阅读说明：本技术 以机器视觉作为对cmp工艺控制算法的输入 (Using machine vision as input to CMP process control algorithms ) 是由 B·切里安 J·钱 N·威斯韦尔 D·J·本韦格努 B·A·斯韦德克 T·H·奥斯特海德 于 2019-08-28 设计创作，主要内容包括：在基板的化学机械抛光期间,通过第一原位监测系统来确定取决于经受抛光的基板上的测量点中的层的厚度的信号值。通过第二原位成像系统产生至少基板的测量点的图像。机器视觉处理(例如卷积神经网络)用于基于图像来确定测量点的特征值。然后,基于特征值和信号值两者来计算测量值。(During chemical mechanical polishing of a substrate, a signal value that depends on a thickness of a layer in a measurement point on the substrate undergoing polishing is determined by a first in-situ monitoring system. An image of at least the measurement point of the substrate is generated by a second in situ imaging system. Machine vision processing (e.g., convolutional neural networks) is used to determine feature values of measurement points based on the image. Then, a measurement value is calculated based on both the characteristic value and the signal value.)

1. a polishing system, comprising:

a support for holding a polishing pad;

a carrier head for holding a substrate in contact with the polishing pad;

a motor for producing relative motion between the support and the carrier head;

a first in-situ monitoring system that generates a signal that is dependent on a thickness of a layer in a measurement spot on the substrate;

a second in-situ imaging system that generates an image of at least the measurement point of the substrate at substantially the same time as the in-situ monitoring system generates the signal for the measurement point on the substrate; and

a controller configured to

Receiving the image from the second in situ imaging system and determining feature values of the measurement points based on the image using machine vision processing,

receiving the signal from the in-situ monitoring system,

generating a measurement value based on both the characteristic value and the signal value, an

Based on the measurements, at least one of: stopping polishing the substrate; or adjusting polishing parameters.

2. The system of claim 1, wherein the machine vision process comprises an artificial neural network.

3. The system of claim 2, wherein the controller is configured to train the artificial neural network by back propagation using training data comprising images and known feature values of the images.

4. The system of claim 2, wherein the first in-situ monitoring system comprises a spectral monitoring system to generate a measured spectrum for the measurement point.

5. The system of claim 4, wherein the artificial neural network is configured to determine a classification of a portion of the substrate corresponding to the measurement point, the classification corresponding to a type of structure on the substrate.

6. The system of claim 5, wherein the structure type includes at least one of an array, a scribe line, a periphery, and a contact pad.

7. The system of claim 2, wherein the first in-situ monitoring system comprises an eddy current monitoring system to generate signal values for the measurement points.

8. The system of claim 7, wherein the artificial neural network is configured to determine a geometric value of a feature affecting a current in the measurement point.

9. The system of claim 8, wherein the geometric value comprises at least one of a distance, a size, or an orientation.

10. A computer program product for controlling processing of a substrate, the computer program product tangibly embodied in a non-transitory computer-readable medium and containing instructions for causing a processor to:

receiving a signal value from a first in-situ monitoring system, the signal value being dependent on a thickness of a layer in a measurement point on a substrate undergoing polishing;

receiving image data for at least the measurement point of the substrate from a second in situ imaging system;

determining feature values of the measurement points based on the image using machine vision processing;

generating a measurement value based on both the characteristic value and the signal value, an

Based on the measurements, at least one of: stopping polishing the substrate; or adjusting polishing parameters.

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

12. The computer program product of claim 10, comprising instructions for synchronizing image data from the second in situ imaging system with signal values from the first in situ monitoring system.

13. The computer program product of claim 10, wherein machine vision processing comprises feeding the image data to an artificial neural network.

14. The computer program product of claim 10, wherein the instructions for performing machine vision processing comprise instructions for determining a classification of a portion of the substrate corresponding to the measurement point, and wherein the instructions for producing the measurement value comprise instructions for selecting one of a plurality of reference spectrum libraries based on the classification.

15. The computer program product of claim 10, wherein the instructions for performing machine vision processing include instructions for determining a geometric value of a feature in a portion of the substrate corresponding to the measurement point.

Technical Field

The present disclosure relates to optical monitoring of substrates, for example, during processing such as chemical mechanical polishing.

Background

Integrated circuits are typically formed on a substrate by the sequential deposition of conductive, semiconductive, or insulative layers onto a silicon wafer. One fabrication step involves depositing a filler layer on a non-planar surface and planarizing the filler layer. For some applications, the filler layer is planarized until the top surface of the patterned layer is exposed. For example, a conductive filler layer can 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 raised 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, the deposited dielectric layer may be planarized for photolithography.

Chemical Mechanical Polishing (CMP) is an accepted planarization method. This planarization method typically requires that the substrate be mounted on a carrier head. The exposed surface of the substrate is typically held against a rotating polishing pad having a durable roughened surface. The carrier head provides a controllable load on the substrate to push 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 to use 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 that has removed a desired amount of material. Variations in the initial thickness of the substrate layer, slurry distribution, polishing pad conditions, relative velocity between the polishing pad and the substrate, and load on the substrate can result in variations in the material removal rate across the substrate and between substrates. These variations result in variations in the time required to reach the polishing endpoint and the amount of removal. Thus, it is not possible to determine the polishing endpoint as a function of polishing time alone, nor to achieve a desired profile by applying a constant pressure alone.

In some systems, the substrate is monitored in-situ during polishing, for example, by an optical monitoring system or eddy current monitoring system. Thickness measurements from the in-situ monitoring system may be used to adjust the pressure applied to the substrate, thereby adjusting the polishing rate and reducing within-wafer non-uniformity (WIWNU).

Disclosure of Invention

A polishing system, comprising: a support for holding a polishing pad; a carrier head for holding a substrate in contact with a polishing pad; a motor for producing relative movement between the support and the carrier head; a first in-situ monitoring system for generating a signal that is dependent on a thickness of a layer in a measurement spot on a substrate; a second in-situ imaging system for generating an image of at least the measurement point of the substrate at substantially the same time as the in-situ monitoring system generates a signal for the measurement point on the substrate; and a controller. The controller is configured to: receiving an image from the second in situ imaging system and determining feature values of the measurement points based on the image using machine vision processing; receiving a signal from an in situ monitoring system; generating a measurement value based on both the feature value and the signal value; and based on the measurements, performing at least one of: stopping polishing the substrate; or adjusting polishing parameters.

In another aspect, a computer program product for controlling processing of a substrate includes instructions for causing one or more processors to: receiving a signal value from a first in-situ monitoring system, the signal value being dependent on a thickness of a layer in a measurement point on a substrate undergoing polishing; receiving image data for at least the measurement point of the substrate from a second in situ imaging system; determining feature values of the measurement points based on the image using machine vision processing; generating a measurement value based on both the feature value and the signal value; and based on the measurements, performing at least one of: stopping polishing the substrate; or adjusting polishing parameters.

Implementations may include one or more of the following features.

Machine vision processing may include processing images 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 comprising the images and known feature values of the images.

The first in-situ monitoring system may include a spectral monitoring system to generate a measured spectrum for the measurement point. The artificial neural network may be configured to determine a classification of a portion of the substrate corresponding to the measurement point, the classification may correspond to a type of structure on the substrate. The structure type may include at least one of an array, a scribe line, a periphery, and a contact pad. One of the plurality of reference spectral libraries may be selected based on the classification.

The first in-situ monitoring system may include an eddy current monitoring system to generate signal values for the measurement points. The artificial neural network may be configured to determine a geometric value of a feature affecting 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.

Certain implementations may have one or more of the following advantages. The process control techniques may be directed to performance sensitive portions of the die. The thickness of a layer on a substrate may be measured more accurately and/or more quickly. Within-wafer thickness non-uniformity and wafer-to-wafer thickness non-uniformity (WIWNU and WTWNU) may be reduced and the reliability of the endpoint system for detecting a desired process endpoint may be improved. post-CMP metrics may be based on yield and/or performance sensitive portions of the product, rather than an average die thickness (which may include die area unrelated to product performance).

The details of one or more embodiments are set forth in the accompanying drawings and the description below. Other features, aspects, and advantages will become apparent from the description, the drawings, and the claims.

Drawings

Fig. 1 shows a schematic cross-sectional view of an example of a polishing apparatus.

FIG. 2A is a schematic diagram of an in-situ optical monitoring system.

FIG. 2B is a schematic view of an in situ eddy current monitoring system.

Fig. 3 is a schematic top view of the polishing apparatus.

Figure 4 shows a schematic diagram of a line scan imaging system.

Figure 5 shows a neural network used as part of the controller of the polishing apparatus.

Fig. 6 shows a graph of the measured values over time.

Like reference numbers and designations in the various drawings indicate like elements.

Detailed Description

Various techniques (e.g., eddy current monitoring and optical monitoring) may be used to monitor the substrate during processing. Such monitoring techniques may be performed in a two-stage manner. First, the raw signal from the monitoring system (e.g., the measured spectrum from a spectrophotometer or the voltage from an eddy current monitoring system) is converted into a more useful form of measurement (e.g., an indicator or thickness value that indicates progress through polishing). The sequence of measurements over time as the process progresses can then be monitored for process control. For example, a function can be fit to the measurement sequence, and the time at which the function is expected to reach a threshold can be used to trigger a polishing endpoint or to control other polishing parameters.

If the sensors of the monitoring system are swept across the substrate, measurements may be taken at different locations on the substrate. Thus, measurements may be made at different areas of the substrate, e.g., within the die rather than the scribe lines, or at different areas within the die (e.g., arrays, contact pads, etc.). These different regions may have different properties and provide different original signals. It would be useful to determine the type of region over which measurements are made in order to correctly convert the raw signal into useful measurements.

While the measured radial position may be determined, for example, due to rotational slippage of the substrate relative to the carrier head, the measured angular position on the substrate may not be known at all. Thus, commercial in-situ monitoring techniques do not take into account where the measurements are made in the die when converting the raw signals to useful measurements.

Also, while some monitoring systems perform filtering to exclude some of the original signal (e.g., exclude the spectrum based on the shape of the spectrum), such techniques do not use information from the surrounding portions of the substrate.

However, the images collected by the in situ imager may be processed by machine learning techniques (e.g., convolutional neural networks) to determine characteristics of the substrate for which measurements are being performed by another monitoring system. This characteristic may be, for example, the type of area being measured (e.g., scribe line, array, periphery), or the relative orientation and/or distance of various features (e.g., guard rings) to the measurement location. The characteristics can then be provided as input to an in situ monitoring system to affect the conversion of the raw signal to a measurement.

Fig. 1 shows an example of a polishing apparatus 100. The polishing apparatus 20 can comprise a rotatable disc-shaped platen 22 with a polishing pad 30 positioned on the rotatable disc-shaped platen 22. The table is operable to rotate about an axis 23. For example, the motor 24 may rotate the drive shaft 26 to rotate the table 22.

The polishing pad 30 may be removably secured to the platen 22, for example, by an adhesive layer. The polishing pad 30 may be a dual layer polishing pad having an outer polishing layer 32 and a softer backing layer 34. A window 36 may be formed in the polishing pad 30.

The polishing apparatus 20 can include a polishing liquid supply port 40 to dispense a polishing liquid 42, such as an abrasive slurry, onto the polishing pad 30. The polishing apparatus 20 may further comprise a polishing pad conditioner for dressing the polishing pad 30 to maintain the polishing pad 30 in a uniform abrasive state.

The carrier head 50 is operable to hold the substrate 10 against the polishing pad 30. Each carrier head 50 also includes a plurality of independently controllable pressurizable chambers, such as three chambers 52a-52c, which chambers 52a-52c may apply independently controllable pressurization to associated zones on substrate 10. The central region on the substrate may be substantially circular and the remaining regions may be concentric annular regions surrounding the central region.

The chambers 52a-52c may be defined by a flexible membrane 54, the flexible membrane 54 having a bottom surface to which the substrate 10 is mounted. The carrier head 50 may also include a retaining ring 56 to hold the substrate 10 under the flexible membrane 54. Although fig. 1 shows only three chambers for ease of illustration, there may be a single chamber, two chambers, or four or more chambers, e.g., five chambers. In addition, other mechanisms may be used in carrier head 50 to adjust the pressure applied to the substrate, such as piezoelectric actuators.

Each carrier head 50 is suspended from a support structure 60 (e.g., a turntable or track) and is connected by a drive shaft 62 to a carrier head rotation motor 64 so that the carrier head can rotate about an axis 51. Alternatively, each carrier head 50 may oscillate laterally (e.g., on a slide on a turntable) by movement along a track or by rotational oscillation of the turntable itself. In operation, the platen 22 rotates about the central axis 23 of the platen 22 and the carrier head 50 rotates about the central axis 51 of the carrier head 50 and translates laterally across the top surface of the polishing pad 30.

The polishing apparatus further includes a first in-situ monitoring system 100 and a second in-situ imaging system 150. The in-situ monitoring system 150 and the in-situ imaging system 150 may be used together to control polishing parameters, such as the pressure in one or more of the chambers 52a-52c, and/or to detect the polishing endpoint and stop polishing.

The first in-situ monitoring system 100 includes a sensor 100a (see FIG. 3), the sensor 100a generating a raw signal that is dependent on the thickness of the layer being polished. The first in-situ monitoring system 100 may be, for example, an eddy current monitoring system or an optical monitoring system, such as a spectroscopic monitoring system.

The sensor may be configured to sweep across the substrate. For example, the sensor may be fixed to the table 22 and rotate with the table 22 such that with each rotation of the table, the sensor sweeps across the substrate in an arc.

Referring to FIG. 2A, as an optical monitoring system, the first in-situ monitoring system 100 may include a light source 102, a light detector 104, and circuitry 106, the circuitry 106 for sending and receiving signals between the controller 90 (e.g., a computer) and the light source 102 and the light detector 104. One or more optical fibers may be used to transmit light from the light source 102 to the window 36 and to transmit light reflected from the substrate 10 to the detector 104. For example, a bifurcated optical fiber 108 may be used to transmit light from the light source 102 to the window 36 and back to the detector 104. In this implementation, the ends of the bifurcated optical fibers 108 may provide sensors that sweep across the substrate. If the optical monitoring system is a spectroscopic system, the light source 102 may be operable to emit white light and the detector 104 may be a spectrometer.

Referring to FIG. 2B, as an eddy current monitoring system, the first in-situ monitoring system 100 may include a magnetic core 112 and at least one coil 114 wound around a portion of the core 114. Drive and sense circuitry 116 is electrically connected to coil 114. The drive and sense circuitry 116 may apply an AC current to the coil 114, the coil 114 generating a magnetic field between the two poles of the core 112, which may enter the substrate 10. In this implementation, the core 112 and coil 114 may provide a sensor that sweeps across the substrate. Circuitry 116 may include a capacitor connected in parallel with coil 114. The coil 114 and the capacitor may together form an LC resonant tank (resonant tank). When the magnetic field reaches the conductive layer, the magnetic field 150 may pass through and generate a current (if the layer is a loop) or generate an eddy current (if the layer is a sheet). This modifies the effective impedance of the LC circuit. The drive and sense circuitry 116 may detect the change in effective impedance and generate a signal that may be sent to the controller 90.

In either case, the output of circuitry 106 or circuitry 116 may be a digital electronic signal that is passed through a rotating coupling 28 (e.g., slip ring) in drive shaft 26 to controller 90 (see fig. 1). Alternatively, circuitry 106 or circuitry 116 may communicate with controller 90 via wireless signals. Some or all of circuitry 106 or circuitry 116 may be mounted in table 22.

The controller 90 may be a computing device, such as a programmable computer, including a microprocessor, memory, and input/output circuitry. Although shown as a single block, the controller 90 may be a networked system with functions distributed over multiple computers.

The controller 90 may be considered to provide part of the first monitoring system, as the controller 90 may perform part of the processing of the signals (e.g., converting the "raw" signals into usable measurements).

As shown in FIG. 3, due to the rotation of the stage (indicated by arrow A), the first in-situ monitoring system takes measurements at the sampling frequency as the sensor 100a travels under the carrier head. As a result, measurements are taken at locations 94 in the arc across the substrate 10 (the number of points is illustrative; depending on the sampling frequency, more or fewer measurements than shown may be taken). The substrate may also rotate (indicated by arrow B) and vibrate radially (indicated by arrow C).

The polishing system 20 can include a position sensor 96, such as an optical interrupter, to sense when the sensor 100a of the first in-situ monitoring system 100 is below the substrate 10 and when the sensor 100a is off the substrate 10. For example, the position sensor 96 may be mounted in a fixed position relative to the carrier head 70. The markings 98 may be attached to the periphery of the table 22. The attachment point and length of the marker 98 are selected so that the marker 98 can send a signal to the position sensor 96 as the sensor 100a sweeps beneath the substrate 10.

Alternatively or additionally, polishing system 20 can include an encoder to determine the angular position of table 22.

In one revolution of the stage, spectra are obtained from different positions on the substrate 10. Specifically, some spectra may be obtained from a location closer to the center of the substrate 10, and some spectra may be obtained from a location closer to the edge. The controller 90 may be configured to calculate the radial position (relative to the center of the substrate 10) from each measurement of the scan based on timing, motor encoder information, stage rotation or position sensor data, and/or optical detection of the edge of the substrate and/or retaining ring. Thus, the controller may associate various measurements with various regions on the substrate. In some implementations, the measured time may be used as an alternative to an accurate calculation of the radial position.

The in-situ imaging system 150 is positioned to generate an image of substantially the same portion of the substrate 10 being measured by the first in-situ monitoring system 100. In short, the camera of the imaging system is co-located with the sensor of the in-situ monitoring system 100.

Referring to FIG. 4, in situ imaging system 150 may include a light source 152, a light detector 154, and circuitry 156, with circuitry 156 for sending and receiving signals between controller 90 and light source 152 and light detector 154.

The light source 152 may be operable to emit white light. In one implementation, the emitted white light includes light having a wavelength of 200 and 800 nanometers. Suitable light sources are arrays of white Light Emitting Diodes (LEDs), or xenon or mercury xenon lamps. The light source 152 is oriented to direct light 158 onto the exposed surface of the substrate 10 at a non-zero angle of incidence α. The angle of incidence α may be, for example, about 30 ° to 75 °, e.g., 50 °.

The light source 152 may illuminate an elongated region that is substantially linear. The elongated regions may span the width of the substrate 10. The light source 152 may include optics (e.g., a beam expander) to spread the light from the light source into the elongated region. Alternatively or additionally, the light source 152 may comprise a linear array of light sources. The light source 152 itself, as well as the area illuminated on the substrate, may be elongated and have a longitudinal axis parallel to the substrate surface.

The diffuser 160 may be placed in the path of the light 168 or the light source 162 may include a diffuser to diffuse the light before it reaches the substrate 10.

The detector 154 is a camera, such as a color camera, that is sensitive to light from the light source 152. The camera includes an array of detector elements. For example, the camera may include a CCD array. In some implementations, the array is a single row of detector elements. For example, the camera may be a line scan camera. The row of detector elements may extend parallel to the longitudinal axis of the elongate area illuminated by the light source 152. Where the light source 165 includes a row of light-emitting elements, the row of detector elements may extend along a first axis that is parallel to the longitudinal axis of the light source 152. A row of detector elements may contain 1024 or more elements.

The detector 154 is provided with suitable focusing optics 162 to project the field of view of the substrate onto the array of detector elements of the detector 154. The field of view may be long enough to view the entire width of the substrate 10, e.g., 150mm to 300mm long. The detector 164 may also be configured such that the pixel width is comparable to the pixel length. For example, an advantage of a line scan camera is its very fast frame rate. The frame rate may be at least 5 kHz. The frame rate may be set to a frequency such that the pixel width is comparable to the pixel length, e.g., equal to or less than about 0.3 millimeters, when the imaged area is scanned across the substrate 10.

The light source 162 and the light detector 164 may be supported in a recess in the table, for example, in the same recess that holds the sensors of the first in-situ monitoring system 100.

A possible advantage of having a line scan camera and light source that move together across the substrate is that the relative angle between the light source and the camera remains constant for different positions across the wafer, for example, as compared to a conventional 2D camera. Thus, artifacts caused by viewing angle changes may be reduced or eliminated. In addition, the line scan camera can eliminate perspective distortion, while the conventional 2D camera shows intrinsic perspective distortion, which needs to be corrected by image transformation.

Optionally, a polarizing filter 164 may be positioned in the light path, for example between the substrate 10 and the detector 154. The polarizing filter 164 may be a Circular Polarizer (CPL). A typical CPL is a combination of a linear polarizer and a quarter-wave plate. Proper orientation of the polarization axis of the polarizing filter 164 may reduce haze in the image and sharpen or enhance desired visual features.

The controller 90 assembles the individual image lines from the light detector 154 into a two-dimensional image. The camera 164 may be a color camera having separate detector elements, for example, for each of the colors red, blue, and green, in which case the controller 90 assembles the individual image lines from the light detector 154 into a two-dimensional color image. The two-dimensional color image may include a monochrome image 204, 206, 208 for each color channel (e.g., for each of a red, blue, and green channel).

Referring to FIG. 5, the controller 90 may convert the raw signals from the in-situ monitoring system into useful measurements. The controller 90 uses both the signal from the first in situ monitoring system 100 and the image data from the second in situ imaging system 150 to calculate the measurements. The images collected from the in situ imaging system 150 may be synchronized with the data stream collected from the first in situ monitoring system 100.

In particular, the controller 90 feeds images from the in-situ imaging system 150 to the machine vision system 200, the machine vision system 200 being configured to derive characteristic values of the substrate portion measured by the first in-situ monitoring system 100. The machine vision system may include, for example, a neural network 210. The neural network 210 may be a convolutional neural network.

The neural network 210 includes a plurality of input nodes 212, such as an input node 212 for each pixel in the image from the in situ imaging system 150. These input nodes 212 may include input node N₁、N₂……N_L. The neural network 210 further includes a plurality of hidden nodes 214 (hereinafter also referred to as "intermediate nodes") and at least one output node 216, the at least one output node 216 to generate at least one characteristic value.

In general, the value output by the hidden node 214 is a non-linear function of a weighted sum of values from nodes connected to the hidden node.

For example, the output of the hidden node 214 designated as node k may be represented as:

wherein tanh is the hyperbolic tangent, a_kxIs the connection weight between the kth intermediate node and the xth input node (among the M input nodes), and I_MIs the value at the mth input node. However, other non-linear functions may be used instead of tanh, such as a rectified linear unit (ReLU) function and its variants.

The architecture of the neural network 210 may vary in depth and width. Although the neural network 210 is shown with one column of intermediate nodes 214, in practice the neural network will include many columns, which may have various connections. The convolutional neural network may perform multiple iterations of convolution and pooling, followed by classification.

The neural network 210 may be trained, for example, in a training mode using backpropagation with sample images and sample feature values. Thus, in operation, the machine vision system 200 generates feature values based on images from the in situ imaging system 150. This may be performed for each value of the "raw signal" received from in situ monitoring system 100.

The raw signal from the in-situ monitoring system 100 and the feature values 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 a measurement value from the feature value and the raw signal.

The measurement is typically the thickness of the outer layer, but may be a related property, such as the removed thickness. In addition, the measurement value may be a more general representation of the progress of the substrate through the polishing process, such as an index value representing the time or number of table rotations for which the measurement value is expected to be observed in the polishing process following a predetermined progress.

The measurements may be fed to the process control subsystem 240 to adjust the polishing process based on a series of characteristic values, for example, to detect a polishing endpoint and stop polishing and/or to adjust a polishing pressure during the polishing process to reduce polishing non-uniformities. The process control module 240 may output process parameters, such as a pressure for a chamber in the carrier head and/or a signal to stop polishing.

For example, referring to fig. 6, a first function 254 may be fitted to the sequence 250 of measured values 252 for a first region, and a second function 264 may be fitted to the sequence 260 of feature values 262 for a second region. The process controller 240 may calculate the times T1 and T2 at which the first and second functions are expected to reach the target value V and calculate an adjusted process parameter (e.g., an adjusted carrier head pressure) that will cause one of the zones to be polished (indicated by line 270) at a modified rate such that the zones reach the target at about the same time.

The polishing endpoint may be triggered by the process controller 240 when the function indicates that the characteristic value reaches the target value V.

In some implementations, multiple measurements may be combined at the conversion algorithm module 220 or the process control module 240. For example, if the system generates multiple measurements from a single scan of the sensor across the substrate, the conversion algorithm module 220 may combine the multiple measurements from the single scan to generate a single measurement per scan or a single measurement per radial region on the substrate. However, in some implementations, a measurement is generated for each position 94, and the sensor 100a generates a raw signal value for each position 94 (see fig. 3).

In some implementations, the neural network 210 generates a plurality of feature values at a plurality of output nodes 216. The one or more additional characteristic values (i.e., characteristic values other than the characteristic value representing the thickness measurement) may represent other characteristics of the substrate, such as wafer orientation, type of structures on the wafer (e.g., memory array, central processing unit). Additional feature values may be fed into the process control 240.

Example 1

The in-situ monitoring system 100 may be a spectroscopic monitoring system. The same window 36 may be used by the sensors of the spectral monitoring system and the in situ imaging system 150. A window of data from the line scan camera of the in-situ imaging system 150, centered at the time of spectral acquisition by the in-situ monitoring system 100, may be used to reconstruct a two-dimensional image of the portion of the substrate 10 from which the spectrum was collected.

The machine vision system 200 may include a Convolutional Neural Network (CNN) 210. To train the neural network 210, a series of images from one or more reference substrates may be manually identified with relevant categories (e.g., array, scribe line, periphery, contact pads, etc.). Assigning a category to an image is sometimes referred to as "annotation". The images and classes from the reference substrate may then be input to the neural network in a training mode (e.g., using back propagation) to train the neural network 210 as an image classifier. It should be noted that such image classifiers can be trained with a relatively small number of annotated images via the use of transfer learning, in which several additional images from a new domain are shown to a pre-trained image classification network.

In operation, during polishing of the product substrate, an image is fed into the neural network 210. The output of the neural network 210 is used in real time to correlate each measured spectrum with the classification of the portion of the substrate from which the spectrum was obtained.

The image classification by the convolutional neural network may be concatenated with the measured spectrum (continate) before being fed into another model for thickness estimation or prediction.

This classification may be used by the conversion algorithm module 220. For example, the controller 90 may store a plurality of libraries of reference spectra, wherein each reference spectrum has an associated measurement value, such as an indicator value. The controller 90 may select one of the libraries based on the classification received from the neural network 210. The reference system from the selected library that best matches the measured spectrum can then be determined, for example, by finding the reference spectrum with the smallest sum of squared differences relative to the measured spectrum. The index value of the best matching reference spectrum can then be used as the measurement value.

Example 2

The in-situ monitoring system 100 may be an eddy current monitoring system. The sensor 100a of the eddy current monitoring system and the sensor of the in situ imaging system 150 are co-located, for example, in the same recess in the table. The line scan camera of in-situ imaging system 150 generates a time-synchronized image that covers the entire scan of 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 geometry (e.g., location, size, and/or orientation) of substrate features (e.g., guard rings) that affect the current may be manually identified. The images and geometric values from the reference substrate may then be input to the neural network in a training mode (e.g., using back propagation) to train the neural network 210 as a feature geometry reconstructor.

In operation, during polishing of the product substrate, an image is fed into the neural network 210. The output of the neural network 210 is used in real time to correlate each measured value from the eddy current monitoring system with the geometric value of the portion of the substrate from which the spectrum was obtained.

The geometry values generated by the neural network 210 may be used by the conversion algorithm module 220. The mapping from eddy current signals to resistance depends on the relative orientation and location of features on the substrate. For example, the sensitivity of the sensor 100a to conductive loops on the substrate may depend on the orientation of the loops. The controller 90 may include a function that calculates the gain based on the geometric value (e.g., orientation). This gain may then be applied to the signal, e.g., the signal value may be multiplied by the gain. Thus, the geometric values may be used to adjust how eddy current sensor data is interpreted.

Conclusion

Embodiments of the invention and all of the functional operations described in this specification can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the structural elements disclosed in this specification and structural equivalents thereof, or in combinations of them. Embodiments of the invention can be implemented as one or more computer program products (i.e., one or more computer programs) tangibly embodied in a machine-readable storage medium for execution by, or to control the operation of, data processing apparatus (e.g., a programmable processor, a computer, or multiple processors or computers). A computer program (also known as a program, software application, or code) can be written in any form of programming language, including compiled or interpreted languages, and it 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. A program can be stored in a portion of an archive that holds other programs or data, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub programs, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a communication network.

The processes and logic flows described in this specification can be performed by one or more programmable processors executing one or more computer programs to perform functions by operating on input data and generating output. The processes and logic flows can also be performed by, and apparatus can also be implemented as, special purpose logic circuitry, e.g., 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. Either the polishing pad or the carrier head or both can be moved to provide relative motion between the polishing surface and the substrate. For example, the table may orbit rather than rotate. The polishing pad may be a circular (or some other shape) pad that is affixed to the platen. The polishing system can be a linear polishing system, for example, wherein the polishing pad is a linearly moving continuous belt or a disk-type belt (reel-to-reel belt). The polishing layer may be a standard (e.g., polyurethane with or without fillers) polishing material, a soft material, or a fixed abrasive material. Relative positioning terms are used with respect to the orientation or positioning of the components; it should be understood that the polishing surface and substrate may 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 may be applicable to other semiconductor processing techniques, such as etching or deposition, e.g. chemical vapor deposition. Instead of a line scan camera, a camera that images a two-dimensional area of the substrate may be used. In this case, it may be necessary to merge multiple images.

Specific embodiments of the present invention have been described. Other embodiments are within the scope of the following claims.

What is claimed is: