Polishing apparatus using neural network for monitoring

Abstract

To provide a method of polishing a layer on a substrate at a polishing station.SOLUTION: A method of polishing a layer on a substrate at a polishing station includes the actions of monitoring the layer during polishing at the polishing station with an in-situ monitoring system to generate a plurality of measured signals for a plurality of different locations on the layer; generating, for each location of the plurality of different locations, an estimated measure of thickness of the location, the generating step including processing the plurality of measured signals through a neural network; and at least one of detecting a polishing endpoint or modifying a polishing parameter based on each estimated measure of thickness.SELECTED DRAWING: Figure 1A

Description

[Technical Field]

[0001] FIELD OF THE DISCLOSURE The present disclosure relates to in situ monitoring during substrate polishing. [Background technology]

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

[0003] One manufacturing step involves depositing a filler layer over a non-planar surface and planarizing the filler layer until the non-planar surface is exposed. For example, a conductive filler layer can be deposited over a patterned insulating layer to fill trenches or holes in the insulating layer. The filler layer is then polished until the raised pattern of the insulating layer is exposed. After planarization, portions of the conductive layer remaining between the raised pattern of the insulating layer form vias, plugs, and lines that provide conductive paths between thin film circuits on the substrate. Furthermore, planarization can be used to flatten the substrate surface for lithography.

[0004] Chemical-mechanical polishing (CMP) is a recognized planarization method. This planarization method typically involves mounting a substrate on a carrier head. The exposed surface of the substrate is placed against a rotating polishing pad. The carrier head exerts a controllable load on the substrate, pressing it against the polishing pad. A polishing fluid, such as an abrasive slurry, is supplied to the surface of the polishing pad.

[0005] During semiconductor processing, it may be important to determine one or more characteristics of a substrate or a layer on a substrate. For example, it may be important to know the thickness of a conductive layer during a CMP process so that the process can be terminated at the appropriate time. A number of methods may be used to determine substrate characteristics. For example, optical sensors may be used for in-situ monitoring of a substrate during chemical mechanical polishing. Alternatively (or additionally), eddy current sensing systems may be used to induce eddy currents in conductive regions of a substrate to determine parameters such as the local thickness of the conductive regions. Summary of the Invention

[0006] In one aspect, 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 for a plurality of different locations on the layer, generating a plurality of estimates 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 polishing parameters or detecting a polishing endpoint based on each thickness estimate.

[0007] In another aspect, corresponding computer systems, devices, and computer programs stored on one or more computer storage devices are configured to implement the method. One or more computer systems may be configured to perform particular operations or functions by software, firmware, hardware, or a combination thereof installed on the system such that the system performs the functions during operation. One or more computer programs are configured to perform particular operations or functions by including instructions that, when executed by a data processing device, cause the device to perform the operations.

[0008] In another aspect, a 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 generating relative motion 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 measurement signals from the in-situ monitoring system and generate, for each of the plurality of different locations, an estimate of a thickness at the location, the generating including processing the plurality of signals through a neural network, and detect a polishing endpoint, modify polishing parameters based on each thickness estimate, 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 estimate of the thickness at that location may be generated based on the measurement signal at that location. Generating the thickness estimate may include using a static formula relating the plurality of values ​​of the measurement signals to the plurality of values ​​of the thickness estimate. A third plurality of measurement signals may be obtained for the layer of the second substrate. Each of the third plurality of measurement signals may correspond to one of the third plurality of locations on the layer on the second substrate. For each of the third plurality of different locations, an estimate of the thickness at that location may be generated based on the measurement signal at that location. Generating the thickness estimate may include using a static formula relating the plurality of values ​​of the thickness estimate to the plurality of values ​​of the measurement signals.

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

[0012] A ground truth measure of thickness may be obtained for each of the one or more locations of the layer. An error measure between the thickness estimate for each location and the corresponding ground truth thickness value for that location may be calculated. Parameters of the neural network system may be updated based on the error measure. The ground truth thickness value may be determined based on a four-point probe method. Updating the parameters of the neural network system based on the error measure may include backpropagating a 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, for example, an eddy current monitoring system, can generate a signal as a sensor scans across the substrate. The system can compensate for distortions in the portion of the signal that corresponds to the substrate edge. The signal can be used for endpoint control and / or closed-loop control of polishing parameters, such as carrier head pressure, which can result 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 the description and drawings, and from the claims. [Brief explanation of the drawings]

[0015] [Figure 1A] 1 is a schematic side view, partially in cross section, of a chemical mechanical polishing station including an eddy current monitoring system; [Figure 1B] FIG. 2 is a schematic top view of a chemical mechanical polishing station. [Figure 2] 2 is a schematic top view of a substrate scanned by a sensor head of a polishing apparatus. FIG. [Figure 3] 1 is a schematic graph of a static formula for determining substrate thickness based on a measurement signal. [Figure 4] 4 is a schematic graph of a measurement signal obtained during monitoring of a location on a substrate. [Figure 5] 1 is an exemplary neural network. [Figure 6] FIG. 1 is a flow diagram of an exemplary process for polishing a substrate. [Figure 7] FIG. 1 is a flow diagram of an exemplary process for generating thickness estimates using a neural network. [Figure 8] FIG. 1 is a flow diagram of an exemplary process for training a neural network to generate correction signals for a set of measurement signals.

[0016] Like reference symbols in the various drawings indicate like elements. DETAILED DESCRIPTION OF THE INVENTION

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

[0018] In-situ monitoring systems can suffer from signal distortion when measuring near the substrate edge. For example, eddy current monitoring systems can generate magnetic fields. Near the substrate edge, the signal can be artificially low because the magnetic field only partially overlaps with the conductive layer of the substrate. However, if the polishing apparatus uses a neural network to generate a modified signal based on the measurement signal generated by the in-situ monitoring system, the apparatus can compensate for the distortion, e.g., reduce the signal strength at the substrate edge.

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

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

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

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

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

[0024] Although only one carrier head 140 is shown, more carrier heads can be provided to hold additional substrates, thereby allowing for efficient use of the surface area of ​​the polishing pad 110.

[0025] The polishing apparatus 100 includes an in-situ monitoring system 160 that generates a sequence of time-varying values ​​that depend on the thickness of a layer on the substrate. The in-situ monitoring system 160 includes a sensor head from which measurements are generated, and relative motion between the substrate and the sensor head allows measurements to be taken at different locations on the substrate.

[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 detection 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 wrapped around a portion of the core 162, and a drive and detection 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 protrudes above the top surface of the platen 120, for example, into a recess 118 in the bottom of the polishing pad 110.

[0027] The drive and sense circuitry 166 is configured to apply an oscillating electrical signal to the coil 164 and measure the resulting eddy currents. Various configurations are possible for the drive and sense circuitry and for the configuration and location of the coils, as described, for example, 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. The drive and sense circuitry 166 can be located in the same recess 128 or in different portions of the platen 120, or can be located outside the platen 120 and coupled to the platen components by a rotary electrical coupling 129.

[0028] During operation, the drive and sense circuitry 166 drives the coil 164 to generate an oscillating magnetic field. At least a portion of the magnetic field extends through the polishing pad 110 to the substrate 10. If a conductive layer is present on the substrate 10, the oscillating magnetic field generates eddy currents in the conductive layer. The eddy currents cause the conductive layer to act as an impedance source that is coupled to the drive and sense circuitry 166. Changes in the thickness of the conductive layer result in a change in impedance that can be detected by the drive and sense circuitry 166.

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

[0030] The CMP apparatus 100 may also include a position sensor 180, such as an optical isolator, to detect when the core 162 is under the substrate 10. For example, the optical isolator may be mounted in a fixed position opposite 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 so that the flag 182 blocks the optical signal of the sensor 180 while the core 162 passes under the substrate 10. Alternatively or additionally, the CMP apparatus may include an encoder for determining the angular position of the platen.

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

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

[0033] As the sensor 162 passes under the substrate with each rotation of the platen, conductive layer thickness information 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 measurements from the monitoring system when the substrate is generally over the core 162 (as determined by the position sensor). As polishing progresses, the conductive layer thickness changes, causing the sampled signal to vary with time. The time-varying sampled signal is sometimes referred to as a trace. During polishing, measurements from the monitoring system are displayed on an output device 192, allowing an operator of the device to visually monitor the progress of the polishing operation.

[0034] During operation, CMP apparatus 100 can use 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 for the detector logic and endpoint criteria include local minima or maxima, change in slope, amplitude or slope thresholds, or combinations thereof.

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

[0036] The controller 190 can use a correlation curve that relates the signal measured by the in-situ monitoring system 160 to the thickness of the layer being polished on the substrate 10 to generate an estimate of the thickness of the layer being polished. 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 on the substrate 10. For a given signal value, the controller 190 can use the correlation curve 303 to generate a corresponding thickness value. The correlation curve 303 can be considered a "static" formula in that it predicts a thickness value for each signal value regardless of the time or location at which the sensor head acquired the signal. The correlation curve can be represented by a variety of functions, such as a polynomial function or a look-up table (LUT) coupled with linear interpolation.

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

[0038] 4 is a graph 420 showing one signal 401 from the in-situ monitoring system 160 during a single pass of the sensor head under the substrate 10. The signal 401 is made up of a series of individual measurements from the sensor head as it passes under the substrate. The graph 420 can be a function of 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 locations on the substrate 10 scanned by the sensor head. Thus, the graph 420 shows corresponding measurement signal values ​​from the signal 401 for a given location on the substrate scanned by the sensor head.

[0039] 2 and 4, signal 401 includes a first portion 422 corresponding to a location within the edge region 203 of substrate 10 when the sensor head crosses the leading edge of substrate 10, a second portion 424 corresponding to a location within the central region 201 of substrate 10, and a third portion 426 corresponding to a location within the edge region 203 when the sensor head crosses the trailing edge of substrate 10. The signal may also include a portion 428 corresponding to off-substrate measurements, i.e., the signal generated when the sensor head scans an area beyond the edge 204 of substrate 10 in FIG.

[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 interior region 205 surrounded by the anchor region 202. The sensor head scans these regions on its path 210 and generates a series of measurements corresponding to a series of points along the path 210.

[0041] In the first portion 422, the signal strength increases from an initial strength (typically a signal generated when the substrate and carrier head are not present) to a high strength. This is caused by the monitoring location initially moving from only a small amount of overlap with the substrate at the edge 204 of the substrate (producing an initial low value) to a monitoring location that is almost entirely overlapping the substrate (producing a high value). Similarly, in the third portion 426, the signal strength decreases as the monitoring location moves from the edge 204 to the substrate.

[0042] Second portion 424 is depicted as flat for simplicity's sake; the actual signal at second portion 424 may include fluctuations due to noise and layer thickness variations. Second portion 424 corresponds to a monitoring location scanning central region 201. Second portion 424 includes subportions 421 and 423 caused by a monitoring location scanning anchor region 202 of central region 201, and subportion 427 caused by a monitoring location scanning interior region 205 of central region 201.

[0043] As discussed above, variations in signal strength at regions 422, 426 are caused in part by the sensor's measurement area overlying the substrate edge, rather than by inherent variations in the thickness or conductivity of the layer being monitored. As a result, this distortion of signal 401 can be a source of error when calculating substrate characteristics, such as the thickness of a layer near the substrate edge. To address this issue, controller 190 can include a neural network (e.g., neural network 500 of FIG. 5) that generates correction signals corresponding to one or more locations on substrate 10 based on the measurement signals corresponding to those locations.

[0044] 5, a neural network 500 is configured, when properly trained, to reduce and / or eliminate distortion in calculated signal values ​​near the substrate edge. The 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 the neural network 500 include an input layer 510, an output layer 530, and one or more hidden layers 520.

[0045] Each layer of the neural network 500 includes one or more neural network nodes. Each neural network node of a neural network layer receives one or more input values ​​(either from inputs 504 to the neural network 500 or from the outputs of one or more nodes of a previous neural network layer), processes the node input values ​​according to one or more parameter values ​​to generate activation values, and optionally applies a nonlinear transformation function (e.g., a sigmoid function or a tanh function) to the activation values ​​to generate the neural network node's output.

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

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

[0048] Typically, the plurality of distinct locations includes a plurality of locations in the edge region 203 and anchor region 202 of the substrate 10. In some implementations, the plurality of distinct locations is only within the edge region 203 and anchor region 202. In other implementations, the plurality of distinct locations is spread across the entire area of ​​the substrate.

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

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

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

[0052] For example, the number of signal input nodes 544 may be equal to the number of measurements of edge regions 203 and anchor regions 202, and may be the same number as the number of output nodes 550. Thus, each output node 550 generates a correction signal corresponding to each measurement signal provided as 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 can be equal to the number of measurements of edge regions 203, or can be equal to the number of measurements of edge regions 203 and anchor regions 202. Again, each output node 550 of the output layer 530 generates a modified signal corresponding to each measurement signal provided as a signal input node 504 (e.g., a first modified signal 551 for the first measurement signal 501, but only for signal input nodes 554 that receive signals from edge regions 203).

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

[0055] The modified signal values ​​430 may be converted into thickness measurements using a static formula, e.g., a correlation curve. For example, the controller 190 may use a neural network 500 to identify the thickness of the edge locations and one or more anchor locations of the substrate. In contrast, the controller 190 may use a static formula directly to generate thickness measurements for other regions, e.g., the interior region 205. That is, the signal values ​​of other regions, e.g., the interior region 205, may be converted into thickness values ​​without being transformed by a neural network.

[0056] In some implementations, for a modified signal value corresponding to a given measurement location, neural network 500 may be configured such that only signal values ​​for measurement locations within a predetermined distance from the given location are used to determine the modified signal value. For example, signal values ​​S1, S2, ..., S3 corresponding to measurements at N consecutive locations on path 210 may be used to determine the modified signal value. M , …S N When received, N When received, the Mth part (R M (denoted as M is S M-L(min 1) , …S M , …S M+L(max N) Only use the corrected signal value S' M The value of L is calculated by taking the measured values ​​at a maximum distance of approximately 2-4 mm from each other and calculating the corrected signal value S'. M is used to generate the measurement S M The value of L may be selected such that measurements within about 1-2 mm (e.g., 1.5 mm) of the point can be used. For example, L can be a number in the range of 0 to 4 (e.g., 1 or 2). For example, if measurements within 3 mm are used and the spacing between measurements is 1 mm, L can be 1; if the spacing is 0.5 mm, L can be 2; and if the spacing is 0.25 mm, L can be 4. However, this depends on the polishing machine configuration and processing conditions. Other parameters (e.g., pad wear) can still be used to determine the corrected signal value S'. M can be used to calculate

[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, with each hidden node 570 corresponding to a respective signal input node 544. Each hidden node 570 may be isolated (or may have a parameter value of zero) from input nodes 544 corresponding to measurements that are more than a predetermined distance away from the measurement location of the corresponding input node. For example, the Mth hidden node may be isolated from the first through (ML-1)th input nodes 544 and from the (M+L+1)th through Nth input nodes. Similarly, each output node 560 may be isolated (or may have a parameter value of zero) from hidden nodes 570 corresponding to modified signals that are more than a predetermined distance away from the measurement location of the output node. For example, the Mth output node may be isolated from the first through (ML-1)th hidden nodes 570 and from the (M+L+1)th through Nth hidden nodes.

[0058] In some embodiments, the polishing apparatus 100 can use a static formula to determine thickness values ​​at multiple locations (e.g., locations within the edge region of a first group of substrates). These substrates can be used to generate data used to train the neural network. The polishing apparatus 100 can then use the neural network 500 to generate a correction signal used to determine thickness values ​​at multiple locations (e.g., locations within the edge region of a second group of substrates). For example, the polishing apparatus 100 can apply the static formula to determine thickness values ​​for the first group of substrates and use the trained neural network 500 to generate a correction signal used to determine thickness values ​​for the second group of substrates.

[0059] 6 is a flow diagram of an exemplary process 600 for polishing substrate 10. Process 600 can be performed by polishing apparatus 100.

[0060] The polishing apparatus 100 polishes (602) a 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 the 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 located within the central region 201 of the substrate, and therefore is not affected by distortion caused 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 an interior region 205 of the central region 201. The number of anchor locations may depend on the measurement spot size and measurement frequency of the in-situ monitoring system 160. In some embodiments, the number of anchor locations may not exceed a maximum value (e.g., a maximum of 4).

[0061] The polishing apparatus 100 generates (606) a thickness estimate for each of the different locations based on the measurement signals at that location, which includes processing the measurement signals through the neural network 500.

[0062] Inputs to the neural network 500 can be raw measurement signals generated by the in-situ monitoring system 160 for different locations, or updated measurement signals. In some embodiments, the apparatus 100 updates each measurement signal by normalizing the signal's value. Such normalization increases the likelihood that at least some of the inputs 504 to the neural network system 500 will fall within a particular range, which in turn can improve the quality of the neural network training and / or the accuracy of the inferences made by the neural network 500.

[0063] The outputs of the neural network 500 are corrected signals, each corresponding to an input measurement signal. If the measurement signals are normalized values, the corrected signals corresponding to the measurement signals will also be normalized values. Therefore, the polishing apparatus 100 may need to convert such corrected signals to non-normalized values ​​before using the corrected signals to estimate the thickness of the substrate.

[0064] The polishing apparatus 100 modifies polishing parameters and / or detects a polishing endpoint based on each thickness estimate (608).

[0065] 7 is a flow diagram of an exemplary process 700 for generating a thickness estimate using the neural network 500. The process 700 may be performed by the polishing apparatus 100.

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

[0067] The polishing apparatus 100 normalizes (706) each measurement signal at the anchor points based on the measurement signals at the anchor points, for example, by dividing each measurement signal by the measurement signal at the anchor points to update the measurement signals. The polishing apparatus 100 then processes (708) the updated measurement signals through the neural network 500 to generate each normalized measurement signal, and converts (710) the modified signals to unnormalized signals using the measurement signals at the anchor points, for example, by dividing each measurement signal by the measurement signal at the anchor points to update the measurement signals. The polishing apparatus 100 then uses (612) the unnormalized signals to generate thickness estimates at each of the group of points in the neural network 500.

[0068] 8 is a flow diagram of an example process 800 for training neural network 500 to generate correction signals for a set 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 thickness estimates generated by neural network 500 based on inputs including measurements for each of a set of substrate locations. The system also obtains 804 ground truth thickness values ​​for each of the set of locations. The system may generate the ground truth thickness values ​​using an electrical impedance measurement method, such as a four-point probe method.

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

[0071] The monitoring system can be used in a variety of polishing systems. Either the polishing pad or the carrier head, or both, can move to cause relative motion 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 return roller, or a continuous belt. The polishing pad can be attached to the platen, can be incrementally fed over the platen between polishing operations, or can be continuously driven over the platen during polishing. The pad can be fixed to the platen during polishing, or a fluid bearing can be present between the platen and the polishing pad during polishing. The polishing pad can be a standard coarse pad (e.g., polyurethane with or without fillers), a soft pad, or a fixed-abrasive pad.

[0072] Although the above description focuses on eddy current monitoring systems, the modification techniques are applicable to other types of monitoring systems (e.g., optical monitoring systems that scan over the edge of a substrate). Additionally, although the above description focuses on polishing systems, the modification techniques are applicable to other types of substrate processing systems (e.g., deposition or etch systems that include in-situ monitoring systems that scan over the edge of a substrate).

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

Claims

1. A method for polishing a substrate, comprising: polishing a layer on the substrate at a polishing station; monitoring the layer being polished at the polishing station with an in-situ monitoring system to generate a plurality of measurement signals for a plurality of different locations on the layer; generating, for each of the plurality of different locations, an estimate of thickness at the location, the estimate comprising processing the plurality of measurement signals through a neural network; At least one of modifying polishing parameters based on each thickness estimate or detecting a polishing endpoint; A method for polishing a substrate, comprising:

2. obtaining a ground truth thickness value for each of one or more different locations of the layer; calculating an error measure between the estimated thickness for each location and the ground truth thickness for the corresponding location; updating the parameters of the neural network system based on the error amount; The method of claim 1 further comprising:

3. The method of claim 2 , wherein the ground truth values ​​of thickness are determined based on a four-point probe method.

4. a carrier for holding the substrate; a support for the polishing surface; an in-situ monitoring system having a sensor configured to generate measurement signals for a plurality of different locations of the formation; a motor that generates relative motion between the sensor and the substrate; a controller, receiving the plurality of measurement signals from the in-situ monitoring system; generating, for each of the plurality of different locations, an estimate of a thickness of the location, said generating including processing the plurality of measurement signals through a neural network; configured to modify polishing parameters based on each thickness estimate, detect a polishing endpoint, or both; A controller; A polishing system comprising:

5. 5. The polishing system of claim 4, wherein the controller is further configured to generate, for each of a second plurality of different locations on the layer, an estimate of the thickness of the location based on the measurement signal at the location, and the generating includes using a static formula relating the plurality of values ​​of the thickness estimate to the plurality of values ​​of the measurement signal.

6. 5. The polishing system of claim 4, wherein the controller is further configured to generate, for each of a third plurality of different locations on the layer of the second substrate, an estimate of the thickness of the location based on the measurement signal of the location, and the generating includes using a static formula relating the plurality of values ​​of the thickness estimate to the plurality of values ​​of the measurement signal.

7. The polishing system of claim 4 , wherein the in-situ monitoring system comprises an eddy current sensor.

8. A computer storage medium encoded with instructions that, when executed by one or more computers, cause the one or more computers to perform operations, the operations including: receiving a plurality of measurement signals from an in-situ monitoring system for a plurality of different locations on a layer of a substrate while the layer is being polished at a polishing station; generating, for each of the plurality of different locations, an estimate of a thickness of the location, the estimate comprising processing the plurality of measurement signals through a neural network; At least one of modifying polishing parameters based on each thickness estimate or detecting an end-of-polishing point; a computer storage medium,

9. The operation further comprises: receiving a second plurality of measurement signals for a second plurality of different locations of the layer; generating, for each of the second plurality of different locations, an estimate of a thickness of the location based on the measurement signals at the location, the estimate including using a static formula relating the plurality of values ​​of the thickness estimate to the plurality of values ​​of the measurement signals; 9. The computer storage medium of claim 8, comprising:

10. The operation further comprises: receiving a third plurality of measurement signals of a layer on a second substrate, each of the third plurality of measurement signals corresponding to one location of a third plurality of locations of the layer on the second substrate; generating, for each of the third plurality of different locations, an estimate of a thickness of the location based on the measurement signals at the location, the estimate including using a static formula relating the plurality of values ​​of the thickness estimate to the plurality of values ​​of the measurement signals; 9. The computer storage medium of claim 8, comprising:

11. the neural network includes one or more neural network layers, including an input layer, an output layer, and one or more hidden layers; Each neural network layer includes one or more neural network nodes; 9. The computer storage medium of claim 8, wherein each neural network node is configured to process inputs according to a set of parameters to generate outputs.

12. 12. The computer storage medium of claim 11, wherein the inputs to the neural network nodes of the input layer include measurements of pad wear of the polishing station.

13. 12. The computer storage medium of claim 11, wherein the one or more different locations include anchor locations, and wherein identifying each first measurement of thickness includes normalizing each measurement signal based on the measurement signal for the anchor location to update the measurement signal.

14. 14. The computer storage medium of claim 13, wherein each thickness estimate is a normalized value, and the operation further comprises converting each thickness estimate to a non-normalized value using the measurement signal for the anchor point to update the thickness estimate.

15. receiving a ground truth thickness value for each of one or more distinct locations of the layer; calculating an error measure between the estimate of thickness for each location and the corresponding ground truth value of thickness at the location; updating the parameters of the neural network system based on the error amount; The computer storage medium of claim 8 further comprising:

Images