing steps applied to a specimen. The various features and structural levels of a semiconductor device are formed by these processing steps. For example, lithography, among other processes, is one semiconductor manufacturing process that involves creating patterns on a semiconductor wafer. Further examples of semiconductor manufacturing processes include, but are not limited to, chemical-mechanical polishing, etching, deposition, and ion implantation. Multiple semiconductor devices can be fabricated on a single semiconductor wafer and then separated into individual semiconductor devices.

[0003] Metrology processes are used at various steps during the semiconductor manufacturing process to detect defects on wafers, promoting higher yields. Optical and X-ray-based metrology techniques offer the potential for high throughput without the risk of specimen destruction. Several metrology-based techniques, including implementations of scatterometry, reflectometry, and ellipsometry, and associated analysis algorithms are commonly used to characterize critical dimensions, film thickness, composition, overlay, and other parameters of nanoscale structures.

[0004] Many metrology techniques are indirect methods of measuring the physical properties of a sample under test. In most cases, the raw measurement signal cannot be used to directly determine the physical properties of the sample. Instead, a measurement model is used to estimate the values ​​of one or more parameters of interest based on the raw measurement signal. For example, ellipsometry is an indirect method of measuring the physical properties of a sample under test. Generally, a physics-based or machine learning-based measurement model is used to estimate the values ​​of one or more parameters of interest based on the raw measurement signal (e.g., α meas and β meas) is needed to determine the physical properties of a sample based on the

[0005] In some examples, a physics-based measurement model is created that estimates the magnitude of the raw measurement signal (e.g., α) based on assumed values ​​of one or more model parameters. meas and β meas As shown in equations (1) and (2), the measurement model attempts to predict the parameters associated with the metrology tool itself, such as the system parameters (P system ) and parameters associated with the sample being measured. When determining the parameters of interest, some sample parameters are treated as fixed values ​​(P spec-fixed ), and other sample parameters of interest are floated (P spec-float ), i.e., solved based on the raw measurement signals. α model =f(P system ,P spec-fixed ,P spec-float )(1) β model =g(P system ,P spec-fixed ,P spec-float )(2)

[0006] System parameters are parameters used to characterize a metrology tool (e.g., the ellipsometer 101). Exemplary system parameters include the angle of incidence (AOI), analyzer angle (AO), polarization angle (PO), illumination wavelength, numerical aperture (NA), compensators or waveplates (if present), etc. Sample parameters are parameters used to characterize a sample (e.g., material and geometric parameters that characterize the structure(s) under measurement). For thin-film samples, exemplary sample parameters include the refractive index, dielectric function tensor, nominal layer thicknesses of all layers, layer order, etc. For CD samples, exemplary sample parameters include geometric parameter values ​​associated with different layers, refractive indices associated with different layers, etc. For measurement purposes, the system parameters and many sample parameters are treated as known, fixed-value parameters. However, the values ​​of one or more sample parameters are treated as unknown, floating-object parameters.

[0007] In some examples, the values ​​of the floating parameters of interest are solved by an iterative process (e.g., regression) that produces the best fit between theoretical predictions and experimental data. The values ​​of the floating parameters of interest that are unknown are varied, and the model output values ​​(e.g., α model and β model ) is a set of sample parameter values ​​that are compared between model output values ​​and experimentally measured values ​​(e.g., α meas and β meas ) is determined to provide a sufficient match between the floating parameters. In some other instances, the floating parameters are solved by searching a library of pre-computed solutions to find the closest match.

[0008] In some other examples, a trained machine learning-based measurement model is used to directly estimate values ​​of the parameters of interest based on raw measurement data. In these examples, the machine learning-based measurement model receives raw measurement signals as model inputs and produces values ​​of the parameters of interest as model outputs.

[0009] Both physics-based and machine learning-based measurement models need to be trained to generate useful estimates of parameters of interest for specific measurement applications. Typically, model training is based on raw measurement signals collected from samples with known values ​​of the parameters of interest (i.e., design of experiments (DOE) data).

[0010] Machine learning-based measurement models are parameterized by a number of weight parameters. Traditionally, machine learning-based measurement models are trained by a regression process (such as ordinary least squares regression). The values ​​of the weight parameters are iteratively adjusted to minimize the difference between a known reference value of the parameter of interest and the value of the parameter of interest estimated by the machine learning-based measurement model based on the measured raw measurement signals.

[0011] As previously mentioned, physics-based measurement models are parameterized by a number of system and sample parameters. Traditionally, physics-based measurement models are also trained by a regression process (such as ordinary least squares regression). One or more system and sample parameters are iteratively adjusted to minimize the difference between the raw measurement data and the modeled measurement data. In each iteration, the value of a particular sample parameter of interest is maintained at a known DOE value.

[0012] Both trained machine learning-based measurement models and physics-based models estimate the values ​​of parameters of interest based on measurement data. The accuracy of the estimation depends on the measurement data. For example, measurement data outside the domain of the measurement data observed during the model's training phase can cause large prediction errors due to extrapolation.

[0013] Traditionally, several approaches have been considered to quantify the measurement uncertainty associated with each measurement site. Several standard quality metrics exist. Some exemplary measurement quality metrics are not specific to a particular measurement parameter. In these examples, the quality of the measurement is determined based on the quality of the fit. Typical quantitative metrics used to characterize the quality of the fit include chi-square, goodness of fit, etc.

[0014] In these examples, the simulated or expected signal (S) is compared with the actual measured signal (M) by an appropriate norm. In some examples, the difference is weighted by an uncertainty threshold (w). For more complex systems, a mathematical transformation of the simulated and measured signals is performed (e.g., converting the photon signal into an independent vector of principal components) before the difference is weighted. In these cases, the fitting quality can be expressed by equation (3), where N is the number of signals.

number

[0015] In general, equation (3) can be modified by any normalization, or the quality can be expressed as a function based on equation (3).

[0016] Other measurement quality metrics are specific to the uncertainty of one or more measurement parameters, e.g., a projected model. In one example, a noise model is multiplied by a sensitivity function to obtain an estimate of the measurement uncertainty associated with a particular parameter or group of parameters.

[0017] Different measurement applications have different signal volumes (N), different thresholds for each use case, and different issues that specifically affect quality. Unfortunately, traditional approaches to assessing measurement quality have difficulty capturing the specific issues that affect measurement quality, especially across a wide range of measurement applications. [Prior art documents] [Patent documents]

[0018] [Patent Document 1] U.S. Patent Application Publication No. 2020 / 0200525 Summary of the Invention [Problem to be solved by the invention]

[0019] Future metrology applications will present metrology challenges due to ever-smaller resolution requirements, multi-parameter correlations, increasingly complex geometries, and the increasing use of opaque materials. Therefore, methods and systems for improving the assessment of measurement quality are desired. [Means for solving the problem]

[0020] Presented herein are methods and systems for monitoring semiconductor measurement quality in a targeted manner. Rather than relying on one or more general indicators to determine overall measurement quality, one or more target measurement quality indicators are determined. Each target measurement quality indicator provides insight into whether a specific operational issue is adversely affecting measurement quality. In this manner, the one or more target measurement quality indicators not only highlight measurement deficiencies but also provide insight into the specific operational issues causing the measurement deficiencies.

[0021] In some embodiments, a value of a target measurement quality indicator is determined for each measurement sample during measurement inference. In some of these embodiments, semiconductor measurements involve inferring values ​​of one or more parameters of interest from measurement data using a trained measurement model. If the estimated value of the target measurement quality indicator indicates that the measurement quality is insufficient, the measurement is flagged. In some examples, repeated insufficient measurement quality indicates that the measurement model is out of date and requires retraining.

[0022] In some embodiments, the trained quality monitor determines values ​​of one or more target measurement quality indicators based on the measurement dataset and the corresponding simulated measurement signal. In particular, the trained quality monitor classifies the quality of the measurement based on features extracted from one or more indicators of comparison between the measured data and the simulated data. The indicators of comparison between the measured data and the simulated data are quantitative results of any suitable analysis used to extract differences or similarities between the measured data and the simulated data, such as pixel-by-pixel differences, relative differences, correlations, etc.

[0023] In some embodiments, the trained quality monitor determines one or more target parameters based on the difference between the measured data and the simulated data, and the one or more target parameters are compared to reference values ​​for the one or more target parameters to determine the status of each corresponding target measurement quality indicator.

[0024] In some embodiments, the trained quality monitor module determines values ​​of one or more target measurement quality indicators based on differences between values ​​of one or more parameters of interest individually determined based on the measurement dataset and the simulated measurement dataset.

[0025] In some embodiments, the trained quality monitor determines values ​​of one or more target measurement quality indicators based on a difference between values ​​of one or more target parameters individually determined based on a measurement dataset by the trained monitor model and the trained measurement model.

[0026] In some embodiments, the trained quality monitor determines values ​​of one or more target measurement quality indicators based on the measurement dataset. In particular, the trained quality monitor classifies the quality of the measurement based on features extracted from the measurement data. In one example, if an initial release of a measurement application is deemed healthy, the similarity of the measurement signal over time is analyzed to determine whether unhealthy changes have occurred.

[0027] The foregoing is a summary and, as such, necessarily contains simplifications, generalizations, and omissions of detail; accordingly, those skilled in the art will appreciate that this summary is merely illustrative and is not intended to be limiting in any way. Other aspects, unique features, and advantages of the devices and / or processes described herein will become apparent in the non-limiting detailed description set forth herein. [Brief explanation of the drawings]

[0028] [Figure 1] 1 illustrates a semiconductor metrology system 100 for monitoring the measurement quality of measurements performed on a wafer in one embodiment according to exemplary methods presented herein. [Figure 2] FIG. 1 illustrates an exemplary measurement quality engine 160 in one embodiment. [Figure 3] FIG. 16 illustrates an exemplary trained quality monitor module 167A in one embodiment. [Figure 4] FIG. 16 illustrates an exemplary trained quality monitor module 167B in one embodiment. [Figure 5] FIG. 16 illustrates an exemplary trained quality monitor module 167C in one embodiment. [Figure 6] FIG. 16 illustrates an exemplary trained quality monitor module 167D in one embodiment. [Figure 7] FIG. 16 illustrates an exemplary trained quality monitor module 167E in one embodiment. [Figure 8] FIG. 3 illustrates a semiconductor metrology system 300 for monitoring measurement quality of measurements performed on a wafer in another embodiment according to exemplary methods presented herein. [Figure 9] 4 is a flowchart of a method 400 for estimating values ​​of one or more target measurement quality indicators indicative of measurement quality associated with operational components of a semiconductor measurement system. DETAILED DESCRIPTION OF THE INVENTION

[0029] Reference will now be made in detail to background examples and certain embodiments of the invention, examples of which are illustrated in the accompanying drawings.

[0030] Presented herein are methods and systems for monitoring semiconductor measurement quality in a targeted manner. Rather than relying on one or more general indicators to determine overall measurement quality, one or more target measurement quality indicators are determined. Each target measurement quality indicator provides insight into whether a specific operational issue is adversely affecting measurement quality. In this manner, the one or more target measurement quality indicators not only highlight measurement deficiencies but also provide insight into the specific operational issues causing the measurement deficiencies.

[0031] In general, the quality of semiconductor measurements depends on the health of several operational components. By way of non-limiting example, operational components include the metrology system, the measurement model, and the underlying manufacturing process. Each of these components can drift slowly or rapidly, causing measurement quality degradation. In some embodiments, the metrology system includes multiple target measurement quality indicators, each focused on a different operational component. In this manner, the cause of overall measurement degradation, e.g., an overall increase in measurement uncertainty, can be narrowed down to one or more specific operational components of the metrology system.

[0032] Target measurement quality indicators may be used, among other things, to characterize uncertainties associated with measurement parameters of interest, defect classification, etc. However, in general, target measurement quality indicators may be used for a variety of other purposes. In some examples, target measurement quality indicators are used as filters to detect measurement outliers. In one example, the value of a target measurement quality indicator associated with a particular measurement is compared to a predetermined threshold. If the target measurement quality indicator value exceeds the predetermined threshold, the measurement is considered an outlier. In some examples, target measurement quality indicators are used as triggers to adjust a semiconductor process. For example, if the average value of multiple target measurement quality indicators, each associated with different instances of the same measurement, exceeds a predetermined threshold, the process is adjusted to bring the average value of the target measurement quality indicator back into a desired range.

[0033] Similarly, in some examples, the target measurement quality index value is used to detect wafers that are not within manufacturing specifications. In some of these examples, the target measurement quality index value is used to detect when the bias of a metrology tool deviates from specifications. In other examples, the target measurement quality index value is used to detect when the measurement error bias deviates from specifications.

[0034] In some embodiments, a value of the target measurement quality indicator is determined for each measurement sample during measurement inference. In some of these embodiments, semiconductor measurements involve inferring values ​​of one or more parameters of interest from measurement data using a trained measurement model. In some of these embodiments, the trained measurement model is physics-based. In some other embodiments, the trained measurement model is a machine learning-based measurement model (e.g., a linear model, a neural network model, a convolutional network model, etc.). If the estimate of the target measurement quality indicator indicates that the measurement quality is insufficient, the measurement is flagged. In some examples, repeated insufficient measurement quality indicates that the measurement model is out of date and requires retraining.

[0035] In some embodiments, the trained measurement model is a defect classification model trained to detect and classify defects from the measurement data. In these embodiments, the trained quality monitor model is used to determine the quality of the defect classification. In some of these embodiments, the measurement data is image data and the trained quality monitor model is used to determine the quality of defect detection from the image data.

[0036] In some embodiments, the trained measurement model is a metrology model trained to estimate values ​​of a geometric parameter of interest (e.g., critical dimension, overlay, etc.), a dispersion parameter of interest (e.g., refractive index, etc.), a process parameter of interest (e.g., lithography focus, dose, etc.), an electrical property of interest (e.g., bandgap, etc.), or any combination thereof. In these embodiments, the trained quality monitor model is used to determine the quality of the estimated parameter of interest.

[0037] Generally, the trained quality monitor model is either a machine learning based model, a physics based model, or a heuristic rule based model.

[0038] In one aspect, the measurement system includes a trained quality monitor model that determines values ​​of one or more target measurement quality indicators, each indicative of a measurement quality associated with a different operational element of the measurement system, and the target measurement quality indicator values ​​are determined based on measurement data collected from an in-line wafer at each measurement site without the use of a reference metrology.

[0039] FIG. 1 illustrates a system 100 for measuring properties of a sample and determining values ​​of one or more target measurement quality indicators associated with each measurement, according to exemplary methods presented herein.

[0040] In at least one novel aspect, an embodiment of a transmission, small-angle X-ray scatterometry (T-SAXS) metrology tool 100 for measuring properties of a sample is shown in Figure 1. As shown in Figure 1, the system 100 can be used to perform T-SAXS measurements on an examination area 102 of a sample 101 illuminated by an illumination beam spot.

[0041] In the illustrated embodiment, metrology tool 100 includes an x-ray illumination source 110 configured to generate x-ray radiation suitable for T-SAXS measurements. In general, any suitable high-brightness x-ray illumination source capable of generating high-brightness x-rays at sufficient flux levels to enable high-throughput in-line metrology may be considered to provide x-ray illumination for T-SAXS measurements. In some embodiments, the x-ray source includes a tunable monochromator, which allows the x-ray source to deliver x-ray radiation at a variety of selectable wavelengths.

[0042] In some embodiments, one or more x-ray sources emitting radiation at photon energies greater than 15 keV are used to ensure that the x-ray source provides light of a wavelength sufficient to penetrate the entire device and wafer substrate. By way of non-limiting example, any of the following may be used as the x-ray illumination source 110: a particle accelerator source, a liquid anode source, a rotating anode source, a stationary solid anode source, a microfocus source, a microfocus rotating anode source, a plasma-based source, and an inverse Compton source. In one example, an inverse Compton source available from Lyncean Technologies, Inc. of Palo Alto, California, USA may be considered. An inverse Compton source has the added advantage of being able to generate x-rays over a range of photon energies, thereby enabling the x-ray source to deliver x-ray radiation at a variety of selectable wavelengths.

[0043] Exemplary x-ray sources include electron beam sources configured to bombard a solid or liquid target to stimulate x-ray emission. A method and system for producing high-brightness liquid metal x-ray illumination is described in U.S. Patent No. 7,929,667, issued April 19, 2011 to KLA-Tencor Corp., the entire contents of which are incorporated herein by reference.

[0044] An x-ray illumination source 110 generates x-ray radiation over a source area having a finite lateral dimension (i.e., a non-zero dimension orthogonal to the beam axis). A focusing optic 111 focuses the source radiation onto a metrology target located on the sample 101. The finite lateral dimension of the source results in a finite-sized spot 102 on the target, the extent of which is defined by rays 117 entering from the edge of the source. In some embodiments, the focusing optic 111 comprises an elliptical focusing optic.

[0045] The beam divergence control slit 112 is located in the beam path between the focusing optical element 111 and the beam shaping slit mechanism 120. The beam divergence control slit 112 limits the divergence of the illumination delivered to the sample during measurement. In some embodiments, an additional intermediate slit 113 is located in the beam path between the beam divergence control slit 112 and the beam shaping slit mechanism 120. The intermediate slit 113 provides further beam shaping. The beam shaping slit mechanism 120 is located in the beam path just before the sample 101. The slit of the beam shaping slit mechanism 120 is placed close to the sample 101 to minimize the expansion of the incident beam spot size due to the beam divergence defined by the finite source size. In one example, the beam spot size expansion due to the shadow caused by the finite source size is approximately 1 μm when the x-ray source size is 10 μm and the distance between the beam shaping slit and the sample 101 is 25 mm. In another example, the beam divergence is controlled by a beam shaping slit placed less than 100 mm from the sample 101 .

[0046] In some embodiments, the beam-shaping slit mechanism 120 includes multiple independently actuated beam-shaping slits. In one embodiment, the beam-shaping slit mechanism 120 includes four independently actuated beam-shaping slits. These four beam-shaping slits effectively block a portion of the incident beam 115 to produce the illumination beam 116 having a box-shaped illumination cross-section.

[0047] Generally, the x-ray optics shape and direct the x-ray radiation toward the sample 101. In some examples, the x-ray optics include an x-ray monochromator to monochromatize the x-ray beam incident on the sample 101. In some examples, the x-ray optics use multi-layer x-ray optics to collimate or focus the x-ray beam onto the measurement area 102 of the sample 101 with a divergence of less than 1 milliradian. In these examples, the multi-layer x-ray optics also function as a beam monochromator. In some embodiments, the x-ray optics include one or more x-ray collimating mirrors, x-ray apertures, x-ray beam stops, refractive x-ray optics, diffractive optics such as zone plates, Montell optics, specular x-ray optics such as grazing incidence elliptical mirrors, polycapillary optics such as hollow capillary x-ray waveguides, multi-layer optics, or systems, or any combination thereof. Further details are provided in U.S. Patent Publication No. 2015 / 0110249, the entire contents of which are incorporated herein by reference.

[0048] The x-ray detector 119 collects the scattered x-ray radiation 114 from the sample 101 and generates an output signal 135 indicative of a property of the sample 101 that is sensitive to the incident x-ray radiation according to the modality of the T-SAXS measurement. In some embodiments, the scattered x-rays 114 are collected by the x-ray detector 119 while the sample positioning system 125 positions and orients the sample 101 to generate the angle-resolved scattered x-rays.

[0049] In some embodiments, the T-SAXS system has a high dynamic range (e.g., 10 5 In some embodiments, a single photon-counting detector detects the location and number of detected photons.

[0050] In some embodiments, a T-SAXS system is used to determine a property (e.g., a structural parameter value) of the sample based on one or more diffraction orders of the scattered light. As shown in Figure 1, the metrology tool 100 includes a computing system 130 that is used to acquire a signal 135 generated by the detector 119 and to determine a property of the sample based at least in part on the acquired signal.

[0051] In some examples, T-SAXS-based metrology involves determining specimen dimensions through the inverse solution of a predetermined measurement model using measurement data. The measurement model includes several (on the order of 10) adjustable parameters that describe the geometry and optical properties of the sample and the optical properties of the measurement system. Inverse solution methods include, but are not limited to, model-based regression, tomographic techniques, machine learning, or any combination thereof. In this method, target profile parameters are estimated by solving for the values ​​of a parameterized measurement model that minimizes the error between the measured scattered x-ray intensity and the modeled results.

[0052] In one embodiment, computing system 130 is configured as a measurement quality engine configured to determine values ​​of one or more target measurement quality indicators based on the amount of measurement data collected by a measurement system, e.g., metrology system 100. Each target measurement quality indicator indicates a measurement quality associated with a different operational element of the measurement system.

[0053] The measurement data may include actual measurement data, simulated measurement data, or both. In some embodiments, the measurement data includes simulated measurement data from a simulation of measurement of one or more measurement targets that are structurally different from the measurement targets employed as part of the training data set.

[0054] In some embodiments, the metrology data includes actual data from measurements of a particular target with a particular metrology tool. Typically, the metrology data includes actual measurements of a target on a nominal wafer. In some embodiments, the metrology data is collected from multiple wafers. In some embodiments, the metrology data is collected from a design of experiments (DOE) wafer. In some embodiments, the metrology data is collected by multiple metrology systems.

[0055] In some embodiments, the features of the measurement dataset are used directly as input data to the trained quality monitor model. However, in other embodiments, the measurement dataset is converted to reduced dimensions, and the features of the reduced measurement dataset are used directly as input data to the trained quality monitor model. In many practical situations, it is preferable to reduce the dimension of the measurement dataset to reduce the computational load. In general, the same data feature dimension reduction techniques used to reduce the training dataset are also used to reduce the measurement dataset.

[0056] In some embodiments, data feature dimensionality reduction is achieved by principal component analysis, in which a measurement data set is decomposed into a set of much smaller principal components. Generally, any suitable data reduction technique may be used, such as Fourier analysis, wavelet analysis, discrete cosine transform analysis, etc.

[0057] Figure 2 shows a diagram of measurement quality engine 160 in one embodiment. As shown in Figure 2, measurement quality engine 160 includes optional data transformation module 161, optional trained measurement model module 163, optional data transformation module 165, and trained quality monitor module 167.

[0058] 2, the measurement dataset 135 is received by a data transformation module 161. In one example, the data transformation module 161 converts the features of the measurement dataset 135 into a set of principal components that are communicated to a trained measurement model module 163 and a trained quality monitor module 167.

[0059] The trained measurement model module 163 determines values ​​of one or more parameters of interest 169 based on the set of principal components 162 using the trained measurement model. The calculated values ​​of the parameters of interest 169 are communicated to the trained quality monitor module 167. Exemplary parameters of interest include geometric parameters, dispersion parameters, process parameters, or electrical parameters that characterize the measured structure.

[0060] Additionally, the trained measurement model module 163 generates a set of simulated measurement signals 164 corresponding to values ​​of one or more parameters of interest 169. The trained measurement model module 163 reconstructs each feature of the measurement dataset (or the reduced measurement dataset) based on the values ​​of the one or more parameters of interest estimated by a measurement model trained on the measurement dataset (or the reduced measurement dataset).

[0061] 2 , the simulated measurement signal 164 is received by a data transformation module 165. In one example, the data transformation module 165 converts features of the measurement dataset 164 into a set of principal components 166 that are transmitted to a trained quality monitor module 167. The trained quality monitor module 167 determines values ​​of one or more target measurement quality indicators 168 based on the measurement dataset 135 (or reduced measurement dataset 162) and, optionally, the simulated measurement signal 164 (or reduced measurement signal 166), values ​​of one or more parameters of interest 169, or both. The values ​​of the one or more target measurement quality indicators 168 are stored in a memory, such as memory 132.

[0062] In some embodiments, the trained quality monitor module 167 determines values ​​of one or more target measurement quality indicators 168 based on the measurement dataset 135 (or reduced measurement dataset 162) and the simulated measurement signal 164 (or reduced measurement signal 166). In particular, the trained quality monitor classifies the quality of the measurement based on features extracted from one or more indicators of comparison between the measured data and the simulated data. The indicators of comparison between the measured data and the simulated data are quantitative results of any suitable analysis used to extract differences or similarities between the measured data and the simulated data, e.g., pixel-by-pixel differences, relative differences, correlations, etc.

[0063] 3 shows a diagram of trained quality monitor module 167A in one embodiment. Trained quality monitor module 167A includes a feature extraction module 171 and a trained classifier module 173. As shown in FIG. 3, trained quality monitor module 167A receives measurement data 135 and simulated measurement data 164 and determines the difference between the two data sets. A difference data set 170 is communicated to feature extraction module 171. Feature extraction module 171 generates a feature set 172 from difference data set 170.

[0064] Feature extraction module 171 is trained to extract any number of different features that describe the quality of the match between the measured data and the simulated data. In some examples, measured data 135 is one or more measured scatterometry images captured by detector 119 shown in FIG. 1 , and simulated data 164 is corresponding scatterometry images simulated by the trained measurement model of module 163. Exemplary features of difference image 170 include image symmetry, correlation, higher order moments, etc.

[0065] In general, the exact combination of features used to characterize the quality of the match between the measurement image and the simulated image will vary depending on the measurement application. To overcome this limitation, feature extraction module 171 extracts a predetermined set of features known to be valuable in assessing the integrity of the measurement. By way of non-limiting example, the predetermined set of features may include correlation, residual, skew, or the like. In addition, feature extraction module 171 determines additional features derived from the variance of the difference data using one or more feature extraction routines, such as principal component analysis (PCA), one or more autoencoders, or the like. In this manner, feature extraction module 171 generates feature set 172 that characterizes the quality of the match between the measurement image and the simulated image across a wide range of measurement applications.

[0066] Additionally, the trained classifier model of module 173 is trained using measurement data sets known to be healthy and unhealthy. The measurement data sets may be real measurements or may be generated synthetically, i.e., by simulation. In this way, the trained classifier model is trained to distinguish between healthy and unhealthy measurements based on feature set 172. Each of the features may represent one or more measurement issues or process values, and classifier 173 determines whether these issues or process values ​​result in a healthy or unhealthy measurement.

[0067] 3, trained classifier module 173 determines the classification of each measurement as healthy or unhealthy based on feature set 172. Target measurement quality indicator 168A generated by trained classifier module 173 is a signal indicating whether the measurement is considered healthy or unhealthy. The results are stored in memory (e.g., memory 132).

[0068] In some examples, the trained quality monitor module 167A operates as a gate for in-line measurements of production wafers. In one example, subsequent inspection of a particular wafer is trusted only if the target measurement quality indicator 168A indicates a healthy measurement. If the target measurement quality indicator 168A indicates an unhealthy measurement, subsequent inspections are deemed unreliable. Furthermore, if the target measurement quality indicator 168A indicates an unhealthy measurement, an alternative metrology technique may be used to measure the wafer. In another example, if the target measurement quality indicator 168A indicates an unhealthy measurement, the measured die is deemed a failure rather than a success. In this manner, the trained quality monitor module 167A operates to determine the yield of the device. In these examples, yield is typically included as part of the training criteria.

[0069] In some embodiments, the trained quality monitor module 167 determines values ​​of one or more target measurement quality indicators 168 based on the measurement data set 135 (or reduced measurement data set 162) and the simulated measurement signal 164 (or reduced measurement signal 166). In particular, the trained quality monitor determines one or more target parameters based on differences between the measurement data and the simulated data. Additionally, the one or more target parameters are compared to reference values ​​for the one or more target parameters to determine the status of each corresponding target measurement quality indicator.

[0070] FIG. 4 shows a diagram of trained quality monitor module 167B in one embodiment. Trained quality monitor module 167B includes trained monitor model module 180. In the embodiment shown in FIG. 4, trained quality monitor module 167B receives reduced measurement data 162 and reduced simulated measurement data 166 and determines the difference between the two data sets. A difference data set 187 is communicated to trained monitor model module 180. Trained monitor model module 180 generates estimates 181 of one or more parameters of interest that characterize the structure being measured from the difference data set 187.

[0071] The trained monitor model is trained to estimate one or more parameters of interest characterizing the structure being measured from the difference between the measurement signal (or a reduced measurement signal) and a corresponding measurement response signal generated by a trained measurement model used to estimate values ​​of one or more parameters of interest characterizing the same structure being measured.

[0072] In some embodiments, the trained monitor model is physics-based. In some other embodiments, the trained monitor model is a machine learning-based model (e.g., a linear model, a neural network model, a convolutional network model, etc.). The monitor model is trained based on measurements of DOE wafers having known values ​​of one or more parameters of interest. In some embodiments, the known values ​​of the parameters of interest may be measured by a trusted reference metrology system. In other embodiments, the known values ​​of the parameters of interest are programmed values ​​of the parameters of interest. In these embodiments, the measurement data used to train the monitor model is generated synthetically, i.e., by simulation.

[0073] As shown in FIG. 4 , trained quality monitor module 167B receives reference values ​​182 for one or more target parameters corresponding to estimated values ​​181. Difference values ​​183 between estimated values ​​181 and reference values ​​182 are compared to standard values ​​186. If the standard is exceeded, signal 184 is transmitted from trained monitor module 167B, indicating that the measurement quality associated with the particular target parameter is unreliable. If difference values ​​183 associated with the particular target parameter are within standard 186, signal 185 is transmitted from trained monitor module 167B, indicating that the measurement quality associated with the particular target parameter is reliable. Thus, target measurement quality indicator 168B generated by trained quality monitor module 167B is one of two signals indicating whether the measurement of each target parameter is considered healthy or unhealthy. The results are stored in memory (e.g., memory 132).

[0074] In some embodiments, the trained quality monitor module 167 determines values ​​of one or more target measurement quality indicators 168 based on differences between values ​​of one or more target parameters determined individually based on the measurement dataset 135 (or reduced measurement dataset 162) and the simulated measurement signal 164 (or reduced measurement signal 166).

[0075] FIG. 5 shows a diagram of trained quality monitor module 167C in one embodiment. Trained quality monitor module 167C includes trained monitor model module 190. In the embodiment shown in FIG. 5, trained monitor model module 190 estimates values ​​191 of one or more target parameters based on reduced measurement data set 162. In addition, trained monitor model module 190 estimates values ​​192 of the same one or more target parameters based on reduced simulated data set 166. Trained quality monitor module 167C determines a difference 193 between the estimated values ​​191 and 192 associated with each target parameter.

[0076] 5, each difference value 193 is compared to a standard value 194. If the difference value 193 associated with a particular target parameter exceeds the standard 194, a signal 195 is communicated from the trained monitor module 167C, indicating that the measurement quality associated with the particular target parameter is unreliable. If the difference value 193 associated with the particular target parameter is within the standard 194, a signal 196 is communicated from the trained monitor module 167C, indicating that the measurement quality associated with the particular target parameter is reliable. Thus, the target measurement quality indicator 168C generated by the trained quality monitor module 167C is one of two signals indicating whether the measurement of each target parameter is considered healthy or unhealthy. The results are stored in memory (e.g., memory 132).

[0077] The trained monitor model is trained to estimate one or more parameters of interest characterizing the structure being measured from both the measurement signal (or reduced measurement signal) and the corresponding measurement response signal generated by the trained measurement model used to estimate values ​​of one or more parameters of interest characterizing the same structure being measured.

[0078] In some embodiments, the trained monitor model is physics-based. In some other embodiments, the trained monitor model is a machine learning-based model (e.g., a linear model, a neural network model, a convolutional network model, etc.). The monitor model is trained based on measurements of DOE wafers having known values ​​of one or more parameters of interest. In some embodiments, the known values ​​of the parameters of interest may be measured by a trusted reference metrology system. In other embodiments, the known values ​​of the parameters of interest are programmed values ​​of the parameters of interest. In these embodiments, the measurement data used to train the monitor model is generated synthetically, i.e., by simulation.

[0079] As shown in Figure 5, the trained monitor model estimates a value of a parameter of interest from a measurement signal and estimates another value of the same parameter of interest from the corresponding measurement response signal generated by the trained measurement model. This allows for prediction of possible errors in the trained measurement model. If the trained measurement model performs well, the values ​​predicted by the trained monitor model based on two different data sets will be close. If the values ​​predicted by the trained monitor model based on two different data sets differ, this indicates an error in the prediction associated with a particular parameter of interest by the trained measurement model.

[0080] In one example, the trained monitor model is a machine learning-based model trained to predict values ​​for system parameters, critical dimension (CD) parameters, and tilt parameters. The model is sensitive to all of these parameters, particularly since each of these parameters has a different measurement response. Conversely, the model is insensitive to general noise or systematic errors that affect all image pixels in the same way.

[0081] By training monitor models on a variety of target parameters, it is possible to determine indicators of measurement quality related to many different operational aspects of the overall measurement. For example, if one or more system parameters deviate from specifications, this indicates that the metrology tool itself needs to be inspected and repaired.

[0082] In another example, the target measurement quality indicator 168C is determined relative to a known reference wafer, such as a monitor wafer, in which case only the system may be at fault.

[0083] In another example, a user may select which target measurement quality indicators 168C to consider or ignore based on the measurement application. For example, if a particular measurement recipe is designed to estimate the tilt of a structure, CD-related target measurement quality indicators are ignored, but tilt-related target measurement quality indicators are considered. If the tilt-related target measurement quality indicators are within specifications, the measurement recipe is accepted, even if the overall measurement uncertainty indicates that the measurement recipe should be discarded.

[0084] In some embodiments, the trained quality monitor module 167 determines values ​​of one or more target measurement quality indicators 168 based on the difference between the values ​​of one or more target parameters determined individually by the trained monitor model and the trained measurement model based on the measurement dataset 135 (or reduced measurement dataset 162).

[0085] FIG. 6 shows a diagram of the trained quality monitor module 167D in one embodiment. The trained quality monitor module 167D includes a trained monitor model module 200. In the embodiment shown in FIG. 6, the trained monitor model module 200 estimates values ​​201 of one or more target parameters based on the reduced measurement data set 162 (or measurement data set 135). In addition, the trained quality monitor module 167D receives values ​​169 of the same one or more target parameters estimated by a trained measurement model, as shown in FIG. 2. The trained measurement model estimates values ​​169 of the same one or more target parameters based on the same reduced measurement data set 162 (or measurement data set 135). The trained quality monitor module 167D determines a difference 202 between the estimated values ​​201 and 169 associated with each target parameter.

[0086] 6, each difference value 202 is compared to a standard value 203. If the difference value 202 associated with a particular target parameter exceeds the standard 203, a signal 204 is communicated from the trained monitor module 167D, indicating that the measurement quality associated with the particular target parameter is unreliable. If the difference value 202 associated with the particular target parameter is within the standard 203, a signal 205 is communicated from the trained monitor module 167D, indicating that the measurement quality associated with the particular target parameter is reliable. Thus, the target measurement quality indicator 168D generated by the trained quality monitor module 167D is one of two signals indicating whether the measurement of each target parameter is considered healthy or unhealthy. The results are stored in memory (e.g., memory 132).

[0087] The trained monitor model is trained to estimate one or more parameters of interest that characterize the structure under measurement from the measurement signal (or a reduced measurement signal). In some embodiments, the monitor model is trained based on measurements of DOE wafers having known values ​​of one or more parameters of interest. In some embodiments, the known values ​​of the parameters of interest may be measured by a trusted reference metrology system. In other embodiments, the known values ​​of the parameters of interest are programmed values ​​of the parameters of interest. In these embodiments, the measurement data used to train the monitor model is generated synthetically, i.e., by simulation.

[0088] In some embodiments, the trained monitor model is physics-based, while in some other embodiments, the trained monitor model is a machine learning-based model (e.g., a linear model, a neural network model, a convolutional network model, etc.).

[0089] In some embodiments, the trained quality monitor module 167 determines values ​​of one or more target measurement quality indicators 168 based on the measurement dataset 135 (or the reduced measurement dataset 162). In particular, the trained quality monitor classifies the quality of the measurement based on features extracted from the measurement data. If the initial release of a measurement application is deemed healthy, the quality of subsequent measurement signals can be used as an indicator of the health of the measurement. In some embodiments, the similarity of the measurement signals over time is analyzed to determine whether unhealthy changes have occurred.

[0090] 7 shows a diagram of trained quality monitor module 167E in one embodiment. Trained quality monitor module 167E includes a feature extraction module 210 and a trained classifier module 212. As shown in FIG. 7, trained quality monitor module 167E receives reduced measurement data 162 (or measurement data 135). Trained feature extraction module 210 generates a feature set 211 from measurement data set 162.

[0091] The feature extraction module 210 is trained to extract any number of different features that describe the quality of the measurement data. In some examples, the measurement data 135 is one or more measured scatterometry images captured by the detector 119 shown in Figure 1. Exemplary features of the images 135 include image symmetry, correlation, higher order moments, etc.

[0092] In general, the exact combination of features used to characterize the quality of a measurement image varies depending on the measurement application. To overcome this limitation, feature extraction module 210 extracts a predetermined set of features known to be valuable in assessing the integrity of a measurement. In addition, feature extraction module 210 determines additional features derived from the variance of the measurement data using one or more feature extraction routines, such as principal component analysis (PCA), one or more autoencoders, etc. In this way, feature extraction module 210 generates feature set 211 that characterizes the quality of measurement images across a wide range of measurement applications.

[0093] In some embodiments, the trained classifier model of module 212 is trained using a training measurement dataset that is known to be healthy. In these embodiments, the classifier model is trained to identify healthy measurements as subsequent measurements that have features that fall within predetermined boundaries of the features of the healthy training measurement dataset. The training measurement dataset may be real measurements or may be generated synthetically, i.e., by simulation.

[0094] In some embodiments, the trained classifier model of module 212 is trained using a training measurement dataset known to be healthy and a training measurement dataset known to be unhealthy. In these embodiments, the classifier model is trained to identify healthy measurements as measurements that have features similar to corresponding features in the healthy training measurement dataset. Similarly, the classifier model is trained to identify unhealthy measurements as measurements that have features similar to corresponding features in the unhealthy training measurement dataset.

[0095] 7, the trained classifier module 212 determines the classification of each measurement as healthy or unhealthy based on the feature set 211. Each of the features can represent one or more measurement issues or process values, and the classifier 212 determines whether these issues or process values ​​result in a healthy or unhealthy measurement. The target measurement quality indicator 168E generated by the trained classifier module 212 is a signal indicating whether the measurement is considered healthy or unhealthy according to each feature. The results are stored in a memory (e.g., memory 132).

[0096] In this way, determining the value of one or more target quality indicators involves a comparison between one or more features extracted from the measurement data and a corresponding set of features extracted from training measurement data measured prior to the measurement data.

[0097] The trained classifier module 212 is trained based on healthy measurement data across a range of process DOEs, simulations of expected variation, or both. If necessary, specific issues, such as tool differences or difficult-to-extract correlation parameters, are used to generate unhealthy measurement data sets. To separately classify healthy and unhealthy measurement data, the feature extraction module 210 is trained to extract features, whether real or synthetically generated, that help separate healthy and unhealthy measurement data sets. The trained classifier module 212 separates healthy from unhealthy measurement data using any suitable measure of similarity, such as nearest neighbor, average distance, distance from boundary, etc.

[0098] In a further aspect, target measurement quality index values ​​are determined for measurements of production wafers, monitor wafers, or both, on a single measurement tool or across multiple measurement tools to isolate specific issue(s) contributing to degradation of measurement health, e.g., system health, measurement model health, tool signature, process health, etc.

[0099] In general, the training data and measurement data used to train and use the quality monitor models as described herein may be collected from any suitable semiconductor metrology system, including, by way of non-limiting example, a spectroscopic ellipsometer, a spectroscopic reflectometer, a soft x-ray based metrology system, a small angle x-ray scatterometry system, an imaging system, a hyperspectral imaging system, etc.

[0100] As a non-limiting example, FIG. 8 illustrates a system 300 for measuring sample properties and monitoring measurement quality associated with each measurement in accordance with the exemplary methods presented herein. As shown in FIG. 8, system 300 may be used to perform spectroscopic ellipsometry measurements of structure 301. In this embodiment, system 300 may include a spectroscopic ellipsometer equipped with an illuminator 302 and a spectrometer 304. The illuminator 302 of system 300 is configured to generate and direct illumination in a selected wavelength range (e.g., 100-2500 nm) toward a structure disposed on the surface of the sample on which structure 301 is fabricated. The spectrometer 304 is then configured to receive illumination reflected from structure 301. Note further that light exiting illuminator 302 is polarized using polarization state generator 307 to generate polarized illumination beam 306. Radiation reflected by structure 301 passes through polarization state analyzer 309 and is sent to spectrometer 304. The radiation received by the spectrometer 304 in the collected beam 308 is analyzed with respect to polarization state, allowing the spectrometer to perform spectral analysis of the radiation that passes through the analyzer. These spectra 311 are passed to a computing system 330 for analysis of the structures described herein.

[0101] As shown in Figure 8, system 300 includes a single measurement technique (i.e., SE). However, in general, system 300 can include any number of different measurement techniques. By way of non-limiting example, system 300 can be configured as a spectroscopic ellipsometer (including Mueller matrix ellipsometry), a spectroscopic reflectometer, a spectroscopic scatterometer, an overlay scatterometer, an angle-resolved beam profile reflectometer, a polarization-resolved beam profile reflectometer, a beam profile reflectometer, a beam profile ellipsometer, any single- or multi-wavelength ellipsometer, or any combination thereof. Furthermore, in general, the measurement data collected by different measurement techniques and analyzed according to the methods described herein can be collected from multiple tools, a single tool integrating multiple techniques, or a combination thereof.

[0102] In further embodiments, system 300 may include one or more computing systems 330 used to perform measurements of the structure and determine quality metric values ​​according to the methods described herein. The one or more computing systems 330 may be communicatively coupled to spectrometer 304. In one aspect, one or more computing systems 330 are configured to receive measurement data 311 associated with measurements of the structure being measured (e.g., structure 301).

[0103] In one aspect, the computing system 330 is configured as a measurement quality engine configured to determine values ​​of one or more target measurement quality indicators, each indicative of a measurement quality associated with a different operational element of the metrology system, as described herein.

[0104] Generally, the training data includes actual measurement data, simulated measurement data, or both. In some embodiments, the training data includes simulated measurement data from a simulation of measurement of a particular target with a particular measurement tool. In some embodiments, the training data includes actual measurement data from measurement of a particular target with a particular measurement tool. In some of these embodiments, the measurement data includes actual measurements of targets on nominal wafers. In some of these embodiments, the measurement data includes actual measurements of targets on design-of-experiment (DOE) wafers with known, programmed values ​​of parameters of interest.

[0105] The training data may be simulated, collected, or both, for any of a variety of metrology objectives, which in some embodiments may include measurement accuracy, measurement accuracy tracking, tool-to-tool matching, wafer-to-wafer variation, etc.

[0106] As mentioned above, training data is required to train the measurement models. The training data used to train the quality monitor models can be the same training dataset used to train the corresponding measurement models, a subset of the training dataset used to train the corresponding measurement models, or a different dataset from the training dataset used to train the corresponding measurement models. In a preferred embodiment, the training data used to train the quality monitor models is the same dataset or a subset of the dataset used to train the corresponding measurement models.

[0107] In some embodiments, reference data is used to train the quality monitor model. The reference data includes reliable values ​​of one or more parameters of interest associated with the measurements. In some embodiments, the reference data is obtained from a reference metrology system, i.e., a metrology system that is trusted to provide adequate measurements of the parameters of interest. In some embodiments, the reference data is provided by a user of the metrology system based on experience. In some embodiments, the reference data is provided as known, programmed values ​​of the parameters of interest associated with the measured DOE wafer. In some embodiments, the reference data is derived from statistical analysis of measurements of the parameters of interest by multiple different measurement techniques. In some embodiments, the reference data is derived from known physical constraints.

[0108] Generally, reference data expresses the values ​​of one or more parameters of interest in various terms, such as precision, accuracy (e.g., bias), tracking (e.g., correlation to a reference), tool-to-tool matching, within-wafer variation, wafer average, wafer signature, wafer-to-wafer variation, etc.

[0109] In some embodiments, the features of the training dataset are directly used as input data for training the quality monitor model. However, in other embodiments, the training dataset is converted to reduced dimensions, and the features of the reduced training dataset are directly used to train the quality monitor model. In many practical situations, it is preferable to reduce the dimensionality of the training dataset to reduce the computational load. In one example, a set of measurement spectra used as training data may contain thousands of features (e.g., 15,000 measurement signals). However, using data reduction, the dimensionality of the training dataset can be reduced by an order of magnitude or more (e.g., 200 signals).

[0110] In some embodiments, data feature dimensionality reduction is achieved by principal component analysis, where a measurement data set is decomposed into a much smaller set of principal components. Generally, any suitable data reduction technique may be used, for example, Fourier analysis, wavelet analysis, discrete cosine transform analysis, etc.

[0111] In some embodiments, the quality monitor model is trained based on specific domain knowledge associated with the training data. The domain knowledge is expressed as one or more probability distributions and used to normalize the optimization process used during training. In this way, the optimization process is physically normalized by one or more equations for physics-based probability distributions. By way of non-limiting example, probability distributions related to measurement accuracy, tool-to-tool matching, tracking, within-wafer variation, etc., are used to physically normalize the optimization process. In this way, domain knowledge gained from experience, measurement data, and physics is directly expressed in the objective function that drives the optimization of the quality monitor model. As a result, domain knowledge is fully utilized in the quality monitor model development process.

[0112] In another aspect, the training data and measurement data used to train and use the quality monitor models described herein are simulated and / or collected from multiple targets located closely together on the wafer.

[0113] In some embodiments, the measurement target during the measurements described herein may be an actual device structure rather than a dedicated metrology target.

[0114] In another aspect, the training data and measurement data used to train and use the quality monitor models described herein are collected from one or more targets by multiple different metrology systems. In one example, measurement data from a first target may be collected by a spectroscopic ellipsometer, measurement data from the same target or a different target may be collected by a small-angle x-ray scatterometry (SAXS) system, and measurement data from the same target or a different target may be collected by an imaging-based metrology system.

[0115] In another further aspect, a plurality of different parameters of interest are measured, and a corresponding target measurement quality indicator is evaluated for each different parameter of interest. In some embodiments, the target measurement quality indicators associated with the plurality of different parameters of interest are used to more clearly identify sources of measurement uncertainty associated with a particular measurement recipe, more clearly identify when process parameter changes are needed, or both.

[0116] In yet another aspect, values ​​of the parameters of interest, corresponding target measurement quality indicator values, or both, associated with structures proximate the measurement target are used to more clearly identify when process parameter changes are required.

[0117] In yet another aspect, the measurements described herein can be used to provide active feedback to a process tool (e.g., a lithography tool, an etch tool, a deposition tool, etc.). For example, values ​​of measurement parameters determined based on the measurement methods described herein can be communicated to an etch tool to adjust the etch time to achieve a desired etch depth. Similarly, etch parameters (e.g., etch time, diffusivity, etc.) or deposition parameters (e.g., time, concentration, etc.) can be included in a measurement model to provide active feedback to an etch tool or a deposition tool, respectively. In some examples, process parameter modifications determined based on measured device parameter values ​​can be communicated to a process tool. In one embodiment, the computing system 130 determines values ​​of one or more target parameters. Additionally, the computing system 130 communicates control commands to a process controller based on the determined values ​​of the one or more target parameters. The control commands cause the process controller to modify a state of the process (e.g., stop the etch process, modify the diffusivity, etc.). In one example, the control commands cause the process controller to adjust the focus of the lithography system, the dose of the lithography system, or both. In another example, the control command causes the process controller to change the etch rate to improve the wafer uniformity of the measured CD parameter.

[0118] In some examples, the metrology model is implemented as an element of a SpectraShape® optical critical dimension metrology system available from KLA-Tencor Corporation of Milpitas, Calif. In this manner, the model is created and made available for use immediately after a spectrum is collected by the system.

[0119] In some other examples, the measurement model is implemented offline, for example, by a computing system implementing AcuShape® software available from KLA-Tencor Corporation of Milpitas, Calif. The resulting trained model can be incorporated as an element of an AcuShape® library that can be accessed by the metrology system that performs the measurements.

[0120] 9 illustrates, in at least one novel aspect, a method 300 for estimating the value of a quality metric indicative of one or more performance characteristics of a semiconductor measurement. Method 300 is suitable for implementation by a metrology system, such as metrology system 100 shown in FIG. 1 of the present invention. It is recognized that, in one aspect, the data processing blocks of method 300 may be performed via pre-programmed algorithms executed by one or more processors of computing system 130, or any other general-purpose computing system. It is recognized that the specific structural aspects of metrology system 100 herein are not intended to represent limitations and should be construed as illustrative only.

[0121] In block 401, an amount of illumination radiation is directed from an illumination source of a semiconductor metrology system to a measurement site on the surface of a semiconductor wafer.

[0122] In block 402, the amount of radiation collected from the measurement site as a function of the amount of illumination radiation is detected.

[0123] In block 403, a first amount of measurement data is generated that characterizes the detected amount of radiation from the measurement site.

[0124] At block 404, values ​​of one or more target measurement quality indicators are estimated based on the first amount of measurement data using the trained quality monitor model, wherein each of the one or more target measurement quality indicators is indicative of a measurement quality associated with a different operational element of the semiconductor measurement system.

[0125] In further embodiments, system 100 includes one or more computing systems 130 used to perform measurements on semiconductor structures based on spectroscopic measurement data collected according to the methods described herein. The one or more computing systems 130 may be communicatively coupled to one or more spectrometers, active optical elements, process controllers, etc. In one aspect, the one or more computing systems 130 are configured to receive measurement data related to spectral measurements of structures on wafer 104.

[0126] It should be appreciated that one or more steps described throughout this disclosure may be performed by a single computing system 130 or may be performed by multiple computing systems 130. Furthermore, different subsystems of system 100 may include computing systems suitable for performing at least a portion of the steps described herein. Accordingly, the foregoing description should not be construed as limiting the present invention, but rather as merely illustrative.

[0127] Additionally, computing system 130 may be communicatively coupled to the spectrometer in any manner known in the art. For example, one or more computing systems 130 may be coupled to a computing system associated with the spectrometer. In another example, the spectrometer may be directly controlled by a single computing system coupled to computing system 130.

[0128] Computing system 130 of system 100 may be configured to receive and / or acquire data or information from subsystems of the system (e.g., spectrometer, etc.) over a transmission medium, which may include wired and / or wireless portions. In this manner, the transmission medium may serve as a data link between computing system 130 and other subsystems of system 100.

[0129] The computing system 130 of the system 100 may be configured to receive and / or acquire data or information (e.g., measurement results, modeling inputs, modeling results, reference measurement results, etc.) from other systems via a transmission medium, which may include wired and / or wireless portions. In this manner, the transmission medium may serve as a data link between the computing system 130 and other systems (e.g., memory onboard the system 100, external memory, or other external systems). For example, the computing system 130 may be configured to receive measurement data from a storage medium (i.e., memory 132 or external memory) via a data link. For example, spectral results obtained using the spectrometers described herein may be stored in a permanent or semi-permanent memory device (e.g., memory 132 or external memory). In this regard, spectral results may be imported from onboard memory or an external memory system. Additionally, the computing system 130 may transmit data to other systems via the transmission medium. For example, measurement models or estimated parameter values ​​determined by the computing system 130 may be communicated to and stored in external memory. In this regard, measurement results may be exported to another system.

[0130] Computing system 130 may include, but is not limited to, a personal computer system, a mainframe computer system, a workstation, a graphics computer, a parallel processor, or any other device known in the art. In general, the term "computing system" may be broadly defined to encompass any device having one or more processors that execute instructions from a memory medium.

[0131] The program instructions 134 implementing the methods as described herein may be transmitted via a transmission medium, such as a wire, cable, or wireless transmission link. For example, as shown in Figure 1, the program instructions 134 stored in memory 132 are transmitted to the processor 131 via bus 133. The program instructions 134 are stored in a computer-readable medium (e.g., memory 132). Exemplary computer-readable media include read-only memory, random-access memory, a magnetic or optical disk, or a magnetic tape.

[0132] As used herein, the term "critical dimension" includes any critical dimension of a structure (e.g., bottom critical dimension, middle critical dimension, top critical dimension, sidewall angle, grating height, etc.), the critical dimension between any two or more structures (e.g., the distance between two structures), and the displacement between two or more structures (e.g., the overlay displacement between overlapping grating structures, etc.). Structures may include three-dimensional structures, patterned structures, overlay structures, etc.

[0133] As used herein, the terms "critical dimension application" or "critical dimension measurement application" include any critical dimension measurement.

[0134] As described herein, the term "metrology system" includes any system used at least in part to characterize a specimen in any manner, including metrology applications such as critical dimension metrology, overlay metrology, focus / dose metrology, and composition metrology. However, such terminology does not limit the scope of the term "metrology system" as described herein. Additionally, system 100 may be configured to measure patterned and / or unpatterned wafers. The metrology system may be configured as an LED inspection tool, an edge inspection tool, a backside inspection tool, a macro inspection tool, or a multi-mode inspection tool (including data from more than one platform simultaneously), as well as any other metrology or inspection tool that would benefit from calibration of system parameters based on critical dimension data.

[0135] Described herein are various embodiments of semiconductor metrology systems that can be used to measure specimens in any semiconductor processing tool (e.g., an inspection system or a lithography system). The term "specimen" is used herein to refer to a wafer, a reticle, or any other specimen that can be processed (e.g., printed or inspected for defects) by means known in the art.

[0136] As used herein, the term "wafer" generally refers to a substrate formed of a semiconductor or non-semiconductor material. Examples include, but are not limited to, monocrystalline silicon, gallium arsenide, and indium phosphide. Such substrates may be commonly found and / or processed in semiconductor foundries. In some cases, a wafer may include only the substrate (i.e., a bare wafer). Alternatively, a wafer may include one or more layers of different materials formed on a substrate. One or more layers formed on a wafer may be "patterned" or "unpatterned." For example, a wafer may include multiple dies having repeatable pattern features.

[0137] A "reticle" can be a reticle at any stage in the reticle manufacturing process or a finished reticle that may or may not be released for use in a semiconductor foundry. A reticle, or "mask," is generally defined as a substantially transparent substrate having substantially opaque regions formed thereon and arranged in a pattern. The substrate may comprise, for example, a glass material such as amorphous SiO2. The reticle may be placed over a resist-covered wafer during the exposure step of a lithography process so that the pattern on the reticle can be transferred to the resist.

[0138] One or more layers formed on a wafer may be patterned or unpatterned. For example, a wafer may include multiple dies, each with repeatable pattern features. The formation and processing of such layers of material may ultimately result in a completed device. Many different types of devices may be formed on a wafer, and the term wafer, as used herein, is intended to encompass a wafer having any type of device fabricated thereon, as known in the art.

[0139] In one or more exemplary embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media, including any medium that facilitates transfer of a computer program from one place to another. Storage media may be any available medium that can be accessed by a general-purpose or special-purpose computer. By way of example, and not limitation, such computer-readable media may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code means in the form of instructions or data structures and that can be accessed by a general-purpose or special-purpose computer, or a general-purpose or special-purpose processor. Furthermore, any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, microwave, etc., the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, microwave, etc. are included in the definition of medium. As used herein, disk and disc include compact discs (CDs), laser discs, optical discs, digital versatile discs (DVDs), floppy disks, and Blu-ray discs, where disks typically reproduce data magnetically, while discs reproduce data optically using lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0140] Although several specific embodiments are described above for illustrative purposes, the teachings of this patent document are generally applicable and are not limited to the specific embodiments described above. Accordingly, various modifications, adaptations, and combinations of various features of the described embodiments may be made without departing from the scope of the invention as set forth in the claims.

Claims

1. an illumination source configured to provide a quantity of illumination radiation toward a measurement site on a surface of the semiconductor wafer during semiconductor measurement; a detector configured to detect an amount of radiation collected from the measurement site in response to the amount of illumination radiation; 1. A computing system comprising: generating a first amount of measurement data characterizing the amount of the detected radiation from the measurement site; and a computing system configured to use a trained quality monitor model to determine values ​​of one or more target measurement quality indicators based on the first amount of measurement data, where the values ​​of the one or more target measurement quality indicators determine whether one or more operational elements of the semiconductor measurement are within specification, the one or more operational elements including any of a measurement model used to estimate a value of a parameter of interest characterizing the measurement site from the first amount of measurement data, a system used to perform the semiconductor measurement, and an underlying manufacturing process used to fabricate a structure on the semiconductor wafer at the measurement site; A system comprising:

2. The computing system further comprises: estimating values ​​of one or more object parameters characterizing the measurement site from the measurement data of the first quantity characterizing the amount of detected radiation based on a trained measurement model; generating a first amount of simulated measurement data corresponding to estimates of the one or more parameters of interest using the trained measurement model; determining an indication of a comparison between the measurement data of the first quantity and the simulated measurement data of the first quantity; extracting one or more features from the indicator of the comparison between the first amount of measurement data and the first amount of simulated measurement data, wherein the determination of values ​​of the one or more target measurement quality indicators is based on the one or more features extracted from the indicator of the comparison between the first amount of measurement data and the first amount of simulated measurement data. The system of claim 1 .

3. The computing system further comprises: estimating values ​​of one or more of a first plurality of object parameters characterizing the measurement site from the first amount of measurement data characterizing the amount of detected radiation based on a trained measurement model; generating a first amount of simulated measurement data corresponding to estimates of the one or more parameters of interest using the trained measurement model; determining a difference between the measurement data of the first quantity and the simulated measurement data of the first quantity; estimating values ​​of one or more of a second plurality of target parameters characterizing the measurement site from the difference between the measurement data of the first quantity and the simulated measurement data of the first quantity; determining a difference between the value of one or more of the parameters of the second plurality of subjects and a respective reference value of the one or more of the parameters of the second plurality of subjects, wherein the determination of values ​​of the one or more target measurement quality indicators is based on the difference between the value of one or more of the parameters of the second plurality of subjects and the reference value of the one or more of the parameters of the second plurality of subjects. The system of claim 1 .

4. The computing system further comprises: estimating values ​​of one or more of a first plurality of object parameters characterizing the measurement site from the first amount of measurement data characterizing the amount of detected radiation based on a trained measurement model; generating a first amount of simulated measurement data corresponding to estimates of values ​​of one or more of the parameters of the first plurality of subjects using the trained measurement model; estimating a first value of at least one parameter of the first plurality of subject parameters characterizing the measurement site from the first amount of measurement data by a trained monitor model; estimating a second value of the at least one parameter characterizing the measurement site by the trained monitor model from simulated measurement data of the first quantity; determining a difference between the first value and the second value, wherein the determination of values ​​of the one or more target measurement quality indicators is based on the difference between the first value and the second value. The system of claim 1 .

5. The computing system further comprises: estimating first values ​​of one or more object parameters characterizing the measurement site from the first amount of measurement data characterizing the amount of detected radiation based on a trained measurement model; estimating second values ​​of the one or more subject parameters characterizing the measurement site from the first amount of measurement data based on a trained monitor model; determining a difference between the first and second values ​​of the one or more target parameters, wherein the determination of values ​​of the one or more target measurement quality indicators is based on the difference between the first and second values ​​of the one or more target parameters. The system of claim 1 .

6. The computing system further comprises: extracting one or more features from the first amount of measurement data, wherein the determination of values ​​of the one or more target measurement quality indicators is based on the one or more features extracted from the first amount of measurement data. The system of claim 1 .

7. 7. The system of claim 6, wherein the determining the value of the one or more target quality indicators comprises a comparison between the one or more features extracted from the first amount of measurement data and a corresponding set of the one or more features extracted from a quantity of measurement data measured before the first amount of measurement data.

8. 3. The system of claim 2, wherein the extraction of the one or more features from the indicator of the comparison between the first amount of measurement data and the first amount of simulated measurement data comprises reducing a dimensionality of the indicator of the comparison.

9. The system of claim 8 , wherein the reduction of the dimensionality of the measure of the comparison includes any of principal component analysis, Fourier analysis, wavelet analysis, and discrete cosine transform analysis.

10. 10. The system of claim 1, wherein the illumination source and the detector are configured as part of one of a spectroscopic ellipsometer, a spectroscopic reflectometer, a soft x-ray reflectometer, a small angle x-ray scatterometer, an imaging system, and a hyperspectral imaging system.

11. The system of claim 1 , wherein the operational element is one of a system parameter, a process parameter, and a parameter characterizing the structure being measured.

12. providing an amount of illumination radiation from an illumination source of a semiconductor measurement system toward a measurement site on a surface of the semiconductor wafer during semiconductor measurement; detecting an amount of radiation collected from the measurement site in response to the amount of illumination radiation; generating a first amount of measurement data characterizing the amount of the detected radiation from the measurement site; using the trained quality monitor model to estimate values ​​of one or more target measurement quality indicators based on the first quantity of measurement data, wherein the values ​​of the one or more target measurement quality indicators determine whether one or more operational elements of the semiconductor measurement are within specifications, the one or more operational elements including any of a measurement model used to estimate a value of a parameter of interest characterizing the measurement site from the first quantity of measurement data, a system used to perform the semiconductor measurement, and an underlying manufacturing process used to fabricate a structure on the semiconductor wafer at the measurement site; A method comprising:

13. estimating values ​​of one or more object parameters characterizing the measurement site from the measurement data of the first amount characterizing the amount of detected radiation based on a trained measurement model; determining a first amount of simulated measurement data from the estimates of the one or more parameters of interest by the trained measurement model; determining an indication of a comparison between the measurement data of the first quantity and the simulated measurement data of the first quantity; extracting one or more features from the indicator of the comparison between the first amount of measurement data and the first amount of simulated measurement data, wherein the determination of values ​​of the one or more target measurement quality indicators is based on the one or more features extracted from the indicator of the comparison between the first amount of measurement data and the first amount of simulated measurement data; The method of claim 12 further comprising:

14. estimating values ​​of one or more of a first plurality of object parameters characterizing the measurement site from the first amount of measurement data characterizing the amount of detected radiation based on a trained measurement model; generating, by the trained measurement model, a first amount of simulated measurement data corresponding to estimates of values ​​of one or more of the parameters of the first plurality of subjects; determining a difference between the measurement data of the first quantity and the simulated measurement data of the first quantity; estimating values ​​of one or more of a second plurality of target parameters characterizing the measurement site from the difference between the measurement data of the first quantity and the simulated measurement data of the first quantity; determining a difference between the values ​​of one or more of the parameters of the second plurality of subjects and a respective reference value of the one or more of the parameters of the second plurality of subjects, wherein the determination of values ​​of the one or more target measurement quality indicators is based on the difference between the values ​​of one or more of the parameters of the second plurality of subjects and the reference value of the one or more of the parameters of the second plurality of subjects; The method of claim 12 further comprising:

15. estimating values ​​of one or more of a first plurality of object parameters characterizing the measurement site from the first amount of measurement data characterizing the amount of detected radiation based on a trained measurement model; generating, by the trained measurement model, a first amount of simulated measurement data corresponding to estimates of values ​​of one or more of the parameters of the first plurality of subjects; estimating, by a trained monitor model from the first amount of measurement data, a first value of at least one parameter of the first plurality of subject parameters characterizing the measurement site; estimating, by the trained monitor model, a second value of the at least one parameter characterizing the measurement site from simulated measurement data of the first quantity; determining a difference between the first value and a second value, wherein the determination of values ​​of the one or more target measurement quality indicators is based on the difference between the first value and the second value; The method of claim 12 further comprising:

16. estimating first values ​​of one or more object parameters characterizing the measurement site from the measurement data characterizing the amount of detected radiation based on a trained measurement model; estimating second values ​​of the one or more subject parameters characterizing the measurement site from the first amount of measurement data based on a trained monitor model; determining a difference between the first and second values ​​of the one or more target parameters, wherein the determination of values ​​of the one or more target measurement quality indicators is based on the difference between the first and second values ​​of the one or more target parameters; The method of claim 12 further comprising:

17. 13. The method of claim 12, further comprising extracting one or more features from the first amount of measurement data, wherein the determining of the values ​​of the one or more target measurement quality indicators is based on the one or more features extracted from the first amount of measurement data.

18. 20. The method of claim 17, wherein the determining the value of the one or more target quality indicators comprises a comparison between the one or more features extracted from the first amount of measurement data and a corresponding set of the one or more features extracted from a quantity of measurement data measured before the first amount of measurement data.

19. an illumination source configured to provide a quantity of illumination radiation toward a measurement site on a surface of the semiconductor wafer during semiconductor measurement; a detector configured to detect an amount of radiation collected from the measurement site in response to the amount of illumination radiation; 1. A non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to: generating a first amount of measurement data characterizing the amount of the detected radiation from the measurement site; and a non-transitory computer-readable medium for causing a trained quality monitor model to be used to determine values ​​of one or more target measurement quality indicators based on the first amount of measurement data, wherein the values ​​of the one or more target measurement quality indicators determine whether one or more operational elements of the semiconductor measurement are within specifications, the one or more operational elements including any of a measurement model used to estimate a value of a parameter of interest characterizing the measurement site from the first amount of measurement data, a system used to perform the semiconductor measurement, and an underlying manufacturing process used to fabricate a structure on the semiconductor wafer at the measurement site; A system comprising:

Images