Polishing apparatus using a neural network for monitoring

Abstract

A method for polishing a substrate is provided that compensates for distortion in a portion of a signal corresponding to the substrate edge, resulting in improved within-wafer non-uniformity (WIWNU) and wafer-to-wafer non-uniformity (WTWNU). [Solution] A method for polishing a layer on a substrate in a polishing station includes monitoring the layer being polished in the polishing station using an in-situ monitoring system to generate a plurality of measurement signals at a plurality of different locations on the layer, generating an estimate of the thickness at each of the plurality of different locations, including processing the plurality of measurement signals via a neural network, and modifying polishing parameters or detecting a polishing endpoint based on each thickness estimate.

Description

Technical Field

[0001] The present disclosure relates to in-situ monitoring during substrate polishing.

Background Art

[0002] Integrated circuits are typically formed on a substrate (e.g., a semiconductor wafer) by successively depositing conductive, semiconductive, or insulating layers on the silicon wafer and then processing the layers.

[0003] A manufacturing step includes depositing a fill layer on a non-planar surface and planarizing the fill layer until the non-planarity is exposed. For example, a conductive fill layer can be deposited on a patterned insulating layer to fill trenches or holes in the insulating layer. The fill layer is then polished until the raised pattern of the insulating layer is exposed. 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. Additionally, planarization can be used to flatten the substrate surface for lithography.

[0004] Chemical mechanical polishing (CMP) is one of the recognized planarization methods. In this planarization method, it is typically necessary to mount the substrate on a carrier head. The exposed surface of the substrate is placed against a rotating polishing pad. The carrier head applies a controllable load to the substrate to press the substrate against the polishing pad. A polishing liquid such as a slurry containing abrasive grains is supplied to the surface of the polishing pad.

[0005] During semiconductor processing, it may be important to identify one or more characteristics of a substrate or a layer on the substrate. For example, it may be important to know the thickness of a conductive layer during CMP processing so that the process can be terminated at an appropriate time. A number of methods can be used to identify substrate characteristics. For example, an optical sensor can be used for in-situ monitoring of a substrate during chemical mechanical polishing. Alternatively (or additionally), an eddy current detection system can be used to induce eddy currents within a conductive region of the substrate to identify parameters such as the local thickness of the conductive region. SUMMARY OF THE INVENTION

[0006] In one aspect, a method of polishing a layer on a substrate at a polishing station includes using an in-situ monitoring system to monitor the layer being polished at the polishing station to generate a plurality of measurement signals for a plurality of different locations on the layer, generating a plurality of estimated values for the plurality of different locations on the layer, including processing the plurality of measurement signals via a neural network, and at least one of modifying a polishing parameter or detecting a polishing endpoint based on each estimated value of the thickness.

[0007] In another aspect, a corresponding computer system, apparatus, and computer program recorded on one or more computer storage devices are configured to perform this method. One or more computer systems can be configured to perform a particular operation or action by software, firmware, hardware, or a combination thereof installed on the system such that the system can perform the action during operation. One or more computer programs are configured to perform a particular operation or action by including instructions that, when executed by a data processing apparatus, cause the apparatus to perform the operation.

[0008] In another aspect, the polishing system includes a carrier for holding a substrate, a support for a polishing surface, an in-situ monitoring system having a sensor, a motor for creating relative movement between the substrate and the sensor, and a controller. The in-situ monitoring system is configured to generate measurement signals for a plurality of different locations on the layer. The controller is configured to receive the plurality of measurement signals from the in-situ monitoring system and generate an estimated value of the thickness at each of the plurality of different locations, generating including processing the plurality of signals via a neural network, detecting a polishing endpoint, modifying polishing parameters based on each estimated value of the thickness, or both.

[0009] Implementations of any of the above aspects may include one or more of the following features.

[0010] A second plurality of measurement signals may be obtained for a second plurality of different locations on the layer. For each of the second plurality of different locations, an estimated value of the thickness at that location may be generated based on the measurement signal at that location. Generating the estimated value of the thickness may include using a static formula related to the plurality of values of the measurement signal for the plurality of values of the estimated value of the thickness. A third plurality of measurement signals is obtained for the layer of the second substrate. Each of the third plurality of measurement signals may correspond to one of a third plurality of locations on the layer on the second substrate. For each of the third plurality of different locations, an estimated value of the thickness at that location may be generated based on the measurement signal at that location. Generating the estimated value of the thickness may include using a static formula related to the plurality of values of the measurement signal for the plurality of values of the estimated value of the thickness.

[0011] The in-situ monitoring system may include an eddy current sensor. The neural network includes one or more neural network layers including an input layer, an output layer, and one or more hidden layers, and each neural network layer includes one or more neural network nodes. Each neural network node may be configured to process an input according to a set of parameters for generating an output. The input to the neural network nodes of the input layer may include a measured value of the wear of the pad of the polishing station. One or more different locations include anchor locations, and identifying each first measured value of the thickness may include normalizing each measurement signal based on the measurement signal of the anchor location to update the measurement signal. The anchor location may be spaced apart from the edge of the substrate. Each estimated value of the thickness may be a normalized value, and the method may further include an operation of converting each estimated value of the thickness into a non-normalized value using the measurement signal of the anchor location for updating the estimated value of the thickness.

[0012] A ground truth measure of the thickness is obtained for each of one or more locations of the layer. An error amount between the estimated value of the thickness for each location and the corresponding ground truth measure of the thickness of that location can be calculated. The parameters of the neural network system can be updated based on the error amount. The ground truth measure of the thickness can be specified based on the four-point probe method. Updating the parameters of the neural network system based on the error amount may include backpropagating the gradient of the measure of error through multiple layers of the neural network.

[0013] Certain implementations may include one or more of the following advantages. An in-situ monitoring system, such as an eddy current monitoring system, can generate signals as the sensor scans across the substrate. The system can compensate for some distortion of the signals corresponding to the substrate edges. The signals can be used for endpoint control and / or closed-loop control of polishing parameters such as carrier head pressure, thereby resulting in improved within-wafer non-uniformity (WIWNU) and wafer-to-wafer non-uniformity (WTWNU).

[0014] Details of one or more implementations are set forth in the accompanying drawings and the description below. Other aspects, features, and advantages will be apparent from these descriptions and drawings, as well as from the claims.

Brief Description of the Drawings

[0015]

Figure 1A

Figure 1B

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

[0016] Similar reference numerals in the various drawings refer to similar elements.

DETAILED DESCRIPTION OF THE INVENTION

[0017] The polishing apparatus can use an in-situ monitoring system, such as an eddy current monitoring system, to detect the thickness of the outer layer used for polishing the substrate. During polishing of the outer layer, the in-situ monitoring system can identify the thickness at different locations of the layer on the substrate. The measured thickness values can be used to trigger the end of polishing and / or to adjust the processing parameters of the polishing process in real time. For example, the substrate carrier head can adjust the pressure on the back side of the substrate to accelerate or decelerate the polishing rate at multiple locations of the outer layer. The polishing rate can be adjusted so that the thickness after polishing is substantially the same at multiple locations of the layer. The CMP system can adjust the polishing rate so that polishing at multiple locations of the layer is completed in approximately the same time. Such profile control is sometimes referred to as real-time profile control (RTPC).

[0018] The in-situ monitoring system may be subject to signal distortion for measurements near the substrate edge. For example, an eddy current monitoring system can generate a magnetic field. Near the substrate edge, since the magnetic field only partially overlaps with the conductive layer of the substrate, the signal can be artificially low. However, when the polishing apparatus uses a neural network to generate a corrected signal based on the measurement signal generated by the in-situ monitoring system, the apparatus can compensate for the distortion, for example, attenuate the signal intensity at the substrate edge.

[0019] Figures 1A and 1B show an example of a polishing apparatus 100. The polishing apparatus 100 includes a rotatable disk-shaped platen 120 with a polishing pad 110 positioned thereon. The platen is operable to rotate about an axis 125. For example, a motor 121 can rotate a drive shaft 124 to rotate the platen 120. The polishing pad 110 may be a two-layer polishing pad having an outer polishing layer 112 and a more flexible backing layer 114.

[0020] The polishing apparatus 100 can include a port 130 for dispensing a polishing liquid 132, such as slurry, onto the polishing pad 110. The polishing apparatus can also include a polishing pad conditioner for abrading the polishing pad 110 to maintain the polishing pad 110 in a consistent polishing state.

[0021] The polishing apparatus 100 includes at least one carrier head 140. The carrier head 140 is operable to hold a substrate 10 against the polishing pad 110. The carrier head 140 can individually control polishing parameters associated with each substrate, such as pressure.

[0022] Specifically, each carrier head 140 can include a retaining ring 142 that holds the substrate 10 under a flexible membrane 144. The carrier head 140 also includes a plurality of individually controllable and pressurizable chambers (e.g., three chambers 146a - 146c) defined by the membrane, and these chambers can apply individually controllable pressure to associated zones on the flexible membrane 144 (and thus on the substrate 10). For simplicity of explanation, only three chambers are illustrated in FIG. 1, but there may be one or two chambers, or four or more chambers, e.g., five chambers.

[0023] The carrier head 140 is suspended from a support structure 150, such as a carousel or track, so that the carrier head can rotate about the axis 155 and is connected by a drive shaft 152 to a carrier head rotation motor 154. Optionally, the carrier head 140 can vibrate laterally, for example, by a slider on the carousel 150 or track, or by the rotational vibration of the carousel itself. During operation, the platen rotates about its central axis 125, the carrier head rotates about its central axis 155, and translates laterally parallel from end to end across the upper surface of the polishing pad.

[0024] Only one carrier head 140 is shown, but more carrier heads can be provided to hold additional substrates, thereby enabling the surface area of the polishing pad 110 to be used efficiently.

[0025] The polishing apparatus 100 includes one in-situ monitoring system 160. The in-situ monitoring system 160 generates a sequence of values that vary over time and depend on the thickness of the layer on the substrate. The in-situ monitoring system 160 includes a sensor head at which measurements are generated, and the measurements are made at different locations on the substrate by the relative movement between the substrate and the sensor head.

[0026] The in-situ monitoring system 160 can be an eddy current monitoring system. The eddy current monitoring system 160 includes a drive system that induces eddy currents in a conductive layer on the substrate and a sensing system that detects the eddy currents induced in the conductive layer by the drive system. The monitoring system 160 includes a core 162 disposed in a recess 128 that rotates with the platen, at least one coil 164 wound around a portion of the core 162, and a drive and sensing circuit 166 connected to the coil 164 by wiring 168. The core 162 and the coil 164 can provide a sensor head. In some implementations, the core 162 projects above the upper surface of the platen 120, for example, into a recess 118 in the bottom of the polishing pad 110.

[0027] The drive and detection circuit 166 is configured to apply a vibrating electrical signal to the coil 164 and measure the resulting eddy current. For example, as described in U.S. Patent Nos. 6,924,641, 7,112,960, and 8,284,560, and U.S. Patent Application Publication Nos. 2011-0189925 and 2012-0276661, various configurations are possible for the drive and detection circuit, as well as for the configuration and position of the coil. The drive and detection circuit 166 can be disposed in the same recess 128 or in different portions of the platen 120, or can be disposed outside the platen 120 and can be coupled to components of the platen by a rotary electrical coupling 129.

[0028] During operation, the drive and detection circuit 166 drives the coil 164 to generate a vibrating magnetic field. At least a portion of the magnetic field extends through the polishing pad 110 to the substrate 10. When a conductive layer is present on the substrate 10, the vibrating magnetic field generates eddy currents in the conductive layer. The eddy currents cause the conductive layer to act as an impedance source coupled to the drive and detection circuit 166. When the thickness of the conductive layer changes, the impedance changes, which can be detected by the drive and detection circuit 166.

[0029] Alternatively or additionally, an optical monitoring system that can function as a reflectometer or an interferometer can be fixed to the platen 120 of the recess 128. When both systems are used, the optical monitoring system and the eddy current monitoring system can monitor the same portion of the substrate.

[0030] The CMP apparatus 100 can also include a position sensor 180, such as a light interrupter, to detect when the core 162 is under the substrate 10. For example, the light interrupter can be mounted at a fixed position on the opposite side of the carrier head 140. A flag 182 is attached to the periphery of the platen. The attachment position and length of the flag 182 are selected such that the flag 182 blocks the light signal of the sensor 180 while the core 162 passes under the substrate 10. Alternatively or additionally, the CMP apparatus can include an encoder for specifying the angular position of the platen.

[0031] A controller 190, such as a general-purpose programmable digital computer, receives an intensity signal from the eddy current monitoring system 160. The controller 190 may include a processor, a memory, an I / O device, and an output device 192 such as a monitor and an input device 194 such as a keyboard.

[0032] The signal can pass from the eddy current monitoring system 160 to the controller 190 via the rotary electrical coupling 129. Alternatively, the circuit 166 can communicate with the controller 190 by wireless signals.

[0033] Since the sensor 162 passes under the substrate with each rotation of the platen, information on the thickness of the conductive layer is accumulated in situ and on a continuous real-time basis (once per rotation of the platen). The controller 190 can be programmed to sample the measured values from the monitoring system when the substrate is generally over the core 162 (as identified by the position sensor). As polishing progresses, the thickness of the conductive layer changes and the sampled signal also varies with time. The time-varying sampled signal may be referred to as a trace. During polishing, the measured values from the monitoring system are displayed on the output device 192 and the operator of the device can visually monitor the progress of the polishing operation.

[0034] During operation, the CMP apparatus 100 can use the eddy current monitoring system 160 to identify when the bulk of the fill layer is removed and / or when the underlying stop layer is substantially exposed. Possible process controls and endpoint criteria for the detector logic include minima or maxima, changes in slope, amplitude or slope thresholds, or combinations thereof.

[0035] Controller 190 can also be connected to a pressure mechanism that controls the pressure applied by the carrier head 140, to a carrier head rotation motor 154 for controlling the rotational speed of the carrier head, to a platen rotation motor 121 for controlling the rotational speed of the platen, or to a slurry distribution system 130 for controlling the composition of the slurry supplied to the polishing pad. Additionally, as described in U.S. Patent No. 6,399,501, the computer 190 can be programmed to separate the measurements from the eddy current monitoring system 160 and the measurements for each pass under the substrate into multiple sampling zones in order to calculate the radial position of each sampling zone and to classify the amplitude measurements for each of a plurality of radial ranges. After classifying the measurements for each radial range, information about the thickness of the film can be sent in real time to a closed-loop controller to periodically or continuously correct the profile of the polishing pressure applied by the carrier head to provide improved polishing uniformity.

[0036] Controller 190 can use a correlation curve that associates the signal measured by the in-situ monitoring system 160 with the thickness of the layer being polished on the substrate 10 to generate an estimate of the thickness of the polished layer. An example of a correlation curve 303 is shown in FIG. 3. In the coordinate system depicted in FIG. 3, the horizontal axis represents the value of the signal received from the in-situ monitoring system 160, while the vertical axis represents the value for the thickness of the layer of the substrate 10. For a given signal value, the controller 190 can use the correlation curve 303 that generates the corresponding thickness value. The correlation curve 303 can be considered a "static" formula in that it predicts the thickness value for each signal value regardless of the time or position at which the sensor head acquired the signal. The correlation curve can be represented by various functions such as a polynomial function or a look-up table (LUT) associated with linear interpolation.

[0037] Referring to FIGS. 1B and 2, a change in the position of the sensor head relative to the substrate 10 may result in a change in the signal of the in-situ monitoring system 160. That is, as the sensor head scans across the substrate 10, the in-situ monitoring system 160 measures a plurality of regions 94 (e.g., measurement spots at different locations on the substrate 10). The regions 94 may partially overlap (see FIG. 2).

[0038] FIG. 4 is a graph 420 showing one signal 401 from the in-situ monitoring system 160 while the sensor head passes under the substrate 10 once. The signal 401 is composed of a series of individual measurement values from the sensor head as it passes under the substrate. The graph 420 can be a function of the measurement time or position (e.g., the radial position of the measurement on the substrate). In either case, different portions of the signal 401 correspond to measurement spots 94 at different positions on the substrate 10 scanned by the sensor head. Thus, the graph 420 shows the corresponding measurement signal values from the signal 401 for a given location on the substrate scanned by the sensor head.

[0039] Referring to FIGS. 2 and 4, the signal 401 includes a first portion 422 corresponding to a location within the edge region 203 of the substrate 10 when the sensor head intersects the front edge of the substrate 10, a second portion 424 corresponding to a location within the central region 201 of the substrate 10, and a third portion 426 corresponding to a location within the edge region 203 when the sensor head intersects the rear edge of the substrate 10. The signal may also include a portion 428 corresponding to a measurement value outside the substrate, i.e., a signal generated when the sensor head scans a region beyond the edge 204 of the substrate 10 in FIG. 2.

[0040] The edge region 203 may correspond to the portion of the substrate where the measurement spot 94 of the sensor head overlaps the substrate edge 204. The central region 201 may include an annular anchor region 202 adjacent to the edge region 203 and an internal region 205 surrounded by the anchor region 202. The sensor head scans these regions on its path 210 and generates a series of measurement values corresponding to a series of locations along the path 210.

[0041] In the first portion 422, the signal intensity rises from an initial intensity (typically, the signal that occurs when the substrate and the carrier head are absent) to a high intensity. This is caused by moving from a monitoring location that initially overlaps the substrate only slightly at the edge 204 of the substrate (generating an initial low value) to a monitoring location where almost the entire substrate overlaps (generating a high value). Similarly, in the third portion 426, the signal intensity decreases when the monitoring location moves from the edge 204 to the substrate.

[0042] The second portion 424 is depicted flat for simplicity, but the actual signal in the second portion 424 may include variations due to noise and changes in the layer thickness. The second portion 424 corresponds to the monitoring locations that scan the central region 201. The second portion 424 includes sub - portions 421 and 423 caused by the monitoring locations that scan the anchor region 202 of the central region 201, and a sub - portion 427 caused by the monitoring locations that scan the internal region 205 of the central region 201.

[0043] As described above, the fluctuations in signal strength in regions 422 and 426 are not due to inherent fluctuations in the thickness or conductivity of the layer being monitored, but are partially caused by the measurement regions of the sensors that overlap the substrate edge. As a result, this distortion of signal 401 can cause errors in calculating characteristic values of the substrate, such as the thickness of the layer near the substrate edge. To address this problem, controller 190 may include a neural network (e.g., neural network 500 of FIG. 5) that generates a correction signal corresponding to one or more locations on substrate 10 based on the measurement signals corresponding to these locations.

[0044] Referring to FIG. 5, neural network 500 is configured to reduce and / or remove the distortion of the calculated signal values near the substrate edge when appropriately trained. Neural network 500 receives a set of inputs 504, processes the inputs 504 through one or more neural network layers, and generates a set of outputs 550. The layers of neural network 500 include an input layer 510, an output layer 530, and one or more hidden layers 520.

[0045] Each layer of neural network 500 includes one or more neural network nodes. Each neural network node in a neural network layer receives one or more input values (from the inputs 504 to neural network 500 or from the outputs of one or more nodes in a preceding neural network layer), processes the node input values according to one or more parameter values to generate an activation value, and optionally applies a non-linear transformation function (e.g., a sigmoid function or a tanh function) to the activation value to generate the output of the neural network node.

[0046] Each node in input layer 510 receives one of the inputs 504 to neural network 500 as the node input value.

[0047] The input 504 of the neural network includes measurement signal values of the in-situ monitoring system 160 for a plurality of different locations on the substrate 10, up to the first measurement signal value 501, the second measurement signal value 502, and the Nth measurement signal value 503. The measurement signal value can be an individual value among a series of values of the signal 401.

[0048] Generally, the plurality of different locations include a plurality of locations in the edge region 203 and the anchor region 202 of the substrate 10. In some implementations, the plurality of different locations are only within the edge region 203 and the anchor region 202. In other implementations, the plurality of different locations extend over the entire area of the substrate.

[0049] These measurement signal values are received at the signal input node 544. Optionally, the input node 504 of the neural network 500 can also include one or more state input nodes 516 that receive one or more process state signals 504 (e.g., a measurement value of the wear of the pad 110 of the polishing apparatus 100).

[0050] The nodes of the hidden layer 520 and the output layer 530 are depicted as receiving inputs from all nodes of the preceding layer. This is the case of a fully connected feedforward network. However, the neural network 500 can be a partially connected feedforward neural network or a non-feedforward neural network. Moreover, the neural network 500 can include at least one of one or more fully connected feedforward layers, one or more partially connected feedforward layers, and one or more non-feedforward layers.

[0051] The neural network generates a set of correction signal values 550 at the nodes of the output layer 530, i.e., the "output nodes" 550. In some implementations, there is an output node 550 for each measurement signal from the in-situ monitoring system input to the neural network 500. In this case, the number of output nodes 550 can correspond to the number of signal input nodes 504 of the input layer 510.

[0052] For example, the number of signal input nodes 544 may be equal to the number of measurement values in the edge region 203 and the anchor region 202, and may be the same as the number of output nodes 550. Accordingly, each output node 550 generates a correction signal corresponding to each measurement signal supplied as an input to the signal input node 544 (e.g., a first correction signal 551 for the first measurement signal 501, a second correction signal 552 for the second measurement signal 502, and an Nth correction signal 553 for the Nth measurement signal 503).

[0053] In some implementations, the number of output nodes 550 is less than the number of input nodes 504. In some implementations, the number of output nodes 550 is less than the number of signal input nodes 544. For example, the number of signal input nodes 544 may be equal to the number of measurement values in the edge region 203, or may be equal to the number of measurement values in the edge region 203 and the anchor region 202. Again, each output node 550 in the output layer 530 generates a correction signal corresponding to each measurement signal supplied as the signal input node 504 (e.g., a first correction signal 551 for the first measurement signal 501, but only for the signal input node 554 that receives a signal from the edge region 203).

[0054] The polishing apparatus 100 can use the neural network 500 to generate correction signals. The correction signals can then be used to specify the thickness for each location within a first group of locations on the substrate, e.g., locations within the edge region (and, optionally, the anchor region). For example, referring back to FIG. 4, the correction signal value for the edge region can provide the corrected portion 430 of the signal 401.

[0055] The corrected signal value 430 can be converted into a thickness measurement value using a static formula, for example, a correlation curve. For example, the controller 190 can use the neural network 500 to identify the thickness of the edge locations and one or more anchor locations of the substrate. Optionally, the controller 190 can directly use a static formula to generate a thickness measurement value for other regions, such as the internal region 205. That is, the signal values of other regions, such as the internal region 205, can be converted into thickness values without being converted by the neural network.

[0056] In some implementations, for the corrected signal value corresponding to a predetermined measurement location, the neural network 500 can be configured such that only the signal values of the measurement locations within a range of a predetermined distance from the predetermined location are used in the determination of the corrected signal value. For example, for the signal values S1, S2, …, S M 、…S N corresponding to the measurement values at N consecutive locations on the path 210, when N is received, for the corrected signal value S’ M corresponding to the Mth location (shown as R M ), S M-L(min 1) 、…S M 、…S M+L(max N) only can be used to calculate the corrected signal value S’ M . The value of L is such that measurement values up to about 2 - 4 mm apart are used in the generation of a predetermined corrected signal value S’ M , and can be selected such that measurement values within a range of about 1 - 2 mm (e.g., 1.5 mm) from the location of the measurement value S M can be used. For example, L can be a number in the range from 0 to 4 (e.g., 1 or 2). For example, if measurement values within 3 mm are used and the interval between measurement values is 1 mm, then L is 1, if the interval is 0.5 mm, then L is 2, and if the interval is 0.25 mm, then L can be 4. However, this depends on the configuration and processing conditions of the polishing apparatus. Other parameters (such as pad wear) can still be used in the calculation of the corrected signal value S’ M .

[0057] For example, there may be a number of hidden nodes 570 (i.e., "hidden nodes" 570) in one or more hidden layers 520 equal to the number of signal input nodes 544, and each hidden node 570 corresponds to a respective signal input node 544. Each hidden node 570 may be separated from (or may have a parameter value of 0) the input node 544 corresponding to the measurement value at a location that is a predetermined distance or more away from the measurement location of the corresponding input node. For example, the Mth hidden node may be separated from the input nodes 544 from the first to the (M - L - 1)th, and from the (M + L + 1)th to the Nth input nodes. Similarly, each output node 560 may be separated from (or may have a parameter value of 0) the hidden node 570 corresponding to the correction signal at a location that is a predetermined distance or more away from the measurement location of the output node. For example, the Mth output node may be separated from the hidden nodes 570 from the first to the (M - L - 1)th, and from the (M + L + 1)th to the Nth hidden nodes.

[0058] In some embodiments, the polishing apparatus 100 can use a static formula to specify the thickness at a plurality of locations (e.g., locations within the edge region of the first group of substrates). These substrates can be used for generating data used in the training of the neural network. Next, the polishing apparatus 100 can use the neural network 500 that generates the correction signal used to specify the thickness at a plurality of locations (e.g., locations within the edge region of the second group of substrates). For example, the polishing apparatus 100 can apply a static formula to specify the value of the thickness of the first group of substrates, and use the trained neural network 500 to generate the correction signal used to specify the value of the thickness of the second group of substrates.

[0059] FIG. 6 is a flowchart of an exemplary process 600 for polishing the substrate 10. The process 600 can be executed by the polishing apparatus 100.

[0060] The polishing apparatus 100 polishes (602) the layer on the substrate 10 and monitors (604) the layer during polishing to generate measurement signal values for different locations on the layer. The locations on the layer may include one or more locations within the edge region 203 of the substrate (corresponding to regions 422 / 426 of signal 401), and one or more locations within the anchor region 202 of the substrate (corresponding to regions 421 / 423 of the signal). The anchor region 202 is spaced from the substrate edge 204 and is within the central region 201 of the substrate, and thus is not affected by the distortion created by the substrate edge 204. However, the anchor region 202 may be adjacent to the edge region 203. The anchor region 202 may also surround the inner region 205 of the central region 201. The number of anchor locations may depend on the measurement spot size and measurement frequency by the in-situ monitoring system 160. In some embodiments, the number of anchor locations cannot exceed a maximum value (e.g., a maximum value of 4).

[0061] Based on the measurement signals of the locations, the polishing apparatus 100 generates (606) an estimated value of the thickness for each of the different locations. This includes processing the measurement signals via the neural network 500.

[0062] The input to the neural network 500 can be the raw measurement signals generated by the in-situ monitoring system 160 for different locations, or the updated measurement signals. In some embodiments, the apparatus 100 updates each measurement signal by normalizing the value of the signal. Such normalization increases the likelihood that at least some of the inputs 504 to the neural network system 500 fall within a specific range, and can then improve the quality of the training of the neural network and / or the accuracy of the inference performed by the neural network 500.

[0063] The outputs of the neural network 500 are correction signals, each corresponding to an input measurement signal. When the measurement signal is a normalized value, the correction signal corresponding to the measurement signal is also a normalized value. Therefore, the polishing apparatus 100 may need to convert such a correction signal to a non-normalized value before using the correction signal for estimating the thickness of the substrate.

[0064] Based on each estimated value of the thickness, the polishing apparatus 100 corrects the polishing parameters and / or detects (608) the polishing end point.

[0065] FIG. 7 is a flowchart of an exemplary process 700 for generating an estimated value of the thickness using the neural network 500. The process 700 may be executed by the polishing apparatus 100.

[0066] The polishing apparatus 100 identifies (702) anchor locations from a group of locations on the substrate and obtains (704) measurement values for each location in the group of locations. In some embodiments, the anchor locations are spaced apart from the edge of the substrate.

[0067] The polishing apparatus 100 normalizes (706) each measurement signal of the anchor locations based on the measurement signals of the anchor locations, for example, by dividing each measurement signal by the measurement signal of the anchor location to update the measurement signal. The polishing apparatus 100 then processes (708) the updated measurement signals via the neural network 500 to generate each normalized measurement signal and converts (710) the correction signal to a non-normalized signal using the measurement signal of the anchor location, for example, by dividing each measurement signal by the measurement signal of the anchor location to update the measurement signal. The polishing apparatus 100 then uses (612) the non-normalized signal to generate an estimated value of the thickness at each location in the group of locations of the neural network 500.

[0068] FIG. 8 is a flowchart of an exemplary process 800 for training a neural network 500 to generate a correction signal for a group of measurement signals. Process 800 may be performed by one or more computer systems configured to train neural network 500.

[0069] The system obtains (802) an estimated thickness value generated by neural network 500 based on input values including measurements for each of a group of locations on a substrate. The system also obtains (804) a ground truth value of the thickness for each of the group of locations. The system can generate the ground truth value of the thickness using an electrical impedance measurement method such as the four-point probe method.

[0070] The system calculates (806) an error amount between the estimated thickness value and the ground truth value of the thickness, and updates one or more parameters of neural network 500 based on the error amount. To do so, the system may use a training algorithm that employs gradient descent by backpropagation.

[0071] The monitoring system can be used in various polishing systems. Either the polishing pad or the carrier head, or both, can move to cause relative movement between the polishing surface and the substrate. The polishing pad can be a circular (or some other shape) pad fixed to a platen, a tape extending between a supply roller and a recovery roller, or a continuous belt. The polishing pad can be attached to the platen, gradually fed onto the platen during the polishing operation, or continuously driven on the platen during polishing. The pad can be fixed to the platen during polishing, or a fluid bearing can exist between the platen and the polishing pad during polishing. The polishing pad can be a standard (e.g., polyurethane with or without filler) coarse pad, a soft pad, or a fixed abrasive pad.

[0072] While the above description has focused on eddy current monitoring systems, the correction techniques are applicable to other types of monitoring systems (e.g., optical monitoring systems that scan over the edges of a substrate). Additionally, while the above description has focused on polishing systems, the correction techniques are applicable to other types of substrate processing systems (e.g., deposition systems or etching systems that include in-situ monitoring systems that scan over the edges of a substrate).

[0073] Numerous embodiments of the present invention have been described. However, it should be understood that various modifications can be made without departing from the essence and scope of the present invention. Accordingly, other embodiments are within the scope of the following claims.

Claims

1. A method for polishing a substrate, polishing a layer on the substrate at a polishing station, monitoring, by an in-situ monitoring system, the layer being polished at the polishing station to generate a plurality of initial values for a first plurality of different locations and a second plurality of different locations on the layer, wherein the plurality of initial values are raw measurement signals or updated measurement signals generated by the in-situ monitoring system, the first plurality of different locations are within a first region on the substrate, the second plurality of different locations are within a second region different from the first region on the substrate, the first region is adjacent to an edge of the substrate, and the second region includes a center of the substrate; monitoring the layer; inputting the plurality of initial values for the first plurality of different locations into a first plurality of corresponding input nodes of a neural network; not inputting the plurality of initial values for the second plurality of different locations into the input nodes of the neural network; receiving a first plurality of output values from a plurality of corresponding output nodes of the neural network; detecting a polishing end point and / or modifying a polishing parameter based on the first plurality of output values and the plurality of initial values for the second plurality of different locations; and A method for polishing a substrate.

2. The method according to claim 1, wherein the number of input nodes of the neural network is greater than or equal to the number of output nodes of the neural network.

3. converting the plurality of initial values for the second plurality of different locations into a plurality of thickness values using a calibration curve, and Converting the plurality of initial values for the first plurality of different locations into values of a plurality of thicknesses using the calibration curve before inputting the plurality of initial values for the first plurality of different locations into the first plurality of corresponding input nodes of the neural network, or converting the first plurality of output values into values of a plurality of thicknesses using the calibration curve, the method according to claim 1.

4. The method according to claim 1, wherein the in-situ monitoring system includes an eddy current monitoring system.

5. A polishing system, A carrier for holding a substrate, A support for a polishing surface, An in-situ monitoring system having a sensor, configured to generate a plurality of initial values for a first plurality of different locations and a second plurality of different locations on a layer on the substrate, the plurality of initial values being raw measurement signals or updated measurement signals generated by the in-situ monitoring system, the first plurality of different locations being within a first region on the substrate, the second plurality of different locations being within a second region different from the first region on the substrate, the first region including an edge of the substrate, the second region including a center of the substrate, an in-situ monitoring system, A motor for generating relative movement between the sensor and the substrate, A controller and comprising, the controller being configured to Input the plurality of initial values for the first plurality of different locations into the first plurality of corresponding input nodes of the neural network and Not input the plurality of initial values for the second plurality of different locations into the first plurality of corresponding input nodes of the neural network, Receive a first plurality of output values from a plurality of corresponding output nodes of the neural network, and Detecting a polishing end point and / or modifying polishing parameters based on the plurality of first output values and the plurality of initial values for the plurality of second different locations is configured to perform a polishing system. **Claim 6** The polishing system according to claim 5, wherein the number of input nodes of the neural network is equal to the number of output nodes of the neural network. **Claim 7** The controller is configured to determine the positions of each of the plurality of initial values and classify the plurality of initial values into a plurality of initial values for the plurality of first different locations and the plurality of second different locations The polishing system according to claim 5. **Claim 8** The controller converts the plurality of initial values for the plurality of second different locations into a plurality of thickness values using a calibration curve, and converts the plurality of initial values for the plurality of first different locations into a plurality of thickness values using the calibration curve before inputting the plurality of initial values for the plurality of first different locations into the first plurality of corresponding input nodes of the neural network, or converts the plurality of first output values into a plurality of thickness values using the calibration curve. The polishing system according to claim 5. **Claim 9** The polishing system according to claim 5, wherein the in-situ monitoring system includes an eddy current monitoring system. **Claim 10** The polishing system according to claim 5, wherein the number of input nodes of the neural network is greater than the number of output nodes of the neural network. **Claim 11** The controller further inputs a plurality of initial values for a third plurality of different locations into a second plurality of corresponding input nodes of the neural network configured such that the third plurality of different locations are different from the first region and the second region on the substrate and are within a third region that is between the first region and the second region, and the first plurality of output values are calculated by the neural network based on the plurality of initial values for the first plurality of different locations and the third plurality of different locations, and the controller is configured to detect the polishing end point or correct the polishing parameters further based on the plurality of initial values for the third plurality of different locations, the polishing system according to claim 10.

12. converting the plurality of initial values for the second plurality of different locations into a plurality of thickness values using a calibration curve, and converting the plurality of initial values for the first plurality of different locations and the third plurality of different locations into a plurality of thickness values using the calibration curve before inputting the plurality of initial values to the first plurality of corresponding input nodes and the second plurality of corresponding input nodes of the neural network, or converting the first plurality of output values into a plurality of thickness values using the calibration curve, the polishing system according to claim 11.

13. A computer program product encoded with instructions that, when executed by one or more computers, cause the one or more computers to perform operations, the operations including receiving, during polishing of a layer of a substrate at a polishing station, a plurality of initial values for a first plurality of different locations and a second plurality of different locations on the layer from an in-situ monitoring system, the plurality of initial values being raw measurement signals or updated measurement signals generated by the in-situ monitoring system, the first plurality of different locations being within a first region on the substrate, the second plurality of different locations being within a second region different from the first region on the substrate, the first region including an edge of the substrate, and the second region including a center of the substrate, receiving the plurality of initial values; Inputting the plurality of initial values for the first plurality of different locations into a first plurality of corresponding input nodes of a neural network; Not inputting the plurality of initial values for the second plurality of different locations into the first plurality of corresponding input nodes of the neural network; Receiving a first plurality of output values from a plurality of corresponding output nodes of the neural network; Detecting at least one of a polishing end point and modifying a polishing parameter based on the first plurality of output values and the plurality of initial values for the second plurality of different locations; A computer program product comprising the above.

14. The computer program product according to claim 13, wherein the number of input nodes of the neural network is equal to the number of output nodes of the neural network.

15. Instructions for determining positions of each of the plurality of initial values and classifying the plurality of initial values into a plurality of initial values for the first plurality of different locations and the second plurality of different locations; The computer program product according to claim 13.

16. Instructions for converting the plurality of initial values for the second plurality of different locations into values of a plurality of thicknesses using a calibration curve, and Instructions for converting the plurality of initial values for the first plurality of different locations into values of a plurality of thicknesses using the calibration curve before inputting the plurality of initial values for the first plurality of different locations into the first plurality of corresponding input nodes of the neural network, or instructions for converting the first plurality of output values into values of a plurality of thicknesses using the calibration curve, the computer program product according to claim 13.

17. The computer program product according to claim 13, wherein the number of input nodes of the neural network is greater than the number of output nodes of the neural network.

18. Instructions for inputting a plurality of initial values for a third plurality of different locations into a second plurality of corresponding input nodes of the neural network, wherein the third plurality of different locations are different from the first region and the second region on the substrate and are within a third region between the first region and the second region, and the first plurality of output values are calculated by the neural network based on the plurality of initial values for the first plurality of different locations and the third plurality of different locations, and further based on the plurality of initial values for the third plurality of different locations, instructions for detecting the polishing end point or modifying the polishing parameters, the computer program product according to claim 17.

19. Instructions for converting the plurality of initial values for the second plurality of different locations into a plurality of thickness values using a calibration curve, and Instructions for converting the plurality of initial values for the first plurality of different locations and the third plurality of different locations into a plurality of thickness values using the calibration curve before inputting the plurality of initial values into the first plurality of corresponding input nodes and the second plurality of corresponding input nodes of the neural network, or instructions for converting the first plurality of output values into a plurality of thickness values using the calibration curve, the computer program product according to claim 18.

Images