Signal processing in chemical mechanical polishing

Abstract

Disclosed herein is a method including storing data representing a sequence of background traces generated by a sequence of scans of a sensor of an in-situ eddy current monitoring system across a blank calibration substrate; monitoring the layer during the polishing with the in-situ eddy current monitoring system; for each respective measured trace in the sequence of measured traces, subtracting a respective background trace that has an equivalent position in the sequence of background traces from the respective measured trace to generate a corrected thickness trace and thus provide a sequence of corrected thickness traces; and at least one of detecting a polishing endpoint or modifying a polishing parameter that affects the polishing based on the sequence of corrected thickness traces.

Description

Attorney DocketNo. 44024776WO01; 05542-1643WO1SIGNAL PROCESSING IN CHEMICAL MECHANICAL POLISHINGTECHNICAL FIELD

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

[0002] An integrated circuit is typically formed on a substrate (e.g. a semiconductor wafer) by the sequential deposition of conductive, semiconductive or insulative layers on a silicon wafer, and by the subsequent processing of the layers.

[0003] One fabrication 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 on a patterned insulative layer to fill the trenches or holes in the insulative layer. The filler layer is then polished until the raised pattern of the insulative layer is exposed. After planarization, the portions of the conductive layer remaining between the raised pattern of the insulative layer form vias, plugs and lines that provide conductive paths between thin film circuits on the substrate. In addition, planarization may be used to planarize the substrate surface for lithography.

[0004] Chemical mechanical polishing (CMP) is one accepted method of planarization. This planarization method typically requires that the substrate be mounted on a carrier head. The exposed surface of the substrate is placed against a rotating polishing pad. The carrier head provides a controllable load on the substrate to push it against the polishing pad. A polishing liquid, such as slurry with abrasive particles, is supplied to the surface of the polishing pad.

[0005] In some systems, a substrate is monitored in-situ during polishing, e.g., through the polishing pad. For example, an eddy current sensing system may be used to induce eddy currents in a conductive layer. The signal from the sensor can be used to determine the thickness of the layer on the substrate during polishing.SUMMARY

[0006] In general, an aspect disclosed herein is a method of chemical mechanical polishing, including storing data representing a sequence of background traces generated by a sequence of scans of a sensor of an in-situ eddy current monitoring system across a blank calibration substrate; polishing a conductive layer on a device substrate; monitoring the layer during theAttorney DocketNo. 44024776WO01; 05542-1643WO1 polishing with the in-situ eddy current monitoring system, including repeatedly sweeping the sensor of the in-situ eddy current monitoring system across substrate such that each sweep generates a sequence of raw signal values, and providing a measured trace from each sequence of raw signal values such that the repeated sweeping provides a sequence of measured traces; for each respective measured trace in the sequence of measured traces, subtracting a respective background trace that has an equivalent position in the sequence of background traces from the respective measured trace to generate a corrected thickness trace and thus provide a sequence of corrected thickness traces; and at least one of detecting a polishing endpoint or modifying a polishing parameter that affects the polishing based on the sequence of corrected thickness traces.

[0007] Examples may include one or more of the following features. Each measured trace can be a thickness profile including thickness values as a function of radial position on the device substrate. The method may include converting the sequence of raw signal values to thickness values using a correlation curve to provide a preliminary7thickness trace. The method may include performing edge reconstruction on preliminary7thickness trace by processing portions of the preliminary thickness trace corresponding to an edge of the device substrate through a neural network to generate a modified thickness trace that provides the measured trace. Each background trace can represent a sequence of raw signal background values that are converted to effective thickness values using the correlation curve to provide a preliminary7effective thickness trace which is then subjected to edge reconstruction by having portions corresponding to an edge of the calibration substrate processed through a neural network to generate a modified effective thickness trace that provides the background trace. The blank calibration substrate may include of a semiconductor substrate. The blank calibration substrate may include a lightly doped semiconductor substrate. The blank calibration substrate may include an unpattemed dielectric layer coating a semiconductor substrate. The unpattemed dielectric layer can be an oxide layer. The oxide layer can be a thermal oxide. The method may include rotating a platen supporting a polishing pad for the polish of the substrate to a first predetermined angular position prior to polishing the layer on the substrate. The method may include counting a number of rotations of the platen to provide an ordinal position of the measured trace in the sequence of measured traces. The method may include measuring an angular orientation of a carrier head prior to polishing and rotating the carrier head to a first predetermined angular position prior to polishing. The method may include storing data representing an ordinal position of each background trace in the sequence of background traces. An ordinal position of each background trace in theAttorney DocketNo. 44024776WO01; 05542-1643WO1 sequence of background traces can be provided by ordering of the data. The method may include instructions to average traces accumulated from different sensors within a single rotation of a platen to generate an averaged trace. The method may include instructions to subtract the respective background trace from the averaged trace. The conductive layer has a resistivity' higher than 1000 OhmA. The conductive layer may include titanium nitride or tungsten doped carbon.

[0008] In general, an aspect disclosed herein is a non-transitory computer readable medium including instructions encoded therein providing a computer program for controlling a polishing system and causing one or more computers to: store data representing a sequence of background traces generated by a sequence of scans of a sensor of an in-situ eddy current monitoring system across a blank calibration substrate; receive from the in-situ eddy current monitoring system, a sequence of raw signal values for each sw eep by the sensor across the substrate, each sequence of raw signal values providing a measured trace such that the repeated sw eeping provides a sequence of measured traces; for each respective measured trace in the sequence of measured traces, subtract a respective background trace that has an equivalent position in the sequence of background traces from the respective measured trace to generate a corrected thickness trace and thus provide a sequence of corrected thickness traces; and at least one of i) detect a polishing endpoint or ii) modify a polishing parameter that affects the polishing based on the sequence of corrected thickness traces..

[0009] Examples may include one or more of the following features. Each measured trace can be a thickness profile including thickness values as a function of radial position on the device substrate. The non-transitory computer readable medium may include converting the sequence of raw signal values to thickness values using a correlation curve to provide a preliminary’ thickness trace. The non-transitory computer readable medium may include performing edge reconstruction on preliminary thickness trace by processing portions of the preliminary' thickness trace corresponding to an edge of the device substrate through a neural network to generate a modified thickness trace that provides the measured trace. Each background trace can represent a sequence of raw signal background values that are converted to effective thickness values using the correlation curve to provide a preliminary effective thickness trace which is then subjected to edge reconstruction by having portions corresponding to an edge of the calibration substrate processed through a neural netw ork to generate a modified effective thickness trace that provides the background trace. The non- transitory computer readable medium may include rotating a platen supporting a polishing pad for the polish of the substrate to a first predetermined angular position prior to polishingAttorney DocketNo. 44024776WO01; 05542-1643WO1 the layer on the substrate. The non-transitory computer readable medium may include counting a number of rotations of the platen to provide an ordinal position of the measured trace in the sequence of measured traces. The non-transitory computer readable medium may include measuring an angular orientation of a carrier head prior to polishing and rotating the carrier head to a first predetermined angular position prior to polishing. The non-transitory computer readable medium may include storing data representing an ordinal position of each background trace in the sequence of background traces. An ordinal position of each background trace in the sequence of background traces can be provided by ordering of the data. The non-transitory computer readable medium may include instructions to average traces accumulated from different sensors within a single rotation of a platen to generate an averaged trace. The non-transitory computer readable medium may include instructions to subtract the respective background trace from the averaged trace.

[0010] In general, an aspect disclosed herein is a polishing system, including a platen to support a polishing pad; a carrier head to hold a substrate in contact with the polishing pad; a motor to generate relative motion between the carrier head and the platen; an eddy current monitoring system to monitor the substrate during polishing, the eddy current monitoring system including a sensor positioned to repeatedly sweep across the substrate such that each sweep generates a sequence of raw signal values, each sequence of raw signal values providing a measured trace such that repeated sweeping provides a sequence of measured traces; and a controller configured to store data representing a sequence of background traces generated by a sequence of scans of a sensor of an in-situ eddy current monitoring system across a blank calibration substrate; receive from the in-situ eddy current monitoring system, sequence of raw signal values for each sweep by the sensor across the substrate, each sequence of raw signal values providing a measured trace such that the repeated sweeping provides a sequence of measured traces; for each respective measured trace in the sequence of measured traces, subtract a respective background trace that has an equivalent position in the sequence of background traces from the respective measured trace to generate a corrected thickness trace and thus provide a sequence of corrected thickness traces; and at least one of i) detect a polishing endpoint or ii) modify a polishing parameter that affects the polishing based on the sequence of corrected thickness traces.

[0011] In general, an aspect disclosed herein is a method, including aligning a platen and a carrier head that is holding a blank calibration substrate by rotating the platen, the carrier head, or both, such that the carrier head and the platen are at a predetermined angular position; generating a sequence of blank calibration signal traces by sweeping one or moreAttorney DocketNo. 44024776WO01; 05542-1643WO1 sensors of an in-situ monitoring system arranged within a recess of the platen across the blank calibration substrate between the platen and the carrier head; labeling the sequence of blank calibration signal traces according to the sequence by which the blank calibrations signal traces are generated; creating a data file for each blank calibrations signal of the sequence of blank calibration signal traces; and storing each data file for each blank calibration signal trace to provide a library of stored blank calibration signal traces.

[0012] Examples may include one or more of the following features. The method may include measuring an angular orientation of the carrier head prior to polishing and rotating the carrier head to a first predetermined angular position prior to polishing. The method may include rotating a platen supporting a polishing pad for the polish of the blank calibration wafer to a second predetermined angular position prior to polishing a layer on the blank calibration wafer. The method may include counting a number of rotations of the platen to provide an ordinal position of the measured trace in the sequence of measured traces.

[0013] Certain implementations can include one or more of the following advantages. An in-situ monitoring system, e.g., an eddy current monitoring system, can generate a signal as one or more sensors scan across a substrate. The system can determine a signal trace from a ‘blank’ oxide wafer and subtract the blank trace from the production signals. The system can perform this subtraction to generate a modified signal which represents the thickness of the less-conductive layers. The modified signal, in turn, can be used for a more accurate thickness calculation. The system can generate the modified signals on a layer-by-layer basis which enhances measurement accuracy for all layers present on the substrate. This can improve end point control and improve adjustment of polishing parameters to reduce within- wafer non-uniformity (WIWNU) and water-to-wafer non-uniformity (WTWNU). Further, the system can begin the polishing process with the platen and carrier head in a consistent starting angular position. This can improve the likelihood that the sensor will follow a consistent series of paths across the substrate on a wafer-to-wafer basis.

[0014] The 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 DESCRIPTION OF DRAWINGS

[0015] FIG. 1 is a schematic side view, partially cross-sectional, of a chemical mechanical polishing station that includes an eddy current monitoring system.Attorney DocketNo. 44024776WO01 ; 05542-1643WO1

[0016] FIG. 2A is a schematic graph of a static formula for determining substrate thickness based on measured signals.

[0017] FIG. 2B is a schematic graph of simplified measured signals obtained while monitoring locations on a substrate.

[0018] FIG. 3 A is a schematic top view of a substrate being scanned by a sensor head of a polishing apparatus for a single pass.

[0019] FIG. 3B is a schematic top view of a earner head being scanned by a sensor head of a polishing apparatus for multiple passes.

[0020] FIG. 3C is a schematic top view of relative angular positions of a carrier head and a platen.

[0021] FIG. 4 A is a schematic graph of an exemplary signal trace.

[0022] FIG. 4B is a schematic graph of three exemplary signal traces changing with temperature.

[0023] FIG. 4C is a schematic graph of a background trace from a scan path.

[0024] FIG. 4D is a schematic graph of a modified trace after subtracting the background trace.

[0025] FIG. 5 is a flow' diagram of an example process for generating a digital library' storing background traces.

[0026] FIG. 6 is a flow diagram of an example process for polishing a substrate.

[0027] Like reference symbols in the various drawings indicate like elements.DETAILED DESCRIPTION

[0028] A polishing apparatus can use an in-situ monitoring system to detect the thickness of an outer layer that is being polished on a substrate. During polishing of the outer layer, the in-situ monitoring system can determine the thickness of different locations of the layer on the substrate. The thickness measurements can be used to trigger a polishing endpoint 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 increase or decrease the polishing rate of the locations of the outer layer. The polishing rate can be adjusted so that the layer has a substantially uniform thickness after polishing. The CMP system can adjust the polishing rate so that polishing of the locations of the layer completes at about the same time. Such profile control can be referred to as real time profile control (RTPC).Attorney DocketNo. 44024776WO01 ; 05542-1643WO1

[0029] One problem in CMP is determining the thickness of the layer using an eddy current sensing system when the layer is formed of a material with a low conductivity. For example, environmental signals and stochastic signals become significant when compared to the eddy current signals generated in the less conductive materials.

[0030] Some in-situ monitoring systems generate energy' that penetrates through the outer layer and into the underlying wafer and even into the carrier head holding the substrate. For example, a magnetic field generated by an eddy current monitoring system can pass through the outer layer, and into the wafer and possibly into the carrier head itself. This can particularly be a problem where the outer layer is formed of a layer having a relatively low conductivity. As another example, acoustic signals generated from a transducer of an acoustic monitoring system can pass through the outer layer and possibly into the wafer and the earner head before being reflected back to the acoustic sensor.

[0031] Consequently, such monitoring systems can be subject to signal distortion due to ‘noise’, i.e., undesired signal contributions, originating from the underlying wafer, or from background signals corresponding to environmental conditions or elements of the carrier head that hold the substrate. These background signals can interfere with the monitoring of the layer of interest, e.g., the conductive layer. In particular, if an eddy current sensor scans across regions which generate low signal intensity' due to the presence of low conductivity' materials, the low signal intensity can be difficult to differentiate from the background noise.

[0032] In some situations, the signal generated by conductive structures of the system can be stronger than signals generated by the low conductivity materials. For example, the carrier head, retaining ring, and slurry' can generate unwanted contributions to the detected eddy current signal. Moreover, the generated signals can vary' through time, e.g., due to changes over time in the temperature of the polishing process, such as the temperature of the slurry, the components in the carrier head, or the substrate.

[0033] A solution to these unwanted contributions is to scan a ‘blank’ calibration substrate. Such a blank calibration substrate can include a semiconductor wafer coated by a bare, i.e., unpattemed, oxide layer. Assuming the semiconductor wafer has the same doping as the wafer to be used in production, the scan of the blank calibration substrate generates a background signal that would occur in the absence of any conductive layers on the substrate.

[0034] In the context of an eddy current monitoring system, such background signals can be generated by conductive components in the carrier head, e.g., metal screws used to attach parts or metal backing plates in the retaining ring, by slurry, or by combinations thereof. Storing this background signal allows the system to use it for downstream polishing ofAttorney DocketNo. 44024776WO01 ; 05542-1643WO1 production wafers. When a production wafer is scanned, the background signal can be subtracted from the generated signal to produce a modified signal which does not include the contributions from the background. As a result, the thickness of a low conductivity film can be more precisely determined. This benefits both within-wafer uniformity (WIWU) and wafer-to-wafer uniformity (WTWU).

[0035] In addition, because each scan of the sensor across the substrate can follow a different path, the signal varies from scan-to-scan. Scan-to-scan differences relative to the underlying signal can introduce noise into the calculation of layer thickness thus reducing the precision of controlling polishing parameters, e.g. endpoint and / or polishing rate, and consequently result in large WIWNU and WIWNU. To mitigate this issue, the system can utilize methods such as scan signal averaging to reduce noise. However, there are situations where this method is ineffective, particularly when the variation from background signals from one scan to another exceeds the signal of the low conductivity film of interest.

[0036] A technique to address this issue is, for each substrate, to align the polishing head and platen consistently to a specific angular orientation relative to the eddy current sensor before commencing the chemical mechanical polishing process for that substate. Maintaining this consistent starting angular orientation on a substrate-to-substrate basis increases the likelihood that the sensor will follow a repeatable path while scanning the polishing head.

[0037] For any given scan, each substrate can experience similar slurry distribution across the surface of the substrate contacting the polishing pad. Similarly, for any given scan, the metal components of the polishing head provide a repeatable effect on the monitored signal when the head and platen are aligned.

[0038] A technique to address this issue is maintaining a library of scans of a calibration substrate. Once the platen and carrier head are aligned, creating a library of scans of the calibration substrate along a sequence of known earner head and platen alignments, and slurry distributions, allows a unique scan profile for each subsequent orientation of the platen and carrier head during the course of a production polishing process. Thus, the environmental noise from the slurry and carrier head metal components can be properly removed and the portion of the signal corresponding to only the film determined by subtracting a library scan containing the signal influence from the slurry and carrier head from the production scan determined by the in-situ monitoring system. As a result, film thickness monitoring precision is enhanced, benefiting both within-wafer uniformity (WIWU) and wafer-to-wafer uniformity (WTWU).Attorney DocketNo. 44024776WO01; 05542-1643WO1Polishing Station

[0039] FIG. 1 is a view of a polishing station 110 of a chemical mechanical polishing system. The polishing station 110 includes a polishing pad 130 supported on a rotatable platen 120. The polishing pad 130 can be a two-layer polishing pad with an outer polishing layer 132 and a softer backing layer 134.

[0040] The polishing station 110 includes a carrier head 140 positioned above the platen 120. The carrier head 140 is held by a support structure 150, e.g., a rotatable carousel or a carriage suspended from a track, that can cause the carrier head 140 to move with respect to the polishing station 110, such as between the polishing station 110 and a different polishing station, or between the polishing station 110 and a transfer station.

[0041] The polishing station 110 can include a port 160 at the end of an arm 162 to dispense polishing liquid 164, such as abrasive slurry. The port 160 is arranged dispense the liquid 164 onto the polishing surface 136 of the polishing pad 130. The polishing station 110 can also include pad conditioning apparatus 170 to abrade the polishing pad 130 to maintain the polishing surface 136 in a consistent abrasive state. For example, the conditioning apparatus can include a conditioning head 172 with a conditioning disk 176 at the end of an arm 174.

[0042] As shown in FIG. 2, the platen 120 at each platen 120 is operable to rotate about an axis 122. For example, a motor 124 can turn a drive shaft 126 to rotate the platen 120. Each motor 124 can include an encoder 125 that measures the angular position or rotation rate of the associated drive shaft 126. The drive shaft 126 can have a reference angular position that is recognized by the encoder 125 to measure the number of revolutions of the drive shaft 126.

[0043] The carrier head 140 is operable to hold a substrate 10 against the polishing pad 130. The carrier head 140 can include a retaining ring 142 to retain the substrate 10 below a flexible membrane 144. The carrier head 140 can also include a plurality of independently controllable pressurizable chambers defined by the membrane 144. The membrane 144 includes three pressurizable chambers 146a-146c which can apply independently controllable pressures to associated zones on the flexible membrane 144 and thus on the substrate 10. Although only three chambers are illustrated in FIG. 1 for ease of illustration, there could be one or two chambers, or four or more chambers, e.g., five chambers.

[0044] The carrier head 140 is suspended from the support structure 150 and connected by a drive shaft 154 to a carrier head rotation motor 156 so that the carrier head 140 can rotate about an axis 152. Optionally the carrier head 140 can oscillate laterally, e.g., driven by aAttorney DocketNo. 44024776WO01 ; 05542-1643WO1 carriage on the track, by a motor that oscillates the carrier head radially, or by rotational oscillation of the support structure 150.

[0045] In operation, the platen 120 is rotated about the platen central axis 122. The carrier head 140 is rotated about the carrier head central axis 152 and translated laterally across the polishing surface 136 of the polishing pad 130.

[0046] An in-situ monitoring system is installed in the platen 120 to monitor the progress of the polishing operation and / or measure a thickness of a layer on the substrate 10 that is being polished. The in-situ monitoring system includes an eddy current monitoring system 180. Optionally, the in-situ monitoring system can include an optical sensor, a spectrometer, a capacitive sensor, or a friction sensor.

[0047] A controller 190, such as a programmable computer, is connected to the platen motor 124 and the carrier head motor 156 to independently control the rotation rate of the platen 120 and the carrier head 140. For example, the motor 156 can include an encoder 158 that measures the angular position or rotation rate of the carrier head drive shaft 154. The drive shaft 154 can have a reference angular position that is recognized by the encoder 158 to measure the number of revolutions of the drive shaft 154.

[0048] The controller 190 is also connected to pressure regulators to control the fluid pressures in the chambers 146a-146c. In particular, the controller 190 can be configured to receive thickness measurements from the in-situ monitoring system 180 and control the pressures in the chambers 146a-146c to provide improved polishing uniformity.

[0049] The controller 190 can include a central processing unit (CPU) 192, a memory 194, and support circuits 196, e.g., input / output circuitry, power supplies, clock circuits, cache, and the like. The memory 194 is connected to the CPU 192. The memory 194 is a non- transitory computable readable medium and can be one or more readily available memory such as random-access memory (RAM), read only memory (ROM), floppy disk, hard disk, or other form of digital storage. In addition, although illustrated as a single computer, the controller 190 could be a distributed system, e.g., including multiple independently operating processors and memories.Eddy Current Monitoring System

[0050] The in-situ-monitoring system includes an eddy current monitoring system 180. The eddy current monitoring system 180 includes a drive system to induce eddy currents in a conductive layer on the substrate and a sensing system to detect eddy currents induced in the conductive layer by the drive system. The eddy current monitoring system 180 includes aAttorney DocketNo. 44024776WO01 ; 05542-1643WO1 core 182 positioned in a recess 184 to rotate with the platen, at least one coil 183 wound around a portion of the sensor head 181. and drive and sense circuitry- 186 connected by wiring 187 to the coil 183. The combination of the core 182 and coil 183 can provide a sensor head 181. In some implementations, the core 182 can project above the top surface of the platen 120, e.g., into a recess 184 in the bottom of the polishing pad 130.

[0051] The drive and sense circuitry 186 is configured to apply an oscillating electric signal to the coil 183 and to measure the resulting eddy current. The drive and sense circuitry- 186 can be located in the same recess 184 or a different portion of the platen 120 or could be located outside the platen 120 and be coupled to the components in the platen 120 through the motor 124.

[0052] In operation, the drive and sense circuitry 186 drives the coil 183 to generate an oscillating magnetic field. At least a portion of magnetic field extends through the polishing pad 130 and into substrate 10. If a conductive layer is present on 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 186. As the thickness of the conductive layer changes, the impedance changes, and this can be detected by the drive and sense circuitry 186.

[0053] The signals can pass from the eddy current monitoring system 180 to the controller 190 through the rotary electrical union 123. Alternatively, the circuitry 186 could communicate with the controller 190 by a wireless signal.

[0054] The polishing stations 110 can also include a position sensor 188, such as an optical interrupter, to sense when the sensor head 181 is beneath the substrate 10. For example, the optical interrupter could be mounted at a fixed point opposite the carrier head 140. A flag 189 is attached to the periphery of the platen. The point of attachment and length of flag 189 is selected so that it interrupts the optical signal of sensor 188 while the sensor head 181 sweeps beneath substrate 10.

[0055] Since the sensor head 181 sweeps beneath the substrate with each rotation of the platen, information on the conductive layer thickness is accumulated in-situ and on a continuous real-time basis (once per platen rotation). The controller 190 can be programmed to sample measurements from the monitoring system 180 when the substrate 10 generally overlies the sensor head 181 (as determined by the position sensor). The measurements from the monitoring system 180 can be displayed on the output device 193 during polishing to permit the operator of the device to visually monitor the progress of the polishing operation, although this is not required.Attorney DocketNo. 44024776WO01 ; 05542-1643WO1

[0056] Assuming the thickness of the layer varies across the substrate 10, the change in the position of the sensor head 181 with respect to the substrate 10 can result in a change in the signal from the in-situ monitoring system 180. The time varying sampled signal resulting from a single sweep of the sensor head 181 below the substrate 10 may be referred to as a trace. Variation in the signal across a trace can indicate variation in the layer thickness across the substrate 10. In addition, as polishing progresses, the thickness of the conductive layer changes, and the sampled signals vary with time. Trace-to-trace differences can indicate variation in the layer thickness over time.

[0057] In operation, the polishing station 110 can use the eddy current monitoring system 180 to determine when the bulk of a filler layer has been removed and / or to determine when an underlying layer has been substantially exposed. Possible process control and endpoint criteria for the detector logic include local minima or maxima, changes in slope, threshold values in amplitude or slope, or combinations thereof.

[0058] The controller 190 can be programmed to divide the measurements from the eddy- current monitoring system 180 from each sweep beneath the substrate into a plurality- of sampling zones, to calculate the radial position of each sampling zone, and to sort the amplitude measurements into radial ranges, as discussed in U.S. Pat. No. 6,399,501. After sorting the measurements into radial ranges, information on the film thickness can be fed in real-time into a closed-loop controller to periodically or continuously modify the polishing pressure profile applied by a carrier head in order to provide improved polishing uniformity.

[0059] The controller 190 can use a correlation curve that relates the signal measured by the in-situ monitoring system 180 to the thickness of the layer being polished on the substrate 10 to generate an estimated measure of the thickness of the layer being polished. An example of a correlation curve 404 is shown in FIG. 2A. In the coordinate system depicted in FIG. 2A, the horizontal axis represents the value of the signal received from the in-situ monitoring system 180, whereas 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 404 to generate a corresponding thickness value. The correlation curve 404 can be considered a "‘static’7formula, in that it predicts a thickness value for each signal value regardless of the position at which the sensor head obtained the signal. The correlation curve can be represented by a variety of functions, such as a polynomial function, or a look-up table (LUT) combined with linear interpolation.Attorney DocketNo. 44024776WO01 ; 05542-1643WO1Signal Processing

[0060] Changes in the position of the sensor head 181 with respect to the substrate 10 can result in a change in the signal from the in-situ monitoring system 180. That is, as the sensor head scans across the substrate 10, the in-situ monitoring system 180 will make measurements at several measurement spots at different locations on the substrate 10 for multiple regions.

[0061] FIG. 3A shows a schematic top view of a substrate 10 being scanned by a sensor head of a polishing apparatus for one scan path 510. The scan path starts at position 512 on the leading edge of the substrate 10 and ends at position 513 on the trailing edge of the substrate 10. As noted above, changes in the position of the sensor head with respect to the substrate 10 can result in a change in the signal from the in-situ monitoring system 180. That is, as the sensor head scans across the substrate 10, the in-situ monitoring system 180 will make measurements for multiple regions, e.g., measurement spots 511, at different locations on the substrate 10.

[0062] FIG. 2B illustrates a graph (show n for illustration of the process; no graph need be generated or displayed in operation) that shows a simplified signal 401 from the in-situ monitoring system 180 during a single pass of the sensor head below the substrate 10.

[0063] Referring to FIGS. 3A and 2B, for a given trace, the signal can be captured (and thus the graph can represent the signal) as a function of measurement of position, e.g., radial or diameter 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 depicts, for a given location of the substrate scanned by the sensor head, a corresponding measured signal value from the signal 401. Although the signal 401 is illustrated as a continuous line, in practice the signal 401 will be composed of a sequence of individual signal values.

[0064] Referring to FIGS. 3A and 2B, the signal 401 includes a first portion 422 that corresponds to locations in an edge region 503 of the substrate 10 when the sensor head crosses a leading edge of the substrate 10, a second portion 424 that corresponds to locations in a central region 501 of the substrate 10, and a third portion 426 that corresponds to locations in edge region 503 when the sensor head crosses a trailing edge of the substrate 10. The signal can also include portions 428 that correspond to off-substrate measurements, i.e., signals generated when the sensor head scans areas beyond the edge 504 of the substrate 10 in FIG. 3A.Attorney DocketNo. 44024776WO01 ; 05542-1643WO1

[0065] The edge region 503 can correspond to a portion of the substrate where measurement spots 94 of the sensor head overlap the substrate edge 504. The sensor head may scan these regions on its path 510 and generate a sequence of measurements that correspond to a sequence of locations along the path 510.

[0066] Although the second portion 424 is illustrated as flat, this is for simplicity, and a real signal in the second portion 424 would likely include fluctuations due both to noise and to variations in the layer thickness and contributions from background signals. The second portion 424 corresponds to the monitoring location scanning the central region 501.

[0067] During a polishing process, the sensor conducts multiple scans, with each scan measuring the thickness at various radial positions across the substrate. Due to the differing rotation rate of the carrier head 140 and the platen 120, and due to the lateral oscillation of the earner head 140, the scan paths are not consistent. As illustrated in FIG. 3B, the carrier head 140 is shown having been scanned by a sensor head along five distinct paths, 510, 520, 530, 540, and 550. Because the multiple scan paths 510-550 are distinct, they cover areas with differing background signals, resulting in scan-to-scan variation due to radial variations in blank calibration signal, variations in how the sensors generate signals from the carrier head 140, or variations in polishing liquid temperature or distribution.

[0068] For example, the carrier head 140 is shown having a gimbal 401 located centrally and two screws 402 and 403 located radially from the center of the carrier head 140. Signals generated from scan paths which intercept the gimbal 401 or screws 402 and 403 include comparatively high background signals associated with the gimbal 401 or screws 402 and 403, particularly in the context of signals generated from films including low7conductivity materials.

[0069] FIG. 4A illustrates a graph (shown for illustration of the process; no graph need be generated or displayed in operation) that shows a signal trace 611 from the eddy current monitoring system 180 along the scan path 510 of the sensor head 181 beneath the substrate 10. As noted above, the trace 611 can be stored as a function of position of the measurement on the substrate, e.g., a radial position. The fluctuations in the trace 611 can be due to variation in the thickness across the substrate of the layer being polished, noise, or the presence of other objects in which the sensor head 181 produces eddy currents.

[0070] For some polishing processes, a portion of the signal strength of the trace 611 attributable to the background can be sufficiently high to make the thickness measurement based on the trace 611 unreliable. For example, the portion of the signal strength of the trace 611 attributable to the background can be comparable to, or even larger than, the portion ofAttorney DocketNo. 44024776WO01 ; 05542-1643WO1 the signal produced by the layer being polished. In this example, the signal trace 611 has multiple background signals that are of greater signal magnitude than the overlaying layer. The signal trace 611 displays peak 612 and peak 613 which can correspond with the sensor head 181 passing beneath a conductive region at an outer edge of the carrier head 140, e.g., a metal backing plate in the retaining ring. The signal trace 611 displays peak 614 and peak 615, which can correspond to the sensor head 181 passing beneath a metallic screw 402 or 403 in the housing of the carrier head 140. The signal trace 611 displays peak 616 which can correspond to the sensor head 181 passing beneath a conductive insert in the gimbal 401 of the carrier head 140. However, other variations are possible.

[0071] Although only one trace is illustrated in FIG. 4 A, the system can acquire two or more traces with each trace obtained from a single pass of a sensor. Multiple traces can be obtained through a single sensor with multiple scans, multiple sensors with a single scan, or multiple sensors with multiple scans, such as scans corresponding with the scan paths 510, 520, 530, or 540 of FIG. 3B. In some examples, a number of traces are obtained above what would be a maximum number of scans for a production wafer. In this manner, the library of background traces is greater than the maximum number of production scans thus ensuring a background trace for every production scan.

[0072] Another source of background signal is due to the temperature of the polishing process increasing over time as the substrate 10 is polished on the polishing station 110. The increasing temperature includes the temperature of the polishing pad 130, the temperature of the liquid 164, the temperature of the substrate 10, the temperature of the interface between the substrate 10 and the polishing surface 136. An example temperature curve of the polishing pad 130 is shown inset to FIG. 4B. While the temperature of the pad typically would not directly affect the signals generated by the eddy-current monitoring system 180. the temperature of the pad is indicative of the amount of heat generated during polishing and thus indicative of the temperature of the substrate 10 and polishing liquid 164 at the interface, and possibly metal components heated due to conductive heat transfer, and any changes in these temperatures do modify conductivity and thus affect the signals generated by the eddy- current monitoring system 180.

[0073] As the temperature changes, the traces generated by the eddy current monitoring system 180 change in response. FIG. 4B shows three exemplary traces illustrating signals generated along the same path of FIG. 4A. Signal trace 617, signal trace 618, and signal trace 619 are generated at points A, B. and C, respectively, on the temperature curve.Attorney DocketNo. 44024776WO01 ; 05542-1643WO1

[0074] Another source of background signal variation is due to the distribution of polishing liquid between the substrate 10 and the polishing pad 130 as the substrate 10 is polished on the polishing station 110. Generally, the polishing liquid 164 is dispensed for each substrate or polishing process while the platen 120 rotates. The polishing liquid 164 traverses the polishing pad 130 under the effects of centrifugal force and ingresses under the carrier head 140 and between the substrate 10 and the polishing pad 140. The polishing liquid 164 distribution path can be impacted by the position of the conditioner head on the polishing pad 130, the position of the carrier head 140, and the position of the dispensing port 16-. At a given radius on the platen 120, the distribution of the polishing liquid 164 is different. Moreover, because the elements of the polishing system move over time, the distribution of the polishing liquid 164 on the polishing pad 130 and under the substrate 10 can change over time during the polishing of a particular substrate. In addition, the point at which the polishing liquid 164 is dispensed relative to a specific sensor position can change from run to run if platen 120 and carrier head 140 are not homed, e.g., brought to the respective predetermined relative positions, resulting in substrate-to-substrate variations in the distribution of the polishing liquid.

[0075] As such, the signal traces can vary region by region as the polishing progresses. In an example, the amount of polishing liquid 164 between the edge region 503 and the polishing pad 130 is different than the amount of polishing liquid 164 between the central region 501 and the polishing pad 130. Furthermore, these amounts can change over time, e.g., due to spread of the polishing liquid 164 over the polishing pad 130 over time, and the change in signal based on the presence of polishing liquid 164 changes with both quantity and temperature of the polishing liquid 164. As such, signals collected at the edge region 503 are more greatly affected by changes in polishing liquid 164 quantity, temperature, and composition.

[0076] To address the issue of relatively high signal ratio between background signals and thickness signals, the controller 190 can incorporate a data processing program which subtracts traces representing background signals, e.g.. background traces, on a scan-by-scan basis. These background traces represent scans of a blank calibration substrate.

[0077] In general, a ‘’blank” substrate is a substrate that is unpattemed, i.e., includes a semiconductor wafer and either no layers thereon or only unpattemed dielectric layers. An example blank calibration substrate can include a lightly-doped semiconductor wafer. Lightly-doped, low-doped, or “undoped” semiconductor wafers are wafers made from semiconductor material that are either not intentionally doped or have been intentionallyAttorney DocketNo. 44024776WO01 ; 05542-1643WO1 doped with other materials, such as phosphorus or boron, but at low levels. This low level can be characterized by resistivity. A resistivity value for a lightly-doped wafer is about 100 microohm-centimeter (pohm-cm).

[0078] Undoped silicon wafers can be characterized by low conductivity, resulting in low background signal contributions from an in-situ eddy current monitoring system. In some implementations, the blank calibration substrate consists of the semiconductor wafer. However, in some implementations, the blank calibration wafer includes one or more unpattemed dielectric layers (but no conductive layers) deposited on the wafer. For example, the blank calibration substrate can include an oxide layer (e.g., a thermal oxide layer), coating the wafer. Briefly, thermal oxidation processes produce a thermal oxide layer e.g., S1O2. which can be useful in providing insulation layers on the substrate 10. In some implementations, the blank calibration substrate consists of the semiconductor wafer and a single unpattemed dielectric layer, e.g., the oxide layer.

[0079] In some implementations, the production substrate includes a conductive layer overlying a patterned dielectric layer, e.g.. a patterned oxide layer. Some production substrates include conductive layers formed of a highly conductive (e.g.. copper). However, there are also applications in which production substrates include conductive layers formed of low conductivity7materials. Such low conductivity' materials generate eddy currents having a lower relative magnitude than high conductivity materials and thus lower detectable signals by the drive and sense circuitry 186. These low conductivity materials’ signal could be in similar or lower level than environmental signal generated by slurry or head metal, therefore subtracting environmental signal using this method is critical for these materials. Whereas for high conductivity materials, the environmental signal is relatively small and can be ignored.

[0080] An example of a high conductivity material is copper (Cu). Some non-limiting examples of materials having a low relative conductivity include titanium nitride (TiN), tungsten doped carbon (WDC), molybdenum (Mo), tungsten (W), cobalt (Co), titanium (Ti), tantalum (Ta), tantalum nitride (TaN), tungsten nitride (WN), or combinations of materials including these examples. Highly conductive materials can have a resistivity value of 300 Ohm Angstroms (OhmA) or less (e g., 300 Ohm- Angstrom = 3 pOhm-cm). One example is copper (Cu) which has a resistivity' value of about 180 OhmA, or aluminum (Al) which has a resistivity' value of about 280 OhmA. In one non-limiting example, the systems and processes provided herein can be used with a substrate 10 having a conductive layer having a resistivity higher than 800 OhmA, e.g., higher than 1000 OhmA. W, Co, and Mo haveAttorney DocketNo. 44024776WO01 ; 05542-1643WO1 similar resistivity values which can be in range from 1000 OhmA to 3000 OhmA. Less conductive materials, such as Ta, TaN. Ti, TiN, WDC. or WN have resistivity values of about 6000 OhmA or higher, e.g., 20000 OhmA, or 30000 OhmA. Briefly, materials having lower resistivity values (e.g., Cu) are more highly conductive materials than materials with higher resistivity' values (e.g., TiN or WDC).

[0081] In some implementations, subtracting background signals includes sequentially labeling both the measured traces and the background traces, starting from the first scan (e.g., labeled as #1) and progressing through to the final scan at the end of the polishing process (e.g., labeled as #N, where N is the total number of scans conducted during a CMP process). The background trace is subtracted from the measured trace bearing the equivalent label number on a scan-by-scan basis.

[0082] In general, during a CMP process, the one or more sensors are repeatedly swept across a substrate to generate a sequence of measured traces. The controller 190 labels the measured trace. In some implementations, the label can be determined based on the ordinal position of the measured trace in the sequence of measured trace. For example, the first measured trace produced during an initial pass of a sensor is assigned the label #1. the second measured trace from the second pass of a sensor is labeled as #2, and so forth, until the polishing process is completed. If there are a total of N sensor scans throughout the entire polishing procedure, the measured traces are labeled from #1 to #N. Referring back to FIG. 4A, the measured trace 611, generated during the scan path 510, can be labeled as #1, whereas a measured trace originating from the scan path 520, can be labeled as #2. Other methods of labeling can be used, so long as it is possible to determine a background trace that occurred at the same scan ordinal position in the sequence of scans as the measured trace.

[0083] The controller 190 obtains a corresponding background trace from a digital library. In particular, the controller 190 can obtain the background trace having the same label as the measured trace.

[0084] Assuming the library of background traces is properly generated, the corresponding background trace would be the one generated using the same scan path as the measured scan trace. For example, FIG. 4C illustrates the background trace derived from the scan path which generated FIG. 4A. Specifically, the background trace 620 originates from the scan path 510 from the first scan (e.g., label #1), and will be the corresponding background trace for the first measured trace 611 (e.g., also label #1). Similarly, the second background trace 752 is generated from the scan path 520 from the second scan (e.g.. label #2), and thus will the corresponding background trace for the measured signal trace 712 (e.g., also label #2).Attorney DocketNo. 44024776WO01; 05542-1643WO1

[0085] The controller 190 subtracts (604) the background trace with the equivalent label from the measured trace to generate a modified trace. The subtraction process is executed on a scan-to-scan basis with equivalent positions in the sequence of signal traces. For example, returning to FIGS. 3A and 4C, the background signal trace 620 is subtracted from the measured signal trace 611 because they share the same position, #1, in the sequence of traces; likewise, the background signal trace 752 is subtracted from the measured signal trace 712 due to their equivalent position, #2. If there are a total of n scans, this identical subtraction process is carried out for each scan.

[0086] The measured signal trace and the background signal trace can have the same labeling system. For example, both measured signal traces and background signal traces can be labeled based on its ordinal position within the sequence of traces. An ordinal position of each trace in the sequence of traces can be provided by ordering of the data. For example, as noted above, the background trace 620 and the measured signal trace 611 can be labeled #1 because they are the first entry in the sequence of traces, generated from the first scan path 510; the background trace 752 and the measured signal trace 712 can be labeled #2 because they are the second entry in the sequence of traces, generated from the second scan path 520.

[0087] The subtraction can be conducted immediately after each sensor scan or following a specified number of sensor scans, which can be defined within the controller 190. For example, if the subtraction is set to be conducted immediately after each scan, then following the first scan path, the background signal trace produced by the scan path #1 is retrieved from the library and then subtracted from the measured signal trace for the scan path # 1 ; likewise, after the second scan path, the background signal trace produced by the scan path #2 is retrieved and then subtracted from the measured signal trace for the scan path #2, and so on.

[0088] If the specified number of scans is set to 2. which means that the subtraction is done after every 2 sensor scans, but the subtraction is still performed on a scan-to-scan basis. For example, after the first scan #1, a subtraction is not performed; but after the second scan #2, the background signal traces for both scan paths #1 and #2 are retrieved together from the library. The background signals are then subtracted from the measured signal traces respectively on scan-by-scan basis, that is. background signal #1 is subtracted from measured signal trace #1, background signal #2 is subtracted from measured signal trace #2.

[0089] In some implementations, the apparatus 100 can be configured to allow for changes in the number of scans required prior to subtraction at different stages of the polishing, making it adaptable to varying conditions. For example, the apparatus 100 can be set to perform subtraction after every 5 scans at the beginning of a polishing process, w hen initialAttorney DocketNo. 44024776WO01 ; 05542-1643WO1 film thickness deviates significantly from a target thickness. As the polishing progresses, especially when measured thickness approaches the target thickness, the apparatus 100 can change to perform the subtraction immediately after each scan for more timely calculation.

[0090] In some implementations, the controller 190 is configured to perform an orientation process on the platen 120, the carrier head 140, or both. The orientation process can be termed ‘homing.’ Homing the platen 120 or carrier head 140 includes rotating the platen 120, the earner head 140. or both to a specified relative angular position before commencing polishing of the substrate 10. Having the platen 120 and the carrier head 140 in known relative angular positions with one another allows for consistent positioning of detectable components of the carrier head 140 which could cause signal variations during the polishing process. Thus, the signal received for a particular scan path, e.g.. scan 510, will have consistent background signal contributions from the carrier head 140 on a substrate-to- substrate basis..

[0091] Having the platen 120 and the carrier head 140 start each polishing operation in a known relative angular relationship also ensures that, assuming the processing conditions such as carrier head sweep, conditioner head sweep are the same, evolution of the distribution of the polishing liquid under the substrate 10 over time should also be the same. Thus, contributions of the polishing liquid for a particular scan should be the same on a substrate- to-substrate basis.

[0092] The controller 190 can be configured such that the earner head 140 and platen 120 are consistently aligned at respective specified angular orientations before the polishing process begins for any substrate. Referring to FIG. 3C, which is a schematic top view of a CMP polishing station, the controller 190 causes the carrier head 140 to rotate to a predetermined angular orientation in a stationary frame of reference, e.g.. the frame of reference of the base that holds the platen 120 pnor to polishing. For example, the predetermined angular orientation 0i, can be relative to an axis 195 that passes through both the axis 122 of the platen 120 and the center 197 of the carrier head 140. In some implementations, this axis can pass through the fixed-position sensor 188. The sensor head 181 can be fixed on the rotating platen 120 such that the sensor head 181 rotates with the platen 120. The predetermined angular orientation of the platen 120, 02, can be relative to the axis 195.

[0093] With both the platen and the carrier head 140 consistently beginning at specific angular orientations, the sensor head 181 should follow a consistent series of paths across the carrier head 140 on a substrate-to-substrate basis. Returning to FIG. 3B, provided that theAttorney DocketNo. 44024776WO01 ; 05542-1643WO1 relative angular position between the carrier head 140 and the sensor head 181 remains consistent from one substrate to another, the initial scan path 510 can consistently commence at the path start position 512 and conclude at the path ending position 513 for each substrate, while for the second scan, the scan path 520 can consistently commence at the path start position 522 and conclude at the path ending position 523 for each substrate. Likewise, for the third, fourth and subsequent scans, the starting and ending positions of each path would also remain consistent from one substrate to another. Consequently, for equivalent positions within the sequence of scans, the sensor's path exhibits repeatability from substrate to substrate.

[0094] Referring back to FIG. 4C, if each consecutive background trace, e.g.. trace 620, is generated with the same angular alignment as the corresponding measured traces, e.g., trace 611, they will trace the same scan paths 510 and 520. This angular alignment of the carrier head 140 and the platen 120 before polishing results in a more accurate subtraction of the underlying signal trace on a scan-by-scan basis.

[0095] For example, FIG. 4D illustrates modified signal traces using background signal subtractions on a scan-by-scan basis, when both the platen 120 and the carrier head 140 consistently begin at specified angular positions. For example, the modified signal trace 661 is obtained by subtracting the background trace 620 from the measured signal 611.Employing background subtraction on a scan-by-scan basis, the modified signal traces offer more accurate representation of ground truth values for the target layer (see FIGS. 4B and 4C).

[0096] The specified angular position can vary among different polishing stations 110. Each polishing stations 110 of the apparatus 100 may each employ a distinct angular alignment for the carrier head 140 and the platen 120. Nevertheless, the angular positioning remaining consistent from one polishing process to another for the same polishing station 110 increases modified signal trace accuracy within each polishing station 110.

[0097] FIG. 5 is a flow-diagram of an example for creating a background signal library. A blank calibration substrate is prepared (702). The blank calibration substrate can be an undoped, or low-doped, substrate suitable for use as a production substrate but prior to deposition of conductive or dielectric layers.

[0098] The controller 190 causes the carrier head 140 and platen 120 to rotate to respective positions (704) such that the carrier head 140 and the platen 120 are at respective predetermined angular positions.Attorney DocketNo. 44024776WO01; 05542-1643WO1

[0099] The platen 120 rotates to repeatedly sweep (706) one or more sensors across the blank calibration substrate to generate a sequence of background traces during a "‘simulated polishing” process. In this simulated polishing process, the carrier head and platen rotate at the same speeds and the carrier had rotates at the same rate as will occur during regular polishing of a product substrate. Similarly, the same polishing liquid can be supplied at the same flow rate as will occur during regular polishing of a product substrate. The thickness of the exposed top layer of the blank calibration substrate should remain unchanged throughout the simulated polishing process. In addition, the pressure in the carrier head can be reduced, e.g., the chambers can be vented to atmosphere, to avoid polishing.

[0100] Process parameters for polishing the blank calibration substrate are set to mirror those of a real polishing process, including, but not limited to, the platen 120 rotation rate, the substrate 10 rotation rate, the angular positions of platen 120, angular position of the arm 162 and associated port 160, polishing liquid 164 flow rate, the angular position of the carrier head 140, or the carrier head 140 sweep posit on. The polishing duration can be set to equal to or longer than that of a real polishing process to encompass all potential unique scan paths generated during a real polishing process. Consequently, for a given product, the total number of background traces in the digital 1 ibrary may equal or exceed that of an actual polishing on a production substrate.

[0101] The controller 190 labels (708) the background traces sequentially from each sensor, such as #1. #2. #3, and so forth, until the simulated polishing process finishes. As noted above, alternative labeling methods can also be utilized.

[0102] The controller 190 stores (712) each individual background trace. The stored background traces provide a library of background traces. In some implementations, each trace is stored as a separate digital file. If there are a total of N traces generated throughout the entire “simulated polishing” process, then there will be a total of N files, with labels from #1 to #N. The file format includes, but is not limited to, a Data Stream Interface (DSI) file.

[0103] In some implementations, there are two or more sensors of the in-situ monitoring system 180 which are supported on the rotating platen 120. Each sensor can travel a distinct path within a single rotation of the platen, and the labeling of the signal traces is specific to each sensor and each scan path. Therefore, a signal trace's position in the sequence of measured traces is contingent upon both the sensor and the scan path it follows. To illustrate, if there are three sensors, and each sensor generates 100 scan paths over the entire polishing process, a total of 300 signal traces would be created, and these can be sequentially labeled from #1 to #300. Consequently, corresponding background traces for the given product areAttorney DocketNo. 44024776WO01; 05542-1643WO1 also generated using the same configuration of three sensors, resulting in a digital library with a minimum of 300 traces. When performing scan-by-scan subtraction, the subtraction process involves matching the position in the sequence of background traces and measured traces to produce a sequence of 300 modified signal traces.

[0104] In some implementations, where multiple sensor heads are supported on the platen, a scan averaging method can be employed to consolidate the modified signal traces collected from the multiple within a single platen rotation. Using the example mentioned earlier, for each platen rotation, three sensors generate three measured signal traces. By subtracting the background signal trace on a scan-by-scan basis from these three measured signal traces, three modified signal traces are obtained. Subsequently, the average of these three modified signal traces is computed, resulting in a single averaged modified signal trace. This averaged trace represents the mean value for that particular platen rotation. If the entire polishing process comprises a total of 100 platen rotations, this scan averaging method would yield 100 averaged traces. These 100 averaged traces are then used to calculate the instant thickness of the substrate 10. In conjunction with scan-by-scan subtraction method, the scan averaging method can further diminish background noise.

[0105] In addition or alternatively, a scan averaging method can be applied prior to the subtraction of background signals. Specifically, an averaged trace is generated by averaging the measured traces gathered from different sensors within a single platen rotation, and then the corresponding background trace is subtracted from the averaged traces. To illustrate using the previously mentioned example, for each platen rotation, three sensors produce three measured signal traces. Before subtraction the average of these three measured signal traces is firstly computed, yielding a single averaged signal trace. Subsequently, the respective background trace is subtracted from this averaged signal trace to yield a modified signal trace. The respective background trace can be obtained by averaging the three corresponding background traces. In other w ords, the subtraction is achieved by deducting an averaged background trace from an averaged measured signal for each platen rotation. The resulting modified signal trace is then employed to calculate the instant thickness of the substrate 10.

[0106] In some implementations, the polishing apparatus 100 can apply edge reconstruction. Edge construction refers to the process of controlling and optimizing the removal of material at the edges of a substrate during the polishing process. The edge of a substrate can present unique challenges during CMP because it is more susceptible to non- uniform material removal and damage compared to the central region. Edge construction can involve various techniques and strategies to achieve uniformity and minimize edge defects. InAttorney DocketNo. 44024776WO01 ; 05542-1643WO1 addition or alternatively, the polishing apparatus 100 can detect a leading edge of the trace and a trailing edge of the trace. The polishing apparatus 100 can be configured to detect such edges.

[0107] In some implementations, the edge reconstruction employs a neural network, as described in U.S. Patent Publication Nos. 2018-0304435. As illustrated in FIG. 2B, the variation in the signal intensity in the portions 422, 426 is caused in part by measurement region of the sensor overlapping the substrate edge, rather than an intrinsic variation in the thickness or conductivity' of the layer being monitored. Consequently, this distortion in the signal 401 can cause errors in the calculating of a characterizing value for the substrate, e.g., the thickness of the layer, near the substrate edge.

[0108] To address this problem, the controller 190 can include a neural network to generate a modified signal corresponding to one or more locations of the substrate 10 based on the measured signals corresponding to those locations. The neural network is configured to, when trained appropriately, generate modified signals that reduce and / or remove the distortion of computed signal values near the substrate edge. The system obtains estimated measures of thickness generated by the neural network based on input values that include measured signals for each location in a group of locations of the substrate. The system also obtains ground truth measures of thickness for each location in the group of locations. The system can generate ground truth measures of thickness using an electrical impedance measuring method, such as the four-points probe method.

[0109] The system then computes a measure of error between the estimated measures of thickness and the ground truth measures of thickness, and updates one or more parameters of the neural network based on the measure of error. If the polishing apparatus uses such a neural network to generate modified signals based on the measured signals generated by the in-situ monitoring system, the apparatus can compensate for the distortions, e.g., reduced signal strength, at the substrate edge. In some embodiments, the system performs the edge reconstruction before, or as part of, the signal conversion to thickness values, e.g., the neural network performs a conversion to thickness values and then performs edge reconstruction.

[0110] Referring back to FIG. 5. after obtaining modified signal traces using various signal processing methods as mentioned above, the controller 190 calculates estimated thickness values based on modified traces. Each estimated thickness value represents the thickness of the layer being polished at the time the measured trace was acquired.

[0111] For a given modified signal value, the polishing apparatus 100 can use the correlation curve 404 to generate a corresponding thickness value. With the correlation curveAttorney DocketNo. 44024776WO01; 05542-1643WO1404, a sequence of estimated thickness values can thus be generated using modified signal values. Each thickness profile includes thickness values as a function of time or radial position on the substrate 10.

[0112] FIG. 6 is a flow-diagram of an example method 800 of chemical mechanical polishing of a substrate 10. Data representing a sequence of background traces generated by a sequence of scans of a sensor of an in-situ eddy current monitoring system across a blank calibration substrate are stored (802), e.g.. in a memory of the controller 190. An example method 700 for generating and storing data representing a digital library of background traces is given in FIG. 5.

[0113] Optionally, the substrate 10 and the platen 120 are positioned at respective predetermined angular positions prior to polishing. The polishing apparatus 100 polishes (804) a layer on the substrate 10 and monitors (806) the layer during the polishing to generate measured signal values for different locations on the layer, e.g., using the in-situ monitoring system 180. Monitoring includes repeatedly sweeping the sensor of the in-situ eddy current monitoring system across the substrate to generate a measured trace including a sequence of raw signal values. Repeatedly sweeping the sensor provides a sequence of measured traces.

[0114] Background traces are subtracted from the measured signal traces on a scan-by-scan basis (808). As described above, this can include determining the background trace that has the same ordinal position in the scanning sequence as the measured trace and subtracting that background trace from the measured trace. As noted above, this scan-by-scan subtraction method can be further combined with other signal processing methods, such as, a scan averaging method by taking an average of multiple signal traces, a neutral network configured to generate modified signals that reduce the distortion of computed signal values near the substrate edge, scaling the signal trace to match position such that each trace extends along a same start time or position to a same end time or position, etc.

[0115] Subtracting the respective background trace from the respective measured trace generates a corrected thickness trace for each location of the different locations based on the difference between the measured and background traces. Each corrected thickness trace for each location then includes a sequence of corrected thickness values.

[0116] A polishing endpoint is detected, a polishing parameter is modified, or both, (810) based on each corrected thickness trace. Polishing parameters which the polishing station 110 can modify based on the corrected thickness traces include, but are not limited to, a chamber pressure, a platen rotation rate, a carrier head rotation rate, a slurry dispensing rate, or a combination thereof.Attorney DocketNo. 44024776WO01; 05542-1643WO1

[0117] 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 provide relative motion between the polishing surface and the substrate. The polishing pad can be a circular (or some other shape) pad secured to the platen, a tape extending between supply and take-up rollers, or a continuous belt. The polishing pad can be affixed on a platen, incrementally advanced over a platen between polishing operations, or driven continuously over the platen during polishing. The pad can be secured to the platen during polishing, or there can be a fluid bearing between the platen and polishing pad during polishing. The polishing pad can be a standard (e g., polyurethane with or without fillers) rough pad, a soft pad, or a fixed-abrasive pad.[Oi l 8] Although the discussion above focuses on an eddy current monitoring system, the correction techniques can be applied to other sorts of monitoring systems in which there is a sensor that scans the substrate and generates energy that can pass through the outer layer to the wafer and carrier head, e.g., active acoustic monitoring systems. Also, although the techniques have been generally discussed as advantageous for low conductivity materials, if demands for accuracy are high enough it may be appropriate to subtract out background signals even from scans of a high conductively material. In addition, although the discussion above focuses on a polishing system, the correction techniques can be applied to other sorts of substrate processing systems, e.g., deposition or etching systems, which include an in-situ monitoring system with a sensor that scans across the substrate.

[0119] A number of embodiments of the invention have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the invention. Accordingly, other embodiments are within the scope of the following claims.

Claims

Attorney DocketNo. 44024776WO01 ; 05542-1643WO1CLAIMSWhat is claimed is:

1. A method of chemical mechanical polishing, comprising: storing data representing a sequence of background traces generated by a sequence of scans of a sensor of an in-situ eddy current monitoring system across a blank calibration substrate; polishing a conductive layer on a device substrate; monitoring the layer during the polishing with the in-situ eddy current monitoring system, including repeatedly sweeping the sensor of the in-situ eddy current monitoring system across substrate such that each sweep generates a sequence of raw signal values, and providing a measured trace from each sequence of raw signal values such that the repeated sweeping provides a sequence of measured traces; for each respective measured trace in the sequence of measured traces, subtracting a respective background trace that has an equivalent position in the sequence of background traces from the respective measured trace to generate a corrected thickness trace and thus provide a sequence of corrected thickness traces; and at least one of detecting a polishing endpoint or modifying a polishing parameter that affects the polishing based on the sequence of corrected thickness traces.

2. The method of claim 1, wherein each measured trace is a thickness profile including thickness values as a function of radial position on the device substrate.

3. The method of claim 2, comprising converting the sequence of raw signal values to thickness values using a correlation curve to provide a preliminary thickness trace.

4. The method of claim 3. comprising performing edge reconstruction on preliminary thickness trace by processing portions of the preliminary thickness trace corresponding to an edge of the device substrate through a neural network to generate a modified thickness trace that provides the measured trace.

5. The method of claim 4. wherein each background trace represents a sequence of raw signal background values that are converted to effective thickness values using theAttorney DocketNo. 44024776WO01 ; 05542-1643WO1 correlation curi e to provide a preliminary effective thickness trace which is then subjected to edge reconstruction by having portions corresponding to an edge of the calibration substrate processed through a neural network to generate a modified effective thickness trace that provides the background trace.

6. The method of claim 1. wherein the blank calibration substrate consists of a semiconductor substrate.

7. The method of claim 1, wherein the blank calibration substrate comprises a lightly doped semiconductor substrate.

8. The method of claim 1, wherein the blank calibration substrate comprises an unpattemed dielectric layer coating a semiconductor substrate.

9. The method of claim 8. wherein the unpattemed dielectric layer is an oxide layer.

10. The method of claim 9, wherein the oxide layer is a thermal oxide.

11. The method of claim 1. comprising rotating a platen supporting a polishing pad for the polish of the substrate to a first predetermined angular position prior to polishing the layer on the substrate.

12. The method of claim 11, comprising counting a number of rotations of the platen to provide an ordinal position of the measured trace in the sequence of measured traces.

13. The method of claim 12, comprising measuring an angular orientation of a carrier head prior to polishing, and rotating the carrier head to a first predetermined angular position prior to polishing.

14. The method of claim 11, comprising storing data representing an ordinal position of each background trace in the sequence of background traces.Attorney DocketNo. 44024776WO01 ; 05542-1643WO115. The method of claim 11, wherein an ordinal position of each background trace in the sequence of background traces is provided by ordering of the data.

16. The method of claim 1, comprising instructions to average traces accumulated from different sensors within a single rotation of a platen to generate an averaged trace.

17. The method of claim 16, comprising instructions to subtract the respective background trace from the averaged trace.

18. The method of claim 1. wherein the conductive layer has a resistivity higher than 1000 OhmA.

19. The method of claim 18, wherein the conductive layer comprises titanium nitride or tungsten doped carbon.

20. A non-transitory computer readable medium comprising instructions encoded therein providing a computer program for controlling a polishing system and causing one or more computers to: store data representing a sequence of background traces generated by a sequence of scans of a sensor of an in-situ eddy current monitoring system across a blank calibration substrate; receive from the in-situ eddy current monitoring system, a sequence of raw signal values for each sweep by the sensor across the substrate, each sequence of raw signal values providing a measured trace such that the repeated sweeping provides a sequence of measured traces; for each respective measured trace in the sequence of measured traces, subtract a respective background trace that has an equivalent position in the sequence of background traces from the respective measured trace to generate a corrected thickness trace and thus provide a sequence of corrected thickness traces; and at least one of i) detect a polishing endpoint or ii) modify a polishing parameter that affects the polishing based on the sequence of corrected thickness traces.Attorney DocketNo. 44024776WO01 ; 05542-1643WO121. The computer readable medium of claim 20, wherein each measured trace is a thickness profile including thickness values as a function of radial position on the device substrate.

22. The computer readable medium claim 21, comprising instructions to convert the sequence of raw signal values to thickness values using a correlation curve to provide a preliminary’ thickness trace.

23. The computer readable medium of claim 22, comprising instructions to perform edge reconstruction on preliminary thickness trace by processing portions of the preliminary’ thickness trace corresponding to an edge of the device substrate through a neural network to generate a modified thickness trace that provides the measured trace.

24. The computer readable medium of claim 23, wherein each background trace represents a sequence of raw signal background values that are converted to effective thickness values using the correlation curve to provide a preliminary effective thickness trace which is then subjected to edge reconstruction by having portions corresponding to an edge of the calibration substrate processed through a neural network to generate a modified effective thickness trace that provides the background trace.

25. The computer readable medium of claim 20, comprising instructions to rotate a platen supporting a polishing pad for the polish of the substrate to a first predetermined angular position prior to polishing the layer on the substrate.

26. The computer readable medium of claim 25, comprising instructions to count a number of rotations of the platen to provide an ordinal position of the measured trace in the sequence of measured traces.

27. The computer readable medium of claim 26. comprising instructions to measure an angular orientation of a carrier head prior to polishing, and rotating the carrier head to a first predetermined angular position prior to polishing.Attorney DocketNo. 44024776WO01 ; 05542-1643WO128. The computer readable medium of claim 25, comprising instructions to store data representing an ordinal position of each background trace in the sequence of background traces.

29. The computer readable medium of claim 25, wherein an ordinal position of each background trace in the sequence of background traces is provided by ordering of the data.

30. The computer readable medium of claim 20, comprising instructions to average traces accumulated from different sensors within a single rotation of a platen to generate an averaged trace.

31. The computer readable medium of claim 30, comprising instructions to subtract the respective background trace from the averaged trace.

32. A polishing system, comprising: a platen to support a polishing pad; a carrier head to hold a substrate in contact with the polishing pad; a motor to generate relative motion between the carrier head and the platen; an eddy current monitoring system to monitor the substrate during polishing, the eddy current monitoring system including a sensor positioned to repeatedly sweep across the substrate such that each sweep generates a sequence of raw signal values, each sequence of raw signal values providing a measured trace such that repeated sweeping provides a sequence of measured traces; and a controller configured to: store data representing a sequence of background traces generated by a sequence of scans of a sensor of an in-situ eddy current monitoring system across a blank calibration substrate. receive from the in-situ eddy current monitoring system a sequence of raw signal values for each sweep by the sensor across the substrate, each sequence of raw signal values providing a measured trace such that the repeated sweeping provides a sequence of measured traces, for each respective measured trace in the sequence of measured traces, subtract a respective background trace that has an equivalent position in the sequence ofAttorney DocketNo. 44024776WO01 ; 05542-1643WO1 background traces from the respective measured trace to generate a corrected thickness trace and thus provide a sequence of corrected thickness traces, and at least one of i) detect a polishing endpoint or ii) modify a polishing parameter that affects the polishing based on the sequence of corrected thickness traces.

33. A method, comprising: aligning a platen and a carrier head that is holding a blank calibration substrate by rotating the platen, the carrier head, or both, such that the carrier head and the platen are at a predetermined angular position; generating a sequence of blank calibration signal traces by sweeping one or more sensors of an in-situ monitoring system arranged within a recess of the platen across the blank calibration substrate between the platen and the carrier head; labeling the sequence of blank calibration signal traces according to the sequence by which the blank calibrations signal traces are generated; creating a data file for each blank calibrations signal of the sequence of blank calibration signal traces; and storing each data file for each blank calibration signal trace to provide a library of stored blank calibration signal traces.

34. The method of claim 33, comprising measuring an angular orientation of the carrier head prior to polishing, and rotating the carrier head to a first predetermined angular position prior to polishing.

35. The method of claim 34, comprising rotating a platen supporting a polishing pad for the polish of the blank calibration wafer to a second predetermined angular position prior to polishing a layer on the blank calibration wafer.

36. The method of claim 35, comprising counting a number of rotations of the platen to provide an ordinal position of the measured trace in the sequence of measured traces.

Images