System and computer program product of using machine vision as input to a cmp process control algorithm

Abstract

During chemical mechanical polishing of a substrate, a first in-situ monitoring system determines signal values ​​that depend on the thickness of the layer at a measurement point on the substrate being polished. A second in-situ imaging system generates an image of at least this measurement point on the substrate. Machine vision processing (e.g., a convolutional neural network) is used to determine feature values ​​of the measurement point based on the image. The measurement value is then calculated based on the feature values ​​and the signal value.

Description

Systems and computer program products for using machine vision as an input to a CMP process control algorithm This disclosure relates to optical monitoring of substrates during processing such as chemical mechanical polishing. Integrated circuits are typically formed on a substrate by sequentially depositing conductive, semiconductor, or insulating layers onto a silicon wafer. One manufacturing step involves depositing a filler layer onto a non-planar surface and planarizing the filler layer until the non-planar surface is exposed. For some applications, the filler layer is planarized until the top surface of a patterned layer is exposed. For example, a conductive filler layer may be deposited on a patterned insulating layer to fill trenches or holes in the insulating layer. After planarization, the portions of the conductive layer remaining between the protruding patterns of the insulating layer form vias, plugs, and lines that provide conductive paths between thin film circuits on the substrate. For other applications, the filler layer is planarized until a predetermined thickness remains on the underlying layer. For example, a deposited dielectric layer may be planarized for lithography. Chemical mechanical polishing (CMP) is an accepted planarization method. This planarization method typically requires mounting the substrate on a carrier head. The exposed surface of the substrate is typically pressed against a rotating polishing pad having a durable rough surface. The carrier head provides a controllable load on the substrate to press the substrate against the polishing pad. A polishing liquid, such as a slurry having abrasive particles, is typically supplied to the surface of the polishing pad. One problem in CMP is using an appropriate polishing rate to achieve a desired profile, e.g., a substrate layer that has been planarized to a desired flatness or thickness or from which a desired amount of material has been removed. Variations in the initial thickness of the substrate layer, slurry distribution, polishing pad conditions, relative speed between the polishing pad and the substrate, and load on the substrate result in variations in the material removal rate across the substrate (and between substrates). These variations result in variations in the time and amount of material removed to reach the polishing endpoint. Thus, it is not possible to determine the polishing endpoint based solely on the polishing time, nor is it possible to obtain a desired profile by simply applying a constant pressure. In some systems, the substrate is monitored in-situ during polishing, e.g., by an optical monitoring system or an eddy current monitoring system. Thickness measurements from the in-situ monitoring system can be used to adjust the pressure applied to the substrate, thereby adjusting the polishing rate and reducing within-wafer non-uniformity (WIWNU). A polishing system includes: a platen for holding a polishing pad; a carrier head for keeping a substrate in contact with the polishing pad; a motor for generating relative movement between the platen and the carrier head; a first in-situ monitoring system for generating a signal that depends on the thickness of a layer at a measurement point on the substrate; a second in-situ imaging system for generating an image of at least the measurement point of the substrate substantially at the same time as the first in-situ monitoring system generates a signal for the measurement point on the substrate; and a controller. The controller is configured to: receive the image from the second in-situ imaging system and determine a characteristic value for the measurement point based on the image using machine vision processing; receive the signal from the in-situ monitoring system; generate a measurement value based on both the characteristic value and the signal value; and based on the measurement value, perform at least one of the following: stop polishing the substrate; or adjust polishing parameters. In another aspect, a computer program product for controlling the processing of a substrate, the computer program product includes instructions that cause one or more processors to: receive a signal value from a first in-situ monitoring system, the signal value depending on the thickness of a layer at a measurement point on a substrate being polished; receive image data for at least the measurement point of the substrate from a second in-situ imaging system; determine a characteristic value for the measurement point based on the image using machine vision processing; generate a measurement value based on both the characteristic value and the signal value, and based on the measurement value, perform at least one of the following: stop polishing the substrate; or adjust polishing parameters. Embodiments may include one or more of the following features. Machine vision processing may include processing the image using an artificial neural network. The artificial neural network may be a convolutional neural network. The controller may be configured to train the artificial neural network by backpropagation using training data that includes the image and known characteristic values of the image. The first in-situ monitoring system may include a spectroscopic monitoring system to generate a measurement spectrum for the measurement point. The artificial neural network may be configured to determine a category of a portion of the substrate corresponding to the measurement point, and the category may correspond to a type of structure on the substrate. The type of structure may include at least one of an array, a scribe line, a periphery, and a contact pad. One of a plurality of reference spectrum libraries may be selected based on the category. The first in-situ monitoring system may include an eddy current monitoring system to generate a signal value for the measurement point. The artificial neural network may be configured to determine a geometric value of a characteristic that affects the current in the measurement point. The geometric value may include at least one of a distance, a dimension, or an orientation. A portion of the image data corresponding to the measurement point may be determined. The image data from the second in-situ imaging system may be synchronized with the signal collected from the first in-situ monitoring system. Some embodiments may include one or more of the following advantages. Process control techniques can be directed to performance-sensitive portions of the die. The thickness of layers on a substrate can be measured more accurately and / or more quickly. Wafer-to-wafer non-uniformity and within-wafer non-uniformity (WIWNU and WTWNU) can be reduced, and the reliability of an endpoint system for detecting a desired process endpoint can be improved. Post-CMP metrology can be based on the yield of the product and / or performance-sensitive portions, rather than the average die thickness (which may include die regions that are not relevant to product performance). Details of one or more specific embodiments are disclosed in the accompanying drawings and the description below. Other possible aspects, features, and advantages will be apparent from the description, drawings, and claims. Various techniques, such as eddy current monitoring and optical monitoring, can be used to monitor a substrate during processing. Such monitoring techniques can be performed in a two-stage manner. First, an original signal from a monitoring system (e.g., a measurement spectrum from a spectrophotometer or a voltage from an eddy current monitoring system) is converted into a more useful measurement form (e.g., an indicator representing a polishing process or a thickness value). Then, the sequence of measurements over time can be monitored for process control. For example, a function can be fit to the measurement sequence, and the time at which the function projects to reach a threshold can be used to trigger a polishing endpoint or control other polishing parameters. If the sensor of the monitoring system scans across the substrate, measurements can be made at different locations on the substrate. Thus, measurements can be made in different regions of the substrate (e.g., within the die rather than within the scribe lines), or in different regions within the die (e.g., arrays, contact pads, etc.). These different regions can have different properties and provide different original signals. It would be useful to determine the type of regions where measurements are made so that the original signals are correctly converted into useful measurements. Although the radial position of the measurement can be determined (e.g., due to rotational sliding of the substrate relative to the carrier head), the angular position of the measurement on the substrate may not be known at all. Thus, in the case of making measurements within a die, commercial in-situ monitoring techniques do not take into account the angular position when converting the original signal into a useful measurement value. Moreover, although some monitoring systems perform filtering to exclude some original signals, such as based on the shape of the spectrum to exclude a spectrum, such techniques do not use information from the surrounding portion of the substrate. However, images collected by an in-situ imager can be processed by machine learning techniques, such as convolutional neural networks, to determine characteristics of a substrate on which measurements are performed by another monitoring system. Such characteristics can be, for example, the type of area where the measurement is performed (e.g., scribe lines, arrays, peripheries), or the relative orientation and / or distance of various features (e.g., guard rings) with respect to the measurement location. This characteristic can then be provided as an input to the in-situ monitoring system to affect the conversion of the raw signal to the measurement. FIG. 1 illustrates an example of a polishing apparatus 20. The polishing apparatus 20 may include a rotatable disk-shaped platform 22, and a polishing pad 30 is located on the disk-shaped platform 22. The platform can be operated to rotate about axis 23. For example, a motor 24 can rotate a drive shaft 26 to rotate the platform 22. The polishing pad 30 can be detachably fixed to the platform 22, for example, by an adhesive layer. The polishing pad 30 can be a two-layer polishing pad having an outer polishing layer 32 and a softer backing layer 34. A window 36 can be formed in the polishing pad 30. The polishing apparatus 20 can include a polishing liquid supply port 40 to release a polishing liquid 42, such as a polishing slurry, onto the polishing pad 30. The polishing apparatus 20 can also include a polishing pad conditioner to polish the polishing pad 30 to maintain the polishing pad 30 in a consistent abrasive state. A carrier head 50 can be operated to hold the substrate 10 against the polishing pad 30. Each carrier head 50 also includes a plurality of independently controllable pressure chambers, such as three chambers 52a - 52c, and the chambers 52a - 52c can apply independently controllable pressure to associated zones on the substrate 10. The central area on the substrate can be substantially circular, and the remaining areas can be concentric annular areas surrounding the central area. The chambers 52a - 52c can be defined by a flexible membrane 54 having a bottom surface on which the substrate 10 is mounted. The carrier head 50 can also include a retaining ring 56 to hold the substrate 10 under the flexible membrane 54. Although only three chambers are illustrated in FIG. 1 for ease of illustration, there can be one or two chambers, or four or more chambers, such as five chambers. Additionally, other mechanisms can be used in the carrier head 50 to adjust the pressure applied to the substrate, such as piezoelectric actuators. Each carrier head 50 is suspended by a support structure 60, such as a rotating turret or track, and is connected to a carrier head rotation motor 64 by a drive shaft 62 to allow the carrier head to rotate about axis 51. Optionally, each carrier head 50 can oscillate laterally (e.g., on a slider on a rotating turret, by an action along a track); or can oscillate laterally by a rotational oscillation of the rotating turret itself. During operation, the platform 22 rotates about the central axis 23 of the platform, and the carrier head 50 rotates about the central axis 51 of the carrier head and is laterally displaced across the top surface of the polishing pad 30. The grinding equipment further includes a first in-situ monitoring system 100 and a second in-situ imaging system 150. The first in-situ monitoring system 100 and the second in-situ imaging system 150 can be used together to control grinding parameters, such as the pressure in one or more chambers 52a - 52c, and / or to detect the grinding end point and / or stop grinding. The first in-situ monitoring system 100 includes a sensor 100a (see Figure 3), and the sensor 100a generates a raw signal that depends on the thickness of the layer being ground. The first in-situ monitoring system 100 can be, for example, an eddy current monitoring system or an optical monitoring system, such as a spectroscopic monitoring system. The sensor can be configured to scan across the substrate. For example, the sensor can be fixed to the platform 22 and rotate with it, such that with each rotation of the platform, the sensor scans across the substrate in an arc. Referring to Figure 2A, as an optical monitoring system, the first in-situ monitoring system 100 can include a light source 102, a photodetector 104, and a circuit system 106 for sending and receiving signals between a controller 90, such as a computer, and the light source 102 and the photodetector 104. One or more optical fibers can be used to transmit light from the light source 102 to the window 36 and to transmit the light reflected from the substrate 10 to the detector 104. For example, a bifurcated optical fiber 108 can be used to transmit light from the light source 102 to the window 36 and then back to the detector 104. In this embodiment, one end of the bifurcated fiber 108 can provide a sensor that scans across the substrate. If the optical monitoring system is a spectroscopic system, the light source 102 can be operable to emit white light, and the detector 104 can be a spectrometer. Referring to Figure 2B, as an eddy current monitoring system, the first in-situ monitoring system 100 can include a magnetic core 112 and at least one coil 114 wound around a portion of the magnetic core 112. A drive and sense circuit system 116 is electrically connected to the coil 114. The drive and sense circuit system 116 can apply an AC current to the coil 114, and the coil 114 generates a magnetic field between two poles of the magnetic core 112, and this magnetic field can enter the substrate 10. In this embodiment, the magnetic core 112 and the coil 114 can provide a sensor that scans across the substrate. The circuit system 116 can include a capacitor connected in parallel with the coil 114. The coil 114 and the capacitor can together form an LC resonant circuit. When the magnetic field reaches a conductive layer, the magnetic field can pass through and generate a current (if this layer is a loop) or generate eddy currents (if this layer is a sheet). This modifies the equivalent impedance of the LC circuit. The drive and sense circuit system 116 can detect the change in the equivalent impedance and generate a signal that can be sent to the controller 90. In either case, the output of circuit system 106 or 116 can be a digital electronic signal that reaches controller 90 (see FIG. 1) through a rotary coupler 28 (such as a slip ring) in drive shaft 26. Alternatively, circuit system 106 or 116 can communicate with controller 90 via a wireless signal. Some or all of circuit systems 106 or 116 can be installed in platform 22. Controller 90 can be a computing device including a microprocessor, memory, and input / output circuitry, such as a programmable computer. Although shown as a single block, controller 90 can be a networked system with functions distributed across multiple computers. Since controller 90 can perform a portion of the signal processing, such as converting the "raw" signal into a usable measurement value, controller 90 can be considered to provide a part of the first monitoring system. As shown in FIG. 3, due to the rotation of the platform (as indicated by arrow A), when sensor 100a of the first in-situ monitoring system travels under the carrier head, measurements are taken at a sampling frequency. As a result, measurements are taken at position 94 in the arc across substrate 10 (the number of dots is illustrative; more or fewer measurements can be taken depending on the sampling frequency). The substrate can also rotate (as indicated by arrow B) and vibrate radially (as indicated by arrow C). The grinding apparatus 20 can include a position sensor 96, such as an optical interrupter, to sense when sensor 100a of the first in-situ monitoring system 100 is under substrate 10 and when sensor 100a leaves substrate 10. For example, position sensor 96 can be mounted in a fixed position opposite carrier head 70. A marker 98 can be attached to the periphery of platform 22. The attachment point and length of flag 98 are selected such that flag 98 sends a signal to sensor 180 as sensor 100a sweeps under substrate 10. Alternatively or additionally, the grinding apparatus 20 can include an encoder to determine the angular position of platform 22. After one full rotation of the platform, spectra are obtained from different positions on substrate 10. Specifically, some spectra can be obtained from positions closer to the center of substrate 10, while some spectra can be obtained from positions closer to the edge. Controller 90 can be configured to calculate the radial position (relative to the center of substrate 10) for each measurement based on a scan based on timing, motor encoder information, platform rotation or position sensor data, and / or optical detection of the substrate and / or clamp ring edge. The controller can thus correlate the various measurement results with various regions on the substrate. In some embodiments, the time of measurement can be used as an alternative for precise calculation of the radial position. The second in-situ imaging system 150 is placed to generate an image of substantially the same portion of the substrate 10 that the first in-situ monitoring system 100 is measuring. Briefly, the camera of the imaging system is located at the same position as the sensor of the in-situ monitoring system 100. Referring to FIG. 4, the second in-situ imaging system 150 may include a light source 152, a light detector 154, and a circuit system 156 for sending and receiving signals between the controller 90 and the light source 152 and the light detector 154. The light source 152 is available for emitting white light. In one embodiment, the emitted white light includes light having a wavelength of 200 - 800 nanometers. Suitable light sources are white light emitting diode (LED) arrays or xenon lamps or xenon mercury lamps. The light source 152 is directed to direct the light 158 onto the exposed surface of the substrate 10 at a non-zero incident angle α. The incident angle α can be, for example, approximately 30° to 75°, such as 50°. The light source 152 can illuminate a substantially linear elongated area. The elongated area can span the width of the substrate 10. The light source 152 can include optics, such as a beam expander, to spread the light from the light source into the elongated area. Alternatively or additionally, the light source 152 can include a linear array of light sources. The light source 152 itself and the area illuminated on the substrate can be elongated and have a longitudinal axis parallel to the surface of the substrate. A diffuser 160 can be placed in the path of the light 158, or the light source 152 can include a diffuser to diffuse the light before it reaches the substrate 10. The light detector 154 is a camera sensitive to the light from the light source 152, such as a color camera. The camera includes an array of detector elements. For example, the camera can include a CCD array. In some embodiments, the array is a single row of detector elements. For example, the camera can be a line-scanning camera. This row of detector elements can extend parallel to the longitudinal axis of the elongated area illuminated by the light source 152. In the case where the light source 152 includes a row of light-emitting elements, this row of detector elements can extend along a first axis parallel to the longitudinal axis of the light source 152. A row of detector elements can contain 1024 or more elements. The light detector 154 is configured with appropriate focusing optics 162 to project the field of view of the substrate onto the array of detector elements of the light detector 154. The field of view can be long enough to view the entire width of the substrate 10, for example 150 to 300 mm long. The light detector 154 can also be configured such that the pixel width is comparable to the pixel length. For example, the advantage of a line-scanning camera is that its frame rate is very fast. The frame rate can be at least 5 kHz. The frame rate can be set to such a frequency that when the imaged area is scanned across the entire substrate 10, the pixel width is comparable to the pixel length, for example equal to or less than approximately 0.3 millimeters. The light source 152 and the photodetector 154 can be supported in a recess in the platform, for example, in the same recess that holds the sensors of the first in-situ monitoring system 100. A possible advantage of moving the line scan camera and the light source together across the entire substrate is that, for example, the relative angle between the light source and the camera remains constant for different positions across the entire wafer as compared to a conventional 2D camera. Thus, artifacts caused by changes in the viewing angle can be reduced or eliminated. Additionally, the line scan camera can eliminate perspective distortion, while a conventional 2D camera exhibits inherent perspective distortion that then needs to be corrected through image transformation. Optionally, a polarization filter 164 can be placed in the optical path, for example, between the substrate 10 and the photodetector 154. The polarization filter 164 can be a circular polarizer (CPL). A typical CPL is a combination of a linear polarizer and a quarter-wave plate. Orienting the polarization axis of the polarization filter 164 correctly can reduce the haze in the image and sharpen or enhance the desired visual features. The controller 90 assembles the individual image lines from the photodetector 154 into a two-dimensional image. The photodetector 154 can be a color camera having, for example, separate detector elements for each of red, blue, and green, in which case the controller 90 combines the separate image lines from the photodetector 154 into a two-dimensional color image. The two-dimensional color image can include monochromatic images 204, 206, 208 for each color channel (e.g., for each of the red, blue, and green channels). Referring to FIG. 5, the controller 90 can convert the raw signals from the in-situ monitoring system into useful measurement values. The controller 90 uses both the signals from the first in-situ monitoring system 100 and the image data from the second in-situ imaging system 150 to calculate the measurement values. The images collected from the second in-situ imaging system 150 can be synchronized with the data stream collected from the first in-situ monitoring system 100. Specifically, the controller 90 feeds the images from the second in-situ imaging system 150 into a machine vision system 200, which is configured to derive characteristic values of a portion of the substrate measured by the first in-situ monitoring system 100. The machine vision system can include, for example, a neural network 210. The neural network 210 can be a convolutional neural network. The neural network 210 includes a plurality of input nodes 212, such as input nodes 212 for each pixel in the image from the second in-situ imaging system 150. These can include input node N 1 、N 2 …N L. The neural network 210 also includes a plurality of hidden nodes 214 (hereinafter also referred to as "intermediate nodes") and at least one output node 216 that will generate at least one eigenvalue. Generally speaking, the value output by the hidden node 214 is a non-linear function of the weighted sum of the values from the nodes connected to the hidden node. For example, the output of the hidden node 214 (designated as node k) can be expressed as: tanh(0.5 * 𝑎 𝑘 1 ( I 1 ) + 𝑎 𝑘 2 ( I 2 ) + … + 𝑎 𝑘 M ( I M ) + 𝑏 𝑘 ) Equation 1 where tanh is the hyperbolic tangent, a kx is the connection weight between the k-th intermediate node and the x-th input node (among the M input nodes), I M is the value at the M-th input node. However, other non-linear functions can be used instead of tanh, such as the rectified linear unit (ReLU) function and its variants. The architecture of the neural network 210 can vary in depth and width. Although the shown neural network 210 has a single row of intermediate nodes 214, in fact the neural network will include multiple rows, which can have various connections. A convolutional neural network can perform multiple iterations of convolution and merging, and then perform classification. The neural network 210 can be trained, for example, in a training mode using backpropagation with sample images and sample eigenvalues. Thus, in operation, the machine vision system 200 generates eigenvalues based on the images from the second in-situ imaging system 150. This operation can be performed for each value of the "raw signal" received from the first in-situ monitoring system 100. The raw signal from the first in-situ monitoring system 100 and the eigenvalue synchronized with the raw signal (e.g., corresponding to the same point on the substrate) are input into the conversion algorithm module 220. The conversion algorithm module 220 calculates the measured value based on the eigenvalue and the raw signal. The measured value is typically the thickness of the outer layer, but can be a related characteristic, such as the removed thickness. Additionally, the measured value can be a more general representation of the substrate progress during the grinding process, such as an indicator value representing the platform rotation time or the number of platform revolutions at which the measured value is expected to be observed during a grinding process following a predetermined progress. The measured value can be fed into the process control subsystem 240 to adjust the grinding process based on a series of eigenvalues, such as detecting the grinding end point and stopping the grinding and / or adjusting the grinding pressure during the grinding process to reduce grinding non-uniformity. The process control module 240 can output process parameters, such as the pressure for the chamber in the carrier head and / or a signal for stopping the grinding. For example, referring to FIG. 6, the first function 254 can be fitted to the sequence 250 of measured values 252 in the first region, and the second function 264 can be fitted to the sequence 260 of eigenvalues 262 in the second region. The process controller 240 can calculate the predicted times T1 and T2 for the first function and the second function to reach the target value V, and calculate the adjusted process parameters (e.g., the adjusted carrier head pressure) that will cause one of the regions to be ground at a modified rate (as shown by line 270) so that these regions reach the target at approximately the same time. When the function indicates that the eigenvalue reaches the target value V, the grinding end point can be triggered by the process controller 240. In some embodiments, multiple measured values can be combined at the conversion algorithm module 220 or the process control module 240. For example, if the system generates multiple measured values from a single scan of the sensor across the substrate, the conversion algorithm module 220 can combine the multiple measured values from the single scan to generate a single measured value for each scan on the substrate or a single measured value for each radial region. However, in some embodiments, a measured value is generated for each position 94 for which the sensor 100a generates a raw signal value (see FIG. 3). In some embodiments, the neural network 210 generates multiple eigenvalues at multiple output nodes 216. One or more additional eigenvalues (i.e., eigenvalues other than those representing thickness measurements) can represent other characteristics of the substrate, such as wafer orientation, the type of structure on the wafer (e.g., memory array, central processing unit). The additional eigenvalues can be fed into the process control 240. Example 1 The first in-situ monitoring system 100 can be a spectroscopic monitoring system. The sensors of the spectroscopic monitoring system and the second in-situ imaging system 150 can use the same window 36. The data window of the line-scanning camera from the second in-situ imaging system 150 can be centered on the spectroscopic acquisition time of the first in-situ monitoring system 100, and this data window can be used to reconstruct a two-dimensional image of the portion of the substrate 10 from which the spectrum is collected. The machine vision system 200 can include a convolutional neural network (CNN) 210. To train the neural network 210, a series of images from one or more reference substrates can be manually identified with relevant categories (such as arrays, scribed lines, peripheries, contact pads, etc.). Assigning categories to the images is sometimes referred to as "annotation". Then, the images and categories from the reference substrates can be input into the neural network in a training mode (such as using backpropagation) to train the neural network 210 as an image classifier. Note that by using transfer learning, such an image classifier can be trained with a relatively small number of annotated images, in which a pre-trained image classification network is shown some additional images from a new domain. In operation, during the grinding of the product substrate, the images are fed into the neural network 210. The output of the neural network 210 is used in real-time to associate each measured spectrum with the category of the portion of the substrate from which the spectrum is obtained. The image classification performed by the convolutional neural network can be merged with the measured spectrum before being fed into another model for thickness estimation or prediction. Such categories can be used by the conversion algorithm module 220. For example, the controller 90 can store a plurality of reference spectrum libraries, where each reference spectrum has an associated measured value, such as an index value. The controller 90 can select one of the resource libraries based on the category received from the neural network 210. Then, the reference spectrum that best matches the measured spectrum from the selected resource library can be determined, for example, by finding the reference spectrum with the minimum sum of squared differences relative to the measured spectrum. Then the index value of the best-matching reference spectrum can be used as the measured value. Example 2 The first in-situ monitoring system 100 can be an eddy current monitoring system. The sensor 100a of the eddy current monitoring system and the sensor of the second in-situ imaging system 150 are located at the same position, such as in the same recess of the platform. The line-scanning camera of the second in-situ imaging system 150 generates time-synchronized images that cover the entire scan of the sensor 100a across the substrate. The machine vision system 200 may include a convolutional neural network (CNN) 210. To train the neural network 210, the geometric shapes (e.g., position, size, and / or orientation) of the substrate features that affect the current (e.g., guard rings) can be manually identified. Then, the images and geometric values from a reference substrate can be input into the neural network in a training mode (e.g., using backpropagation) to train the neural network 210 as a feature geometry reconstructor. In operation, during the grinding of the product substrate, images are fed into the neural network 210. The output of the neural network 210 is used in real-time to correlate each measurement value from the eddy current monitoring system with the geometric values of the substrate portion from which the spectrum is obtained. The geometric values generated by the neural network 210 can be used by the transformation algorithm module 220. The mapping from the eddy current signal to resistance depends on the relative orientation and position of the features on the substrate. For example, the sensitivity of the sensor 100a to the conductive loop on the substrate can depend on the orientation of the loop. The controller 90 may include a function to calculate a gain based on the geometric values (e.g., orientation). This gain can then be applied to the signal, e.g., the signal value can be multiplied by this gain. Thus, the geometric values can be used to adjust how the eddy current sensor data is interpreted. Conclusion The specific embodiments of the invention described in this specification and all functional operations thereof can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the structural components disclosed in this specification and their structural equivalents (or combinations of one or more of them). The specific embodiments of the invention can be implemented as one or more computer program products, i.e., one or more computer programs tangibly embodied on a machine-readable storage medium for execution or control of the operations of a data processing apparatus (e.g., a programmable processor, a computer, or multiple processors or computers). A computer program (also referred to as a program, software, software application, or code) can be written in any form of programming language, including compiled or interpreted languages, and can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program does not necessarily correspond to a file. The program can be stored in a portion of a file that holds other programs or data, in a single file dedicated to the program being discussed, or in multiple cooperating files (e.g., files that store one or more modules, subroutines, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers located at one site or distributed across multiple sites and interconnected by a communication network. The procedures and logical flows described in this specification can be executed by one or more programmable processors that execute one or more computer programs to perform functions by operating on input data and generating output. The procedures and logical flows can also be executed by special-purpose logic circuitry (and the apparatus can also be implemented as special-purpose logic circuitry), such as an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit). The above-described polishing apparatus and method can be applied to various polishing systems. The movable polishing pad or carrier head (or both the polishing pad and the carrier head) is moved to provide relative movement between the polishing surface and the substrate. For example, the platform can rotate in an orbit rather than rotate. The polishing pad can be a circular (or some other shape) pad fixed to the platform. The polishing system can be a linear polishing system, for example, where the polishing pad is a continuous or linearly moving reel-to-reel belt. The polishing layer can be a standard (e.g., polyurethane with or without fillers) polishing material, a soft material, or a fixed polishing material. The term of relative positioning of components is used; it should be understood that the polishing surface and the substrate can be held in a vertical orientation or some other orientation relative to gravity. Although the above description focuses on chemical mechanical polishing, the control system can be applied to other semiconductor processing technologies, such as etching or deposition, such as chemical vapor deposition. Instead of using a line-scanning camera, a camera that images a two-dimensional area of the substrate can be used. In this case, it may be necessary to combine multiple images. Specific specific embodiments of the present invention have been described. Other specific embodiments are within the scope of the appended claims. 10: Substrate 20: Grinding equipment 22: Platform 23: Shaft 24: Motor 26: Drive shaft 28: Rotary coupler 30: Grinding pad 32: Outer grinding layer 34: Backing layer 36: Window 40: Supply port 42: Grinding liquid 50: Carrier head 51: Shaft 52a - 52c: Chambers 54: Flexible film 56: Snap ring 60: Support structure 62: Drive shaft 64: Carrier head rotation motor 70: Carrier head 90: Controller 94: Position 96: Position sensor 98: Flag 100: First in - situ monitoring system 100a: Sensor 102: Light source 104: Photodetector 106: Circuit system 112: Magnetic core 114: Coil 116: Drive and sensing circuit system 150: Second in - situ imaging system 152: Light source 154: Photodetector 156: Circuit system 158: Light 160: Diffuser 162: Focusing optics 164: Filter 200: Machine vision system 210: Neural network 212: Input node 214: Hidden node 216: Output node 220: Conversion algorithm module 240: Process control subsystem 250: Sequence 254: First function 260: Sequence 262: Eigenvalue 264: Second function 270: Line FIG. 1 shows a schematic cross - sectional view of an example of a grinding equipment. FIG. 2A is a schematic diagram of an in - situ optical monitoring system. FIG. 2B is a schematic diagram of an in - situ eddy current monitoring system. FIG. 3 is a schematic top view of the grinding equipment. FIG. 4 shows a schematic diagram of a line - scanning imaging system. FIG. 5 shows a neural network, as part of the controller of the grinding equipment. FIG. 6 shows a graph of measured values over time. In the various figures, like reference numerals and labels indicate like elements. Domestic filing information (please note in the order of filing institution, date, number) None Foreign filing information (please note in the order of filing country, institution, date, number) None 90: Controller 100: First in - situ monitoring system 150: Second in - situ imaging system 200: Machine vision system 210: Neural network 212: Input node 214: Hidden node 216: Output node 220: Conversion algorithm module 240: Process control subsystem

Claims

1. A polishing system comprising: a support for holding a polishing pad; a rotatable bearing head for holding a substrate in contact with the polishing pad; an in-situ monitoring system for generating a signal comprising a series of raw signal values, the series of raw signal values ​​depending on a thickness of a layer at a measurement point on the substrate; an in-situ imaging system for generating a series of images of the substrate while the in-situ monitoring system generates the series of raw signal values; and a controller configured to receive the series of images from the in-situ imaging system, receive the series of raw signal values ​​from the in-situ monitoring system, and determine a series of orientations of the substrate held by the bearing head, wherein the series of images is processed using machine vision. The series of raw signal values ​​from the in-situ monitoring system are converted into a series of characteristic values ​​for the measurement point by providing the series of raw signal values ​​and the series of orientations of the substrate as inputs to a conversion algorithm, and based on the series of characteristic values, performing at least one of the following: stopping the grinding of the substrate; or adjusting a grinding parameter that affects the grinding of the substrate.

2. The system as described in claim 1, wherein the characteristic value includes a measured thickness of the polished layer.

3. The system as described in claim 1, wherein the machine vision processing includes an artificial neural network.

4. The system as described in claim 3, wherein the machine vision processing includes a convolutional neural network.

5. The system as claimed in claim 3, wherein the controller is configured to train the artificial neural network by backpropagation using training data including images and known feature values ​​of such images.

6. The system as claimed in claim 1, wherein the in-situ monitoring system includes an eddy current monitoring system to generate a signal value for the measurement point.

7. The system as claimed in claim 1, wherein the first in-situ monitoring system includes a spectral monitoring system to generate a measurement spectrum for the measurement point.

8. The system as described in claim 1, wherein the controller is configured to determine a portion of the image corresponding to the measurement point.

9. The system as described in claim 8, wherein the controller is configured to synchronize images from the in-situ imaging system with raw signal values ​​from the in-situ monitoring system.

10. A computer program product for controlling the processing of a substrate, the computer program product being tangibly implemented in a non-transitory computer-readable medium and including instructions to cause a processor to: receive a series of images from an in-situ imaging system; receive a series of raw signal values ​​for a measurement point from an in-situ monitoring system; determine a series of orientations of the substrate held by a carrier head, wherein the series of images is processed using machine vision; convert the series of raw signal values ​​from the in-situ monitoring system to a series of feature values ​​for the measurement point, wherein the series of raw signal values ​​and the series of orientations of the substrate are provided as inputs to a conversion algorithm; and based on the series of feature values, perform at least one of the following: stop grinding the substrate; or adjust a grinding parameter affecting the grinding of the substrate.

11. The computer program product as claimed in claim 10, wherein the feature value includes a measured thickness of the polished layer.

12. The computer program product as described in claim 10, the computer program product comprising instructions for determining a portion of the image data corresponding to the measurement point.

13. The system as described in claim 10, wherein the computer program product includes instructions for synchronizing image data from the in-situ imaging system with signal values ​​from the in-situ monitoring system.

14. The computer program product as described in claim 10, wherein machine vision processing includes feeding the image data into an artificial neural network.

15. The computer program product as described in claim 14, wherein the artificial neural network includes a convolutional neural network.

Images