WO2019060577A1 - Semiconductor fabrication using machine learning approach to generating process control parameters - Google Patents

Semiconductor fabrication using machine learning approach to generating process control parameters Download PDF

Info

Publication number
WO2019060577A1
WO2019060577A1 PCT/US2018/052001 US2018052001W WO2019060577A1 WO 2019060577 A1 WO2019060577 A1 WO 2019060577A1 US 2018052001 W US2018052001 W US 2018052001W WO 2019060577 A1 WO2019060577 A1 WO 2019060577A1
Authority
WO
WIPO (PCT)
Prior art keywords
substrate
parameter values
neural network
artificial neural
control parameter
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/US2018/052001
Other languages
French (fr)
Other versions
WO2019060577A4 (en
Inventor
Sivakumar Dhandapani
Jun Qian
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Applied Materials Inc
Original Assignee
Applied Materials Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Applied Materials Inc filed Critical Applied Materials Inc
Priority to JP2020516913A priority Critical patent/JP7036908B2/en
Priority to KR1020207011805A priority patent/KR102450925B1/en
Priority to CN201880062109.4A priority patent/CN111133560B/en
Publication of WO2019060577A1 publication Critical patent/WO2019060577A1/en
Publication of WO2019060577A4 publication Critical patent/WO2019060577A4/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10P—GENERIC PROCESSES OR APPARATUS FOR THE MANUFACTURE OR TREATMENT OF DEVICES COVERED BY CLASS H10
    • H10P52/00—Grinding, lapping or polishing of wafers, substrates or parts of devices
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B24—GRINDING; POLISHING
    • B24B—MACHINES, DEVICES, OR PROCESSES FOR GRINDING OR POLISHING; DRESSING OR CONDITIONING OF ABRADING SURFACES; FEEDING OF GRINDING, POLISHING, OR LAPPING AGENTS
    • B24B37/00—Lapping machines or devices; Accessories
    • B24B37/005—Control means for lapping machines or devices
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B24—GRINDING; POLISHING
    • B24B—MACHINES, DEVICES, OR PROCESSES FOR GRINDING OR POLISHING; DRESSING OR CONDITIONING OF ABRADING SURFACES; FEEDING OF GRINDING, POLISHING, OR LAPPING AGENTS
    • B24B37/00—Lapping machines or devices; Accessories
    • B24B37/11—Lapping tools
    • B24B37/20—Lapping pads for working plane surfaces
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B24—GRINDING; POLISHING
    • B24B—MACHINES, DEVICES, OR PROCESSES FOR GRINDING OR POLISHING; DRESSING OR CONDITIONING OF ABRADING SURFACES; FEEDING OF GRINDING, POLISHING, OR LAPPING AGENTS
    • B24B37/00—Lapping machines or devices; Accessories
    • B24B37/27—Work carriers
    • B24B37/30—Work carriers for single side lapping of plane surfaces
    • B24B37/32—Retaining rings
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
    • G05B23/0243—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model
    • G05B23/0254—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10P—GENERIC PROCESSES OR APPARATUS FOR THE MANUFACTURE OR TREATMENT OF DEVICES COVERED BY CLASS H10
    • H10P52/00—Grinding, lapping or polishing of wafers, substrates or parts of devices
    • H10P52/40—Chemomechanical polishing [CMP]
    • H10P52/403—Chemomechanical polishing [CMP] of conductive or resistive materials
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10P—GENERIC PROCESSES OR APPARATUS FOR THE MANUFACTURE OR TREATMENT OF DEVICES COVERED BY CLASS H10
    • H10P72/00—Handling or holding of wafers, substrates or devices during manufacture or treatment thereof
    • H10P72/04—Apparatus for manufacture or treatment
    • H10P72/0428—Apparatus for mechanical treatment or grinding or cutting
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10P—GENERIC PROCESSES OR APPARATUS FOR THE MANUFACTURE OR TREATMENT OF DEVICES COVERED BY CLASS H10
    • H10P72/00—Handling or holding of wafers, substrates or devices during manufacture or treatment thereof
    • H10P72/06—Apparatus for monitoring, sorting, marking, testing or measuring
    • H10P72/0604—Process monitoring, e.g. flow or thickness monitoring
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10P—GENERIC PROCESSES OR APPARATUS FOR THE MANUFACTURE OR TREATMENT OF DEVICES COVERED BY CLASS H10
    • H10P72/00—Handling or holding of wafers, substrates or devices during manufacture or treatment thereof
    • H10P72/06—Apparatus for monitoring, sorting, marking, testing or measuring
    • H10P72/0612—Production flow monitoring, e.g. for increasing throughput
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10P—GENERIC PROCESSES OR APPARATUS FOR THE MANUFACTURE OR TREATMENT OF DEVICES COVERED BY CLASS H10
    • H10P74/00—Testing or measuring during manufacture or treatment of wafers, substrates or devices
    • H10P74/23—Testing or measuring during manufacture or treatment of wafers, substrates or devices characterised by multiple measurements, corrections, marking or sorting processes
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10P—GENERIC PROCESSES OR APPARATUS FOR THE MANUFACTURE OR TREATMENT OF DEVICES COVERED BY CLASS H10
    • H10P74/00—Testing or measuring during manufacture or treatment of wafers, substrates or devices
    • H10P74/23—Testing or measuring during manufacture or treatment of wafers, substrates or devices characterised by multiple measurements, corrections, marking or sorting processes
    • H10P74/238—Testing or measuring during manufacture or treatment of wafers, substrates or devices characterised by multiple measurements, corrections, marking or sorting processes comprising acting in response to an ongoing measurement without interruption of processing, e.g. endpoint detection or in-situ thickness measurement
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10P—GENERIC PROCESSES OR APPARATUS FOR THE MANUFACTURE OR TREATMENT OF DEVICES COVERED BY CLASS H10
    • H10P95/00—Generic processes or apparatus for manufacture or treatments not covered by the other groups of this subclass
    • H10P95/04—Planarisation of conductive or resistive materials
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10P—GENERIC PROCESSES OR APPARATUS FOR THE MANUFACTURE OR TREATMENT OF DEVICES COVERED BY CLASS H10
    • H10P95/00—Generic processes or apparatus for manufacture or treatments not covered by the other groups of this subclass
    • H10P95/06—Planarisation of inorganic insulating materials
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10P—GENERIC PROCESSES OR APPARATUS FOR THE MANUFACTURE OR TREATMENT OF DEVICES COVERED BY CLASS H10
    • H10P74/00—Testing or measuring during manufacture or treatment of wafers, substrates or devices
    • H10P74/20—Testing or measuring during manufacture or treatment of wafers, substrates or devices characterised by the properties tested or measured, e.g. structural or electrical properties
    • H10P74/203—Structural properties, e.g. testing or measuring thicknesses, line widths, warpage, bond strengths or physical defects

Definitions

  • the present disclosure relates to controlling an integrated circuit fabrication process, e.g., a chemical mechanical polishing process.
  • An integrated circuit is typically formed on a substrate by the sequential deposition of conductive, semiconductive, or insulative layers on a silicon wafer.
  • One fabrication step involves depositing a filler layer over a non-planar surface and planarizing the filler layer.
  • the filler layer is planarized until the top surface of a patterned layer is exposed.
  • a conductive filler layer can be deposited on a patterned insulative layer to fill the trenches or holes in the insulative layer.
  • 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.
  • the filler layer is planarized until a predetermined thickness is left over an underlying layer.
  • a dielectric layer deposited can be planarized for
  • CMP Chemical mechanical polishing
  • This planarization method typically requires that the substrate be mounted on a carrier head.
  • the exposed surface of the substrate is typically placed against a rotating polishing pad with a durable roughened surface.
  • the carrier head provides a controllable load on the substrate to push it against the polishing pad.
  • a polishing liquid such as a slurry with abrasive particles, is typically supplied to the surface of the polishing pad.
  • the processed wafer exhibits a material removal profile, a two dimensional map of the change in a polished layer's thickness after the polishing process.
  • a method of processing substrates includes: subjecting each respective first substrate of a first plurality of substrates to a process that modifies a thickness of an outer layer of the respective first substrate; for each respective first substrate, recording a group of process parameter values used for the process, thus generating a plurality of groups of process parameter values; for each respective first substrate, measuring a removal profile of the outer layer during or after the process with a monitoring system, thus generating a plurality of removal profiles; training an artificial neural network by backpropagation using the plurality of groups of process parameter and plurality of removal profiles as training data, the artificial neural network having a plurality of input nodes to receive respective removal values from the removal profile, a plurality of output nodes to output control parameter values, and a plurality of hidden nodes connecting the input nodes to the output nodes; for each respective second substrate of a second plurality of substrates, determining a target removal profile; for each respective second substrate, determining respective control parameter values to apply to the respective second substrate from the output nodes of the artificial neural
  • a computer program product for controlling processing of a substrate includes instructions for causing a processor to: train an artificial neural network by backpropagation using a plurality of groups of process parameter and a plurality of removal profiles as training data, the artificial neural network having a plurality of input nodes to receive respective removal values from a removal profile, a plurality of output nodes to output control parameter values, and a plurality of hidden nodes connecting the input nodes to the output nodes; obtain a target removal profile for each respective substrate of a plurality of substrates; for each respective substrate, determine respective control parameter values to apply to the respective substrate from the output nodes of the artificial neural network by applying the target removal profile to the input nodes of the artificial neural network; and cause a semiconductor processing system to subject each respective substrate to the process using the respective control parameter values.
  • a polishing system includes: a support to hold a polishing pad; a carrier head to hold a substrate in contact with the polishing pad, the carrier head having a plurality of chambers; a motor to generate relative motion between the support and the carrier head; and a controller configured to: train an artificial neural network by backpropagation using a plurality of groups of process parameter and a plurality of removal profiles as training data, the artificial neural network having a plurality of input nodes to receive respective removal values from a removal profile, a plurality of output nodes to output control parameter values, and a plurality of hidden nodes connecting the input nodes to the output nodes, wherein the control parameter values include pressures for the chambers in the carrier head; obtain a target removal profile for each respective substrate of a plurality of substrates; for each respective substrate, determine respective control parameter values to apply to the respective substrate from the output nodes of the artificial neural network by applying the target removal profile to the input nodes of the artificial neural network, wherein the respective control parameter values include respective pressures
  • Implementations may include one or more of the following features.
  • the process may include chemical mechanical polishing.
  • the control parameter values may include pressures of chambers in a carrier head to hold a substrate against a polishing pad.
  • At least some of the input nodes of the artificial neural network may be configured to receive state parameter values.
  • the state parameter values may include one or more of a retaining ring life, or a polishing pad life.
  • the control parameter values may include one or more of a platen rotation rate, carrier head rotation rate, or polishing time.
  • the monitoring system may include an in-line metrology system.
  • the in-line metrology system may include an optical monitoring system.
  • the determining of the target removal profile may include storing a desired thickness profile, receiving a measured thickness profile of the respective second substrate, and determining a difference between the measured thickness profile and the desired thickness profile.
  • the receiving of the measured thickness profile may include measuring a thickness profile of the respective second substrate with the monitoring system.
  • Control parameter convergence may be achieved more rapidly.
  • Within-wafer thickness non- uniformity and wafer-to-wafer thickness non-uniformity WIWNU and WTWNU
  • Product wafers may be used for model refinement, leading to the actual profile of processed substrates being closer to a desired profile.
  • Complex behavior of the polishing process in response to polishing parameters may be accounted for without explicit knowledge of the functional relationships.
  • FIG. 1 illustrates a schematic cross-sectional view of an example of a polishing apparatus.
  • FIG. 2 illustrates a schematic top view of a substrate having multiple zones.
  • FIG. 3 illustrates a neural network used as part of the integrated advanced process control module of the polishing apparatus.
  • FIG. 4 illustrates an example process for using a neural network model in a CMP process.
  • FIG. 5 illustrates an integrated advanced process control module that includes multiple neural networks.
  • One challenge in CMP is developing a process model that is able to predict a material removal profile as a function of multiple input parameters.
  • Input parameters may include variations in the initial thickness of a substrate layer, a target material removal profile, the polishing pad condition, the retaining ring condition, the relative speed between the polishing pad and a substrate, and the applied pressure on a substrate.
  • the CMP process is a dynamical system in which surface topology is constantly changing as a function of time during the process, which in turn affects the process behavior.
  • the complex relationship of input parameters with the material removal profile and the dynamic nature of a CMP process contribute to the challenge of developing a process model of the polishing process.
  • FIG. 1 illustrates an example of a polishing apparatus 20.
  • the polishing apparatus 20 can include a rotatable disk-shaped platen 22 on which a polishing pad 30 is situated.
  • the platen is operable to rotate about an axis 23.
  • a motor 24 can turn a drive shaft 26 to rotate the platen 22.
  • the polishing pad 30 can be detachably secured to the platen 22, for example, by a layer of adhesive.
  • the polishing pad 30 can be a two-layer polishing pad with an outer polishing layer 32 and a softer backing layer 34.
  • the polishing apparatus 20 can include a polishing liquid supply port 40 to dispense a polishing liquid 42, such as an abrasive slurry, onto the polishing pad 30.
  • the polishing apparatus 20 can also include a polishing pad conditioning disc to abrade the polishing pad 30 to maintain the polishing pad 30 in a consistent abrasive state.
  • a carrier head 50 is operable to hold a substrate 10 against the polishing pad 30.
  • Each carrier head 50 also includes a plurality of independently controllable pressurizable chambers, e.g., three chambers 52a-52c, which can apply independently controllable pressurizes to associated zones 12a-12c on the substrate 10 (see FIG. 2).
  • the center zone 12a can be substantially circular, and the remaining zones 12b- 12c can be concentric annular zones around the center zone 12a.
  • the chambers 52a-52c can be defined by a flexible membrane 54 having a bottom surface to which the substrate 10 is mounted.
  • the carrier head 50 can also include a retaining ring 56 to retain the substrate 10 below the flexible membrane 54.
  • FIGS. 1 and 2 Although only three chambers are illustrated in FIGS. 1 and 2 for ease of illustration, there could be two chambers, or four or more chambers, e.g., five chambers.
  • other mechanisms to adjust the pressure applied to the substrate e.g., piezoelectric actuators, could be used in the carrier head 50.
  • Each carrier head 50 is suspended from a support structure 60, e.g., a carousel or track, and is connected by a drive shaft 62 to a carrier head rotation motor 64 so that the carrier head can rotate about an axis 51.
  • each carrier head 50 can oscillate laterally, e.g., on sliders on the carousel, by motion along the track; or by rotational oscillation of the carousel itself.
  • the platen 22 is rotated about its central axis 23, and the carrier head 50 is rotated about its central axis 51 and translated laterally across the top surface of the polishing pad 30.
  • the polishing apparatus 20 can include an in-line monitoring system for measuring a thickness of a polished layer upon completion of the polishing process.
  • the inline monitoring system can generate a map of layer thicknesses over the zones 12a-12c.
  • the measurements from the in-line monitoring system can be communicated to a controller 90.
  • Examples of the in-line monitoring system include an optical monitoring system, e.g., a spectrographic monitoring system.
  • a spectrographic monitoring system can measure a layer thickness value and an associated goodness of fit (GOF) value for each points of the layer thickness map.
  • a broadband light source can be used to illuminate a location on the layer, and the reflection containing an optical interference spectrum created by the layer, the substrate, and any other layers in between the two can be measured.
  • the measured optical interference spectrum can be analyzed, for example, by fitting it with an equation that describes an optical interference spectrum generated by an expected stacking of films.
  • the fitting produces a determination of the thickness of the layer and a GOF value that is indicative of how closely the measured spectrum agrees with the expected film stack.
  • the GOF value can be used as an indicator of reliability of the determined thickness value.
  • the controller 90 contains a tool control module 92 and an integrated advanced process control module (i-APC) 94.
  • the tool control module 92 and the i-APC module 94 in combination may provide advanced process control functionalities, such as wafer-to-wafer uniformity control.
  • the controller 90 can be a computing device that includes a
  • microprocessor memory
  • input/output circuitry e.g., a programmable computer.
  • controller 90 can be a networked system with functions distributed across multiple computers, and modules 92 and 94 can be located in the same computer or different computers.
  • the described polishing apparatus has many associated process parameters that control the operation of the polishing apparatus or describe the state of the apparatus or the polishing environment.
  • Process parameters that control the operation of the polishing apparatus (and that can be set, at least initially, by the tool control module 92) include the following: rotation rate of platen 22; rotation rate of carrier head 50; pressure of the chambers 52a-52c; and polishing time.
  • Process parameters that reflect the state of the apparatus or polishing environment include the following: polishing retaining ring life; polishing pad life; polishing pad conditioning disc life; type of polishing pad; and type of polishing liquid ('slurry').
  • the retaining ring, pad, and conditioning disc are examples of consumable components ('consumables') within a polishing apparatus. The condition or 'life' of these consumables can be described, for example, as a count of wafers processed; actual wafer polishing time; or total time elapsed since installation.
  • Atypical wafer to be polished by a CMP process has a layer of material with surface topologies to be planarized.
  • a goal of the CMP process is to achieve a desired material removal profile, which is a one or two dimensional map of the change in a polished layer's thickness after the polishing process.
  • the surface topologies vary from wafer-to-wafer due to different die designs having different underlying transistor and interconnect patterns, e.g. due to different pattern densities. These factors interact with the CMP process in a complex manner, which results in different polishing behavior between wafers with different die designs. Furthermore, even wafers with the same die design may have different polishing behaviors due to upstream process variations in deposition or etching.
  • CMP control parameters are typically needed, at least for wafers with different die designs, and possibly for individual wafers with the same die design, to achieve a desired material removal profile.
  • Material removal profiles sometimes have a radial dependence partly due to the axial symmetry and rotation of the polishing head. Accordingly, one control parameter that is often tuned to achieve a desired radial profile is the pressures of the radial chambers 52a-52c.
  • the polishing apparatus 20 can implement a wafer-to-wafer control.
  • the wafer-to- wafer control can provide improved likelihood of achieving target material removal profiles over a wide range of designs and wafer non-uniformities.
  • a wafer-to-wafer feedback control method uses information about previously processed substrates to improve processing of a subsequent substrate.
  • the wafer-to-wafer feedback control method can be implemented by the i-APC module 94.
  • the i-APC module 94 generates an initial set of control parameter values based on an initial process model, and provides the initial set of control parameters values to the tool control module 92.
  • the tool control module 92 can then control the polishing system using the received control parameter values.
  • data about the substrates can be used by the i-APC module 94 to improve the initial process model by generating a new process model or updating the initial process model.
  • the user defined period of time for generation of a new process model or updating of the initial process model can be measured in number of substrates polished, such as 5, 10, 25, or 100 wafers, or a "lot" of wafers (wafers are typically transported and processed in "lots"; a typical lot size is 25 wafers).
  • the user defined period of time may be defined as a portion of a lifetime of a consumable, such as a polishing pad lifetime.
  • Such generation or updating of the model can be performed off-line, or independent of the operation of the polishing apparatus.
  • the improved process model can then be used to generate an improved set of control parameter values for a subsequent substrate. Improvement of the process model by the i- APC module 94 can lead to improved polishing uniformity control by minimizing error between the target material removal profile and the actual profile realized on the substrate.
  • the i-APC module 94 performs tasks including collecting and processing of data from processed wafers to improve processing of future wafers.
  • Collected data can include upstream and downstream metrology data of the product wafers from various wafer metrology tools ("monitoring systems").
  • Upstream metrology data may include thickness map and associated GOF values of a deposited layer.
  • Downstream metrology data may include a thickness map and associated GOF values of the polished layer, or surface roughness values.
  • the GOF values can be used by the i-APC module 94 to determine whether the thickness value are reliable enough to be used in the process model development.
  • the i-APC module 94 pairs these data with the control and state parameters used during the processing of a particular wafer, and stores it in a data log.
  • the data log may be organized in a variety of ways, including grouping by wafer ID, design ID, lot ID, tool ID, etc. This data log is typically used to monitor trends and drifts in the behavior of a CMP process and to take corrective actions. Due to their size, data logs may be stored in one or more servers that are a part of the controller 90.
  • the i-APC module 94 stores a desired thickness profile for each wafer to be processed. Using the desired thickness profile, the i-APC module 94 can generate the target material removal profile by subtracting a desired thickness profile from the thickness map of the deposited layer.
  • a process model can be generated in various ways.
  • a process model for describing the effects of chamber pressures can be generated by processing multiple blanket wafers, which are un-patterned wafers with a uniform layer of a film to be polished, as a proxy for product wafers that have patterns. By polishing the blanket wafers for a fixed amount of time and measuring the resulting material removal profile, a relationship between a set of chamber pressures and a material removal rate can be determined using the Preston equation.
  • Material removal rate (MRR) K P *V*P Equation 1
  • V is the velocity of the polishing pad surface with respect to the substrate surface being polished
  • P is the pressure applied to a radial zone of a wafer to be polished against the polishing pad
  • K p is a proportionality constant known as the Preston coefficient.
  • the velocity, pressure, and polish time are known and controlled variables, so K p of each zone can be determined with algebraic manipulations.
  • the resulting set of Preston equations containing the Preston coefficients from polishing of blanket wafers is an example of a baseline process model.
  • the i-APC module 94 can then use this baseline process model to generate a set of initial control parameters for processing of subsequent product wafers. While a process model for modeling effects of chamber pressures are discussed, process models for other control parameters can be generated in this manner.
  • the baseline process model might not achieve the target material removal profile on the polished product wafer.
  • the i-APC 94 attempts to find, through multiple iterations over polishing of multiple product wafers, a set of correction factors ("offsets") for improving the achieved material removal profile on the product wafers.
  • the i- APC module 94 eventually determines, or converges on, offsets that can be used with the baseline process model for polishing the product wafer.
  • offsets can take polishing of multiple product wafers, during which the polished wafers are less likely to meet the target removal profile thus more likely to be rejected, adversely impacting yield of the polishing process.
  • the i-APC module 94 is configured to provide the improved process model.
  • One way of improving the quality of the process model is to implement a process model that takes in as inputs (1) target material removal profile and (2) state parameters that reflect the conditions of the polishing apparatus, and outputs a set of estimated control parameters (e.g. chamber pressures) that would achieve the target material removal profile.
  • the material removal rate is influenced by state parameters 114 of the polishing apparatus such as the life, or age, of various consumables due to their wear.
  • the polishing pad may experience a reduction in elasticity or a reduction in surface roughness due to aging.
  • the retaining ring may get polished away during the CMP process, becoming thinner.
  • the polishing pad conditioning disc may become dull and have reduced capacity for pad conditioning.
  • a neural network can be used to model behavior (e.g., removal profile) of a semiconductor processing system (e.g., a CMP system) in response to the process parameters.
  • the neural network is based on a large collection of simple neural units
  • Each individual neural unit computes a respective output, or activation, using a summation function.
  • Such threshold function may include a hyperbolic tangent or a sigmoid function.
  • a neural network is trained, rather than explicitly programmed.
  • a neural network does not require a priori knowledge of the functional dependence of the process being modeled. Instead, the neural network can be trained by providing it with a sufficiently large and representative set of training data; this technique is called supervised machine learning. Because a neural network can excel at modeling complex relationships between inputs and the outputs that are difficult to express in functional relationships, it can provide significant advantages for control of a semiconductor processing system, such as a CMP system.
  • FIG. 3 illustrates the functional blocks implemented by the i-APC module 94.
  • the i- APC module 94 includes neural network input variable set 1 10, which includes target removal profile 1 12 and state parameters 1 14, a neutral network 120, and a process initialization module 130 that initializes the tool control module 92 of a semiconductor processing system, e.g., a CMP tool, with the output of the neural network.
  • a semiconductor processing system e.g., a CMP tool
  • the neutral network 120 includes a plurality of input nodes 122 for each input variable from neural network input variable set 1 10, a plurality of hidden nodes 124 (also called “intermediate nodes” below), and a plurality output nodes 126 that will generate the process parameters to be used for initialization.
  • each hidden node 124 can be coupled to each input node 122, and each output node 126 can be coupled to each hidden node 124.
  • a hidden node 124 outputs a value that a non-linear function of a weighted sum of the values from the input nodes 122 to which the hidden node is connected.
  • node k the output of a hidden node 124, designated node k, can be expressed as:
  • Equation 2 tanh (0.5* ⁇ 3 ⁇ 4 1 ( ⁇ 1 )+ ⁇ 3 ⁇ 4 2 ( ⁇ 2 )+ ... + ⁇ 3 ⁇ 4 ⁇ ( ⁇ )+ 3 ⁇ 4) Equation 2 where tanh is the hyperbolic tangent, aia is a weight for the connection between the k th intermediate node and the x th input node (out of M input nodes), and IM is the value at the M* h input node.
  • other non-linear functions can be used instead of tanh, such as a rectified linear unit (ReLU) function and its variants, or a sigmoid function.
  • ReLU rectified linear unit
  • FIG. 4 illustrates an example process 400 for training and using a neural network model in a CMP process.
  • the neutral network 120 need to be configured before being used to generate baseline control parameters.
  • the neural network 120 may be configured through training by using various methods of supervised learning.
  • One example of supervised learning method is backpropagation, the use of forward stimulation to reset weights on the 'front' neural units, or input nodes 122.
  • the i-APC control module 94 collects sufficient amount of training data (410).
  • the training data comprises a set of input and output pairs that have been measured, and therefore known to be true.
  • data log maintained by the i-APC control module 94 can be used to retrieve for each substrate, the state parameters 114 during the process; the target removal profile 112; the control parameters used during the process; and the measured removal profile.
  • the retrieved information is then analyzed to determine its suitability as a training example.
  • the measured removal profile is compared with the target profile. If the difference between the two is sufficiently small to be within the specification of the process, the data is determined to be a suitable training example.
  • the CMP polish behavior may vary substantially between substrates with different designs. Therefore, it may be beneficial to further categorize the training data into different die designs to be used for creating design-specific neural networks. Once a sufficient amount of such training examples have been collected, the training begins.
  • the i-APC control module 94 trains the neural network 120 (420), e.g., by backpropagation.
  • Backpropagation is a two phase cycle of propagation and weight update.
  • the input vector e.g. target removal profile and state parameters from a training example
  • the output of the network is then compared to the recorded output (e.g. control parameter from a training example) using a cost function, and an error value is calculated for each of the neurons in the output layer.
  • a cost function or loss function is a function that maps values of one or more variables onto a real number intuitively representing some 'cost' associated with the event.
  • the error values are then propagated backwards, starting from the output, until each neuron has an associated error value which roughly represents its contribution to the original output.
  • Backpropagation uses these error values to calculate the gradient of the loss function with respect to the weights in the network.
  • back propagation method uses these error values to update the weights, in an attempt to minimize the loss function.
  • the i-APC control module 94 determines whether the trained neural network is accurate (430). After the training phase 420 is completed, the neural network's accuracy should be verified. Verification of the accuracy, for example, may include operating the neural network in its inference mode, feeding it with previously used training input, and determining whether the output is in agreement, or is sufficiently close to the correct output in the training data. If the neural network output is not sufficiently accurate, then the process returns to step 410 to collect additional training data. If the output is accurate, then the neural network is ready to be used.
  • the i-APC control module 94 uses the trained neural network (440) to generate process parameters.
  • the neural network outputs process parameters to be used for initialization in response to target material removal profile 112 and state parameters 114.
  • individual neural network is generated and trained for each wafer design.
  • the i-APC control module 94 maintains a library of neural networks, and the controller selects a neural network trained for the current wafer design to initialize the process.
  • the architecture of the neural network 120 can vary in depth and width.
  • the neural network 120 is shown with a single column of intermediate nodes 124, it can include multiple columns.
  • the number of intermediate nodes 124 can be equal to or greater than the number of input nodes 122.
  • the i-APC control module 94 can be configured to have multiple neural networks 120.
  • the neural networks 120 can operate in parallel to generate the chamber pressure values for each zones.
  • the number of neural networks can match the number of zones. For example, there can be a first neural network 120a for the first zone 12a, a second neural network 120b for the second zone 12b and a third neural network for the third zone 12c.
  • the output of each neural network 120 can be fed to the process initialization module 130.
  • the training of the neural network can be computationally intensive. Accordingly, in some implementations, training of the neural network may be done off-line, e.g., while the CMP tool is undergoing maintenance. In other implementations, the training can be done on a separate server that is part of the controller 90. In some other implementations, the training can be done in a server that is not part of the controller 90, which provides the pre-trained neural network to the controller.
  • the behavior of a CMP process may slowly change over time for reasons that are not factored in (e.g. wear of parts that are not considered consumables) by the input variables of the neural network. Therefore, in some implementations, a periodic retraining of the neural network may be performed to overcome such changes.
  • Embodiments and all of the functional operations described in this specification can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the structural means disclosed in this specification and structural equivalents thereof, or in combinations of them.
  • Embodiments can be implemented as one or more computer program products, i.e., one or more computer programs tangibly embodied in a machine readable storage media, for execution by, or to control the operation of, data processing apparatus, e.g., a programmable processor, a computer, or multiple processors or computers.
  • a computer program also known as a program, software, software application, or code
  • a computer program does not necessarily correspond to a file.
  • a program can be stored in a portion of a file that holds other programs or data, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub programs, or portions of code).
  • a computer program can be deployed to be executed on one computer or on multiple computers at one site or distributed across multiple sites and interconnected by a communication network.
  • the processes and logic flows described in this specification can be performed by one or more programmable processors executing one or more computer programs to perform functions by operating on input data and generating output.
  • the processes and logic flows can also be performed by, and apparatus can also be implemented as, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit).
  • the above described polishing apparatus and methods can be applied in a variety of polishing systems.
  • Either the polishing pad, or the carrier heads, or both can move to provide relative motion between the polishing surface and the substrate.
  • the platen may orbit rather than rotate.
  • the polishing pad can be a circular (or some other shape) pad secured to the platen.
  • the polishing system can be a linear polishing system, e.g., where the polishing pad is a continuous or a reel-to-reel belt that moves linearly.
  • the polishing layer can be a standard (for example, polyurethane with or without fillers) polishing material, a soft material, or a fixed-abrasive material. Terms of relative positioning are used relative orientation or positioning of the components; it should be understood that the polishing surface and substrate can be held in a vertical orientation or some other orientation with respect to gravity.

Landscapes

  • Engineering & Computer Science (AREA)
  • Mechanical Engineering (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • General Physics & Mathematics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Automation & Control Theory (AREA)
  • Mechanical Treatment Of Semiconductor (AREA)
  • Finish Polishing, Edge Sharpening, And Grinding By Specific Grinding Devices (AREA)
  • Constituent Portions Of Griding Lathes, Driving, Sensing And Control (AREA)
  • Feedback Control In General (AREA)

Abstract

A method for processing substrates includes subjecting each respective first substrate of a first plurality of substrates to a process that modifies a thickness of an outer layer of the respective first substrate, generating a plurality of groups of process parameter values; generating a plurality of removal profiles, training an artificial neural network by backpropagation using the plurality of groups of process parameter and plurality of removal profiles as training data where the artificial neural network has a plurality of input nodes to receive respective removal values from the removal profile and a plurality of output nodes to output control parameter values, for each respective second substrate of a second plurality of substrates determining a target removal profile, determining respective control parameter values to apply by applying the target removal profile to the input nodes, and subjecting each respective second substrate to the process using the respective control parameter values.

Description

SEMICONDUCTOR FABRICATION USING MACHINE LEARNING APPROACH TO GENERATING PROCESS CONTROL
PARAMETERS
TECHNICAL FIELD
The present disclosure relates to controlling an integrated circuit fabrication process, e.g., a chemical mechanical polishing process.
BACKGROUND
An integrated circuit is typically formed on a substrate by the sequential deposition of conductive, semiconductive, or insulative layers on a silicon wafer. One fabrication step involves depositing a filler layer over a non-planar surface and planarizing the filler layer. For some applications, the filler layer is planarized until the top surface of a patterned layer 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. 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. For other applications, the filler layer is planarized until a predetermined thickness is left over an underlying layer. For example, a dielectric layer deposited can be planarized for
photolithography.
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 typically placed against a rotating polishing pad with a durable roughened surface. The carrier head provides a controllable load on the substrate to push it against the polishing pad. A polishing liquid, such as a slurry with abrasive particles, is typically supplied to the surface of the polishing pad. The processed wafer exhibits a material removal profile, a two dimensional map of the change in a polished layer's thickness after the polishing process.
SUMMARY
In one aspect, a method of processing substrates includes: subjecting each respective first substrate of a first plurality of substrates to a process that modifies a thickness of an outer layer of the respective first substrate; for each respective first substrate, recording a group of process parameter values used for the process, thus generating a plurality of groups of process parameter values; for each respective first substrate, measuring a removal profile of the outer layer during or after the process with a monitoring system, thus generating a plurality of removal profiles; training an artificial neural network by backpropagation using the plurality of groups of process parameter and plurality of removal profiles as training data, the artificial neural network having a plurality of input nodes to receive respective removal values from the removal profile, a plurality of output nodes to output control parameter values, and a plurality of hidden nodes connecting the input nodes to the output nodes; for each respective second substrate of a second plurality of substrates, determining a target removal profile; for each respective second substrate, determining respective control parameter values to apply to the respective second substrate from the output nodes of the artificial neural network by applying the target removal profile to the input nodes of the artificial neural network; and subjecting each respective second substrate to the process using the respective control parameter values.
In another aspect, a computer program product for controlling processing of a substrate, the computer program product tangibly embodied in a non-transitory computer readable medium, includes instructions for causing a processor to: train an artificial neural network by backpropagation using a plurality of groups of process parameter and a plurality of removal profiles as training data, the artificial neural network having a plurality of input nodes to receive respective removal values from a removal profile, a plurality of output nodes to output control parameter values, and a plurality of hidden nodes connecting the input nodes to the output nodes; obtain a target removal profile for each respective substrate of a plurality of substrates; for each respective substrate, determine respective control parameter values to apply to the respective substrate from the output nodes of the artificial neural network by applying the target removal profile to the input nodes of the artificial neural network; and cause a semiconductor processing system to subject each respective substrate to the process using the respective control parameter values.
In yet another aspect, a polishing system includes: a support to hold a polishing pad; a carrier head to hold a substrate in contact with the polishing pad, the carrier head having a plurality of chambers; a motor to generate relative motion between the support and the carrier head; and a controller configured to: train an artificial neural network by backpropagation using a plurality of groups of process parameter and a plurality of removal profiles as training data, the artificial neural network having a plurality of input nodes to receive respective removal values from a removal profile, a plurality of output nodes to output control parameter values, and a plurality of hidden nodes connecting the input nodes to the output nodes, wherein the control parameter values include pressures for the chambers in the carrier head; obtain a target removal profile for each respective substrate of a plurality of substrates; for each respective substrate, determine respective control parameter values to apply to the respective substrate from the output nodes of the artificial neural network by applying the target removal profile to the input nodes of the artificial neural network, wherein the respective control parameter values include respective pressures for the chambers in the carrier head; and for each respective substrate, cause the carrier head to apply the respective pressures to the chambers in the carrier head during polishing.
Implementations may include one or more of the following features.
The process may include chemical mechanical polishing. The control parameter values may include pressures of chambers in a carrier head to hold a substrate against a polishing pad.
At least some of the input nodes of the artificial neural network may be configured to receive state parameter values. The state parameter values may include one or more of a retaining ring life, or a polishing pad life. The control parameter values may include one or more of a platen rotation rate, carrier head rotation rate, or polishing time.
The monitoring system may include an in-line metrology system. The in-line metrology system may include an optical monitoring system.
The determining of the target removal profile may include storing a desired thickness profile, receiving a measured thickness profile of the respective second substrate, and determining a difference between the measured thickness profile and the desired thickness profile. The receiving of the measured thickness profile may include measuring a thickness profile of the respective second substrate with the monitoring system.
Certain implementations may have one or more of the following advantages. Control parameter convergence may be achieved more rapidly. Within-wafer thickness non- uniformity and wafer-to-wafer thickness non-uniformity (WIWNU and WTWNU) may be reduced. Product wafers may be used for model refinement, leading to the actual profile of processed substrates being closer to a desired profile. Complex behavior of the polishing process in response to polishing parameters may be accounted for without explicit knowledge of the functional relationships.
The details of one or more embodiments are set forth in the accompanying drawings and the description below. Other features, objects, and advantages will be apparent from the description and drawings, and from the claims.
DESCRIPTION OF DRAWINGS
FIG. 1 illustrates a schematic cross-sectional view of an example of a polishing apparatus.
FIG. 2 illustrates a schematic top view of a substrate having multiple zones.
FIG. 3 illustrates a neural network used as part of the integrated advanced process control module of the polishing apparatus.
FIG. 4 illustrates an example process for using a neural network model in a CMP process.
FIG. 5 illustrates an integrated advanced process control module that includes multiple neural networks.
Like reference symbols in the various drawings indicate like elements.
DETAILED DESCRIPTION
One challenge in CMP is developing a process model that is able to predict a material removal profile as a function of multiple input parameters. Input parameters may include variations in the initial thickness of a substrate layer, a target material removal profile, the polishing pad condition, the retaining ring condition, the relative speed between the polishing pad and a substrate, and the applied pressure on a substrate. Furthermore, the CMP process is a dynamical system in which surface topology is constantly changing as a function of time during the process, which in turn affects the process behavior. The complex relationship of input parameters with the material removal profile and the dynamic nature of a CMP process contribute to the challenge of developing a process model of the polishing process.
FIG. 1 illustrates an example of a polishing apparatus 20. The polishing apparatus 20 can include a rotatable disk-shaped platen 22 on which a polishing pad 30 is situated. The platen is operable to rotate about an axis 23. For example, a motor 24 can turn a drive shaft 26 to rotate the platen 22. The polishing pad 30 can be detachably secured to the platen 22, for example, by a layer of adhesive. The polishing pad 30 can be a two-layer polishing pad with an outer polishing layer 32 and a softer backing layer 34.
The polishing apparatus 20 can include a polishing liquid supply port 40 to dispense a polishing liquid 42, such as an abrasive slurry, onto the polishing pad 30. The polishing apparatus 20 can also include a polishing pad conditioning disc to abrade the polishing pad 30 to maintain the polishing pad 30 in a consistent abrasive state.
A carrier head 50 is operable to hold a substrate 10 against the polishing pad 30. Each carrier head 50 also includes a plurality of independently controllable pressurizable chambers, e.g., three chambers 52a-52c, which can apply independently controllable pressurizes to associated zones 12a-12c on the substrate 10 (see FIG. 2). Referring to FIG. 2, the center zone 12a can be substantially circular, and the remaining zones 12b- 12c can be concentric annular zones around the center zone 12a.
Returning to FIG. 1, the chambers 52a-52c can be defined by a flexible membrane 54 having a bottom surface to which the substrate 10 is mounted. The carrier head 50 can also include a retaining ring 56 to retain the substrate 10 below the flexible membrane 54.
Although only three chambers are illustrated in FIGS. 1 and 2 for ease of illustration, there could be two chambers, or four or more chambers, e.g., five chambers. In addition, other mechanisms to adjust the pressure applied to the substrate, e.g., piezoelectric actuators, could be used in the carrier head 50.
Each carrier head 50 is suspended from a support structure 60, e.g., a carousel or track, and is connected by a drive shaft 62 to a carrier head rotation motor 64 so that the carrier head can rotate about an axis 51. Optionally each carrier head 50 can oscillate laterally, e.g., on sliders on the carousel, by motion along the track; or by rotational oscillation of the carousel itself. In operation, the platen 22 is rotated about its central axis 23, and the carrier head 50 is rotated about its central axis 51 and translated laterally across the top surface of the polishing pad 30.
The polishing apparatus 20 can include an in-line monitoring system for measuring a thickness of a polished layer upon completion of the polishing process. For example, the inline monitoring system can generate a map of layer thicknesses over the zones 12a-12c. The measurements from the in-line monitoring system can be communicated to a controller 90. Examples of the in-line monitoring system include an optical monitoring system, e.g., a spectrographic monitoring system. A spectrographic monitoring system can measure a layer thickness value and an associated goodness of fit (GOF) value for each points of the layer thickness map. For example, a broadband light source can be used to illuminate a location on the layer, and the reflection containing an optical interference spectrum created by the layer, the substrate, and any other layers in between the two can be measured. The measured optical interference spectrum can be analyzed, for example, by fitting it with an equation that describes an optical interference spectrum generated by an expected stacking of films. The fitting produces a determination of the thickness of the layer and a GOF value that is indicative of how closely the measured spectrum agrees with the expected film stack. As such, the GOF value can be used as an indicator of reliability of the determined thickness value.
The controller 90 contains a tool control module 92 and an integrated advanced process control module (i-APC) 94. The tool control module 92 and the i-APC module 94 in combination may provide advanced process control functionalities, such as wafer-to-wafer uniformity control. The controller 90 can be a computing device that includes a
microprocessor, memory and input/output circuitry, e.g., a programmable computer.
Although illustrated with a single block, the controller 90 can be a networked system with functions distributed across multiple computers, and modules 92 and 94 can be located in the same computer or different computers.
The described polishing apparatus has many associated process parameters that control the operation of the polishing apparatus or describe the state of the apparatus or the polishing environment. Process parameters that control the operation of the polishing apparatus (and that can be set, at least initially, by the tool control module 92) ('control parameters') include the following: rotation rate of platen 22; rotation rate of carrier head 50; pressure of the chambers 52a-52c; and polishing time.
Process parameters that reflect the state of the apparatus or polishing environment ('state parameters') include the following: polishing retaining ring life; polishing pad life; polishing pad conditioning disc life; type of polishing pad; and type of polishing liquid ('slurry'). The retaining ring, pad, and conditioning disc are examples of consumable components ('consumables') within a polishing apparatus. The condition or 'life' of these consumables can be described, for example, as a count of wafers processed; actual wafer polishing time; or total time elapsed since installation.
Atypical wafer to be polished by a CMP process has a layer of material with surface topologies to be planarized. A goal of the CMP process is to achieve a desired material removal profile, which is a one or two dimensional map of the change in a polished layer's thickness after the polishing process. The surface topologies vary from wafer-to-wafer due to different die designs having different underlying transistor and interconnect patterns, e.g. due to different pattern densities. These factors interact with the CMP process in a complex manner, which results in different polishing behavior between wafers with different die designs. Furthermore, even wafers with the same die design may have different polishing behaviors due to upstream process variations in deposition or etching. Therefore, different CMP control parameters are typically needed, at least for wafers with different die designs, and possibly for individual wafers with the same die design, to achieve a desired material removal profile. Material removal profiles sometimes have a radial dependence partly due to the axial symmetry and rotation of the polishing head. Accordingly, one control parameter that is often tuned to achieve a desired radial profile is the pressures of the radial chambers 52a-52c.
The polishing apparatus 20 can implement a wafer-to-wafer control. The wafer-to- wafer control can provide improved likelihood of achieving target material removal profiles over a wide range of designs and wafer non-uniformities. A wafer-to-wafer feedback control method uses information about previously processed substrates to improve processing of a subsequent substrate. The wafer-to-wafer feedback control method can be implemented by the i-APC module 94.
In an example implementation of the controller 90, the i-APC module 94 generates an initial set of control parameter values based on an initial process model, and provides the initial set of control parameters values to the tool control module 92. The tool control module 92 can then control the polishing system using the received control parameter values.
After processing of one or more substrates over a user defined period of time, data about the substrates can be used by the i-APC module 94 to improve the initial process model by generating a new process model or updating the initial process model. For example, the user defined period of time for generation of a new process model or updating of the initial process model can be measured in number of substrates polished, such as 5, 10, 25, or 100 wafers, or a "lot" of wafers (wafers are typically transported and processed in "lots"; a typical lot size is 25 wafers). As another example, the user defined period of time may be defined as a portion of a lifetime of a consumable, such as a polishing pad lifetime. Such generation or updating of the model can be performed off-line, or independent of the operation of the polishing apparatus.
The improved process model can then be used to generate an improved set of control parameter values for a subsequent substrate. Improvement of the process model by the i- APC module 94 can lead to improved polishing uniformity control by minimizing error between the target material removal profile and the actual profile realized on the substrate.
The i-APC module 94 performs tasks including collecting and processing of data from processed wafers to improve processing of future wafers. Collected data can include upstream and downstream metrology data of the product wafers from various wafer metrology tools ("monitoring systems"). Upstream metrology data may include thickness map and associated GOF values of a deposited layer. Downstream metrology data may include a thickness map and associated GOF values of the polished layer, or surface roughness values. The GOF values can be used by the i-APC module 94 to determine whether the thickness value are reliable enough to be used in the process model development. The i-APC module 94 pairs these data with the control and state parameters used during the processing of a particular wafer, and stores it in a data log. The data log may be organized in a variety of ways, including grouping by wafer ID, design ID, lot ID, tool ID, etc. This data log is typically used to monitor trends and drifts in the behavior of a CMP process and to take corrective actions. Due to their size, data logs may be stored in one or more servers that are a part of the controller 90.
In some implementations, the i-APC module 94 stores a desired thickness profile for each wafer to be processed. Using the desired thickness profile, the i-APC module 94 can generate the target material removal profile by subtracting a desired thickness profile from the thickness map of the deposited layer.
A process model can be generated in various ways. For example, a process model for describing the effects of chamber pressures can be generated by processing multiple blanket wafers, which are un-patterned wafers with a uniform layer of a film to be polished, as a proxy for product wafers that have patterns. By polishing the blanket wafers for a fixed amount of time and measuring the resulting material removal profile, a relationship between a set of chamber pressures and a material removal rate can be determined using the Preston equation.
The Preston equation states:
Material removal rate (MRR) = KP*V*P Equation 1 where V is the velocity of the polishing pad surface with respect to the substrate surface being polished, P is the pressure applied to a radial zone of a wafer to be polished against the polishing pad, and Kp is a proportionality constant known as the Preston coefficient. The velocity, pressure, and polish time are known and controlled variables, so Kp of each zone can be determined with algebraic manipulations. The resulting set of Preston equations containing the Preston coefficients from polishing of blanket wafers is an example of a baseline process model. The i-APC module 94 can then use this baseline process model to generate a set of initial control parameters for processing of subsequent product wafers. While a process model for modeling effects of chamber pressures are discussed, process models for other control parameters can be generated in this manner.
Due to the differences in surface topologies between a blanket wafer and a product wafer, however, the baseline process model might not achieve the target material removal profile on the polished product wafer. In such case, the i-APC 94 attempts to find, through multiple iterations over polishing of multiple product wafers, a set of correction factors ("offsets") for improving the achieved material removal profile on the product wafers. The i- APC module 94 eventually determines, or converges on, offsets that can be used with the baseline process model for polishing the product wafer. However, such the determination of the offsets can take polishing of multiple product wafers, during which the polished wafers are less likely to meet the target removal profile thus more likely to be rejected, adversely impacting yield of the polishing process.
Consumption of product wafers for offset convergence should be minimized as they are valuable, expensive, and limited in quantity. One way of reducing the number of rejected product wafers until offset convergence is reached is to provide an improved process model that more completely and accurately captures the effects of various process parameters. In some implementations, the i-APC module 94 is configured to provide the improved process model. One way of improving the quality of the process model is to implement a process model that takes in as inputs (1) target material removal profile and (2) state parameters that reflect the conditions of the polishing apparatus, and outputs a set of estimated control parameters (e.g. chamber pressures) that would achieve the target material removal profile.
The material removal rate is influenced by state parameters 114 of the polishing apparatus such as the life, or age, of various consumables due to their wear. For example, the polishing pad may experience a reduction in elasticity or a reduction in surface roughness due to aging. The retaining ring may get polished away during the CMP process, becoming thinner. The polishing pad conditioning disc may become dull and have reduced capacity for pad conditioning. These changes to characteristics of the consumables can affect the polishing behavior. Therefore, wafers with an identical design may require different sets of control parameters for different state parameters to achieve the target material removal profile. Therefore, incorporating state parameters into the process model can lead to better control parameter estimation.
A neural network can be used to model behavior (e.g., removal profile) of a semiconductor processing system (e.g., a CMP system) in response to the process parameters. The neural network is based on a large collection of simple neural units
(artificial neurons). Each individual neural unit computes a respective output, or activation, using a summation function. There may be a threshold function or limiting function on each connection and on the unit itself, such that the signal must surpass the limit before propagating to other neurons. Such threshold function may include a hyperbolic tangent or a sigmoid function.
A neural network is trained, rather than explicitly programmed. In other words, a neural network does not require a priori knowledge of the functional dependence of the process being modeled. Instead, the neural network can be trained by providing it with a sufficiently large and representative set of training data; this technique is called supervised machine learning. Because a neural network can excel at modeling complex relationships between inputs and the outputs that are difficult to express in functional relationships, it can provide significant advantages for control of a semiconductor processing system, such as a CMP system. FIG. 3 illustrates the functional blocks implemented by the i-APC module 94. The i- APC module 94 includes neural network input variable set 1 10, which includes target removal profile 1 12 and state parameters 1 14, a neutral network 120, and a process initialization module 130 that initializes the tool control module 92 of a semiconductor processing system, e.g., a CMP tool, with the output of the neural network. These functional blocks can be distributed across multiple computers.
The neutral network 120 includes a plurality of input nodes 122 for each input variable from neural network input variable set 1 10, a plurality of hidden nodes 124 (also called "intermediate nodes" below), and a plurality output nodes 126 that will generate the process parameters to be used for initialization. In a neural network having a single layer of hidden nodes, each hidden node 124 can be coupled to each input node 122, and each output node 126 can be coupled to each hidden node 124.
In general, a hidden node 124 outputs a value that a non-linear function of a weighted sum of the values from the input nodes 122 to which the hidden node is connected.
For example, the output of a hidden node 124, designated node k, can be expressed as:
tanh (0.5*α¾1(Ι1)+<¾2(Ι2)+ ... + <¾Μ(ΙΜ)+ ¾) Equation 2 where tanh is the hyperbolic tangent, aia is a weight for the connection between the kth intermediate node and the xth input node (out of M input nodes), and IM is the value at the M*h input node. However, other non-linear functions can be used instead of tanh, such as a rectified linear unit (ReLU) function and its variants, or a sigmoid function.
FIG. 4 illustrates an example process 400 for training and using a neural network model in a CMP process. The neutral network 120 need to be configured before being used to generate baseline control parameters. The neural network 120 may be configured through training by using various methods of supervised learning. One example of supervised learning method is backpropagation, the use of forward stimulation to reset weights on the 'front' neural units, or input nodes 122.
Before the neural network can be trained, the i-APC control module 94 collects sufficient amount of training data (410). The training data comprises a set of input and output pairs that have been measured, and therefore known to be true. For example, data log maintained by the i-APC control module 94 can be used to retrieve for each substrate, the state parameters 114 during the process; the target removal profile 112; the control parameters used during the process; and the measured removal profile. The retrieved information is then analyzed to determine its suitability as a training example. In the current example, the measured removal profile is compared with the target profile. If the difference between the two is sufficiently small to be within the specification of the process, the data is determined to be a suitable training example. However, it should be noted that the CMP polish behavior may vary substantially between substrates with different designs. Therefore, it may be beneficial to further categorize the training data into different die designs to be used for creating design-specific neural networks. Once a sufficient amount of such training examples have been collected, the training begins.
The i-APC control module 94 trains the neural network 120 (420), e.g., by backpropagation. Backpropagation is a two phase cycle of propagation and weight update. During the propagation cycle, the input vector (e.g. target removal profile and state parameters from a training example) presented to the network is propagated forward through the network, layer by layer, until it reaches the output layer. The output of the network is then compared to the recorded output (e.g. control parameter from a training example) using a cost function, and an error value is calculated for each of the neurons in the output layer. A cost function or loss function is a function that maps values of one or more variables onto a real number intuitively representing some 'cost' associated with the event. The error values are then propagated backwards, starting from the output, until each neuron has an associated error value which roughly represents its contribution to the original output. Backpropagation uses these error values to calculate the gradient of the loss function with respect to the weights in the network. Then in the weight update phase, back propagation method uses these error values to update the weights, in an attempt to minimize the loss function.
The i-APC control module 94 determines whether the trained neural network is accurate (430). After the training phase 420 is completed, the neural network's accuracy should be verified. Verification of the accuracy, for example, may include operating the neural network in its inference mode, feeding it with previously used training input, and determining whether the output is in agreement, or is sufficiently close to the correct output in the training data. If the neural network output is not sufficiently accurate, then the process returns to step 410 to collect additional training data. If the output is accurate, then the neural network is ready to be used.
The i-APC control module 94 uses the trained neural network (440) to generate process parameters. In inference mode, the neural network outputs process parameters to be used for initialization in response to target material removal profile 112 and state parameters 114.
In some implementations, individual neural network is generated and trained for each wafer design. In such implementations, the i-APC control module 94 maintains a library of neural networks, and the controller selects a neural network trained for the current wafer design to initialize the process.
The architecture of the neural network 120 can vary in depth and width. For example, although the neural network 120 is shown with a single column of intermediate nodes 124, it can include multiple columns. The number of intermediate nodes 124 can be equal to or greater than the number of input nodes 122.
In some implementations, referring to FIG. 5, the i-APC control module 94 can be configured to have multiple neural networks 120. The neural networks 120 can operate in parallel to generate the chamber pressure values for each zones. The number of neural networks can match the number of zones. For example, there can be a first neural network 120a for the first zone 12a, a second neural network 120b for the second zone 12b and a third neural network for the third zone 12c. The output of each neural network 120 can be fed to the process initialization module 130.
The training of the neural network can be computationally intensive. Accordingly, in some implementations, training of the neural network may be done off-line, e.g., while the CMP tool is undergoing maintenance. In other implementations, the training can be done on a separate server that is part of the controller 90. In some other implementations, the training can be done in a server that is not part of the controller 90, which provides the pre-trained neural network to the controller.
The behavior of a CMP process may slowly change over time for reasons that are not factored in (e.g. wear of parts that are not considered consumables) by the input variables of the neural network. Therefore, in some implementations, a periodic retraining of the neural network may be performed to overcome such changes. Embodiments and all of the functional operations described in this specification can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the structural means disclosed in this specification and structural equivalents thereof, or in combinations of them. Embodiments can be implemented as one or more computer program products, i.e., one or more computer programs tangibly embodied in a machine readable storage media, for execution by, or to control the operation of, data processing apparatus, e.g., a programmable processor, a computer, or multiple processors or computers. A computer program (also known as a program, software, software application, or code) can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program does not necessarily correspond to a file. A program can be stored in a portion of a file that holds other programs or data, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub programs, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers at one site or distributed across multiple sites and interconnected by a communication network.
The processes and logic flows described in this specification can be performed by one or more programmable processors executing one or more computer programs to perform functions by operating on input data and generating output. The processes and logic flows can also be performed by, and apparatus can also be implemented as, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit).
The above described polishing apparatus and methods can be applied in a variety of polishing systems. Either the polishing pad, or the carrier heads, or both can move to provide relative motion between the polishing surface and the substrate. For example, the platen may orbit rather than rotate. The polishing pad can be a circular (or some other shape) pad secured to the platen. The polishing system can be a linear polishing system, e.g., where the polishing pad is a continuous or a reel-to-reel belt that moves linearly. The polishing layer can be a standard (for example, polyurethane with or without fillers) polishing material, a soft material, or a fixed-abrasive material. Terms of relative positioning are used relative orientation or positioning of the components; it should be understood that the polishing surface and substrate can be held in a vertical orientation or some other orientation with respect to gravity.
Particular embodiments have been described. Other embodiments are within the scope of the following claims.

Claims

WHAT IS CLAIMED IS:
1. A method of processing substrates, comprising:
subjecting each respective first substrate of a first plurality of substrates to a process that modifies a thickness of an outer layer of the respective first substrate;
for each respective first substrate, recording a group of process parameter values used for the process, thus generating a plurality of groups of process parameter values;
for each respective first substrate, measuring a removal profile of the outer layer during or after the process with a monitoring system, thus generating a plurality of removal profiles;
training an artificial neural network by backpropagation using the plurality of groups of process parameter and plurality of removal profiles as training data, the artificial neural network having a plurality of input nodes to receive respective removal values from the removal profile, a plurality of output nodes to output control parameter values, and a plurality of hidden nodes connecting the input nodes to the output nodes;
for each respective second substrate of a second plurality of substrates, determining a target removal profile;
for each respective second substrate, determining respective control parameter values to apply to the respective second substrate from the output nodes of the artificial neural network by applying the target removal profile to the input nodes of the artificial neural network; and
subjecting each respective second substrate to the process using the respective control parameter values.
2. The method of claim 1, wherein the process comprises chemical mechanical polishing.
3. The method of claim 1, wherein at least some of the input nodes of the artificial neural network are configured to receive state parameter values.
4. The method of claim 3, wherein the process comprises chemical mechanical polishing.
5. The method of claim 4, wherein the control parameter values comprise pressures of chambers in a carrier head to hold a substrate against a polishing pad.
6. The method of claim 5, wherein the control parameter values comprise one or more of a platen rotation rate, carrier head rotation rate, or polishing time.
7. The method of claim 4, wherein the state parameter values comprise one or more of a retaining ring life, or a polishing pad life.
8. The method of claim 1, wherein determining the target removal profile comprises storing a desired thickness profile, receiving a measured thickness profile of the respective second substrate, and determining a difference between the measured thickness profile and the desired thickness profile.
9. The method of claim 1, wherein receiving the measured thickness profile comprises measuring a thickness profile of the respective second substrate with the monitoring system.
10. A computer program product for controlling processing of a substrate, the computer program product tangibly embodied in a non-transitory computer readable medium and comprising instructions for causing a processor to:
train an artificial neural network by backpropagation using a plurality of groups of process parameter and a plurality of removal profiles as training data, the artificial neural network having a plurality of input nodes to receive respective removal values from a removal profile, a plurality of output nodes to output control parameter values, and a plurality of hidden nodes connecting the input nodes to the output nodes;
obtain a target removal profile for each respective substrate of a plurality of substrates;
for each respective substrate, determine respective control parameter values to apply to the respective substrate from the output nodes of the artificial neural network by applying the target removal profile to the input nodes of the artificial neural network; and
cause a semiconductor processing system to subject each respective substrate to the process using the respective control parameter values.
11. The computer program product of claim 10, wherein at least some of the input nodes of the artificial neural network are configured to receive state parameter values.
12. The computer program product of claim 11, wherein the control parameter values represent pressures of chambers in a carrier head to hold a substrate against a polishing pad, and the state parameter values represent one or more of a retaining ring life, or a polishing pad life.
13. A polishing system, comprising:
a support to hold a polishing pad;
a carrier head to hold a substrate in contact with the polishing pad, the carrier head having a plurality of chambers;
a motor to generate relative motion between the support and the carrier head; and a controller configured to
train an artificial neural network by backpropagation using a plurality of groups of process parameter and a plurality of removal profiles as training data, the artificial neural network having a plurality of input nodes to receive respective removal values from a removal profile, a plurality of output nodes to output control parameter values, and a plurality of hidden nodes connecting the input nodes to the output nodes, wherein the control parameter values comprise pressures for the chambers in the carrier head;
obtain a target removal profile for each respective substrate of a plurality of substrates;
for each respective substrate, determine respective control parameter values to apply to the respective substrate from the output nodes of the artificial neural network by applying the target removal profile to the input nodes of the artificial neural network, wherein the respective control parameter values comprise respective pressures for the chambers in the carrier head; and
for each respective substrate, cause the carrier head to apply the respective pressures to the chambers in the carrier head during polishing.
14. The polishing system of claim 13, wherein at least some of the input nodes of the artificial neural network are configured to receive state parameter values.
15. The polishing system of claim 14, wherein the control parameter values represent pressures of the plurality of chambers in the carrier head, and the state parameter values represent one or more of a retaining ring life, or a polishing pad life.
PCT/US2018/052001 2017-09-25 2018-09-20 Semiconductor fabrication using machine learning approach to generating process control parameters Ceased WO2019060577A1 (en)

Priority Applications (3)

Application Number Priority Date Filing Date Title
JP2020516913A JP7036908B2 (en) 2017-09-25 2018-09-20 Semiconductor manufacturing using a machine learning approach to generate process control parameters
KR1020207011805A KR102450925B1 (en) 2017-09-25 2018-09-20 Semiconductor Manufacturing Using Machine Learning Approach to Generating Process Control Parameters
CN201880062109.4A CN111133560B (en) 2017-09-25 2018-09-20 Semiconductor manufacturing using machine learning methods to generate process control parameters

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US201762562986P 2017-09-25 2017-09-25
US62/562,986 2017-09-25

Publications (2)

Publication Number Publication Date
WO2019060577A1 true WO2019060577A1 (en) 2019-03-28
WO2019060577A4 WO2019060577A4 (en) 2019-05-09

Family

ID=65809083

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2018/052001 Ceased WO2019060577A1 (en) 2017-09-25 2018-09-20 Semiconductor fabrication using machine learning approach to generating process control parameters

Country Status (6)

Country Link
US (1) US12524675B2 (en)
JP (1) JP7036908B2 (en)
KR (1) KR102450925B1 (en)
CN (1) CN111133560B (en)
TW (1) TWI783037B (en)
WO (1) WO2019060577A1 (en)

Families Citing this family (39)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI816620B (en) 2017-04-21 2023-09-21 美商應用材料股份有限公司 Polishing apparatus using neural network for monitoring
TWI783037B (en) * 2017-09-25 2022-11-11 美商應用材料股份有限公司 Semiconductor fabrication using machine learning approach to generating process control parameters
US20190130303A1 (en) * 2017-10-26 2019-05-02 International Business Machines Corporation Smart default threshold values in continuous learning
US11380594B2 (en) * 2017-11-15 2022-07-05 Kla-Tencor Corporation Automatic optimization of measurement accuracy through advanced machine learning techniques
US11029359B2 (en) * 2018-03-09 2021-06-08 Pdf Solutions, Inc. Failure detection and classsification using sensor data and/or measurement data
TWI845444B (en) 2018-04-03 2024-06-11 美商應用材料股份有限公司 Polishing apparatus, polishing system, method, and computer storage medium using machine learning and compensation for pad thickness
CN111886686B (en) 2018-09-26 2024-08-02 应用材料公司 Compensation of substrate doping in edge reconstruction for in-situ electromagnetic induction monitoring
US11989492B2 (en) 2018-12-26 2024-05-21 Applied Materials, Inc. Preston matrix generator
JP7086835B2 (en) * 2018-12-28 2022-06-20 株式会社荏原製作所 Polishing recipe determination device
WO2020137802A1 (en) * 2018-12-28 2020-07-02 株式会社荏原製作所 Pad temperature adjusting device, pad temperature adjusting method, polishing device, and polishing system
SG10202006423VA (en) * 2019-07-12 2021-02-25 Ebara Corp Substrate processing apparatus and storage medium
WO2021061477A1 (en) 2019-09-27 2021-04-01 Corning Incorporated Devices, systems, and methods of generating and providing a target topographic map for finishing a photomask blank subject to functional requirements on flatness
US12197133B2 (en) * 2019-10-08 2025-01-14 International Business Machines Corporation Tool control using multistage LSTM for predicting on-wafer measurements
JP7412150B2 (en) * 2019-11-29 2024-01-12 東京エレクトロン株式会社 Prediction device, prediction method and prediction program
KR20210071130A (en) * 2019-12-05 2021-06-16 삼성전자주식회사 Computing device, operating method of computing device, and storage medium
JP7517832B2 (en) * 2020-01-17 2024-07-17 株式会社荏原製作所 Polishing head system and polishing apparatus
US12131103B2 (en) * 2020-03-30 2024-10-29 Kla Corporation Semiconductor fabrication process parameter determination using a generative adversarial network
WO2021231427A1 (en) * 2020-05-14 2021-11-18 Applied Materials, Inc. Technique for training neural network for use in in-situ monitoring during polishing and polishing system
KR102732531B1 (en) 2020-06-24 2024-11-21 어플라이드 머티어리얼스, 인코포레이티드 Determination of substrate layer thickness using polishing pad wear compensation
US11693435B2 (en) * 2020-06-25 2023-07-04 Applied Materials, Inc. Ethercat liquid flow controller communication for substrate processing systems
CN114124654B (en) * 2020-08-10 2023-10-27 中国移动通信集团浙江有限公司 Alarm merging method, device, computing equipment and computer storage medium
US11371148B2 (en) * 2020-08-24 2022-06-28 Applied Materials, Inc. Fabricating a recursive flow gas distribution stack using multiple layers
KR102825817B1 (en) 2020-08-28 2025-06-30 삼성전자주식회사 Thickness estimation method and process control method
US12189375B2 (en) * 2020-09-21 2025-01-07 Applied Materials, Inc. Dynamic scheduling based on task dependencies
US11313670B2 (en) * 2020-09-30 2022-04-26 National Tsing Hua University Inspection method for multilayer semiconductor device
US12237158B2 (en) 2020-11-24 2025-02-25 Applied Materials, Inc. Etch feedback for control of upstream process
US12351917B2 (en) * 2020-12-01 2025-07-08 Applied Materials, Inc. Multi-zone heater tuning in substrate heater
US11709477B2 (en) * 2021-01-06 2023-07-25 Applied Materials, Inc. Autonomous substrate processing system
CN112992735B (en) * 2021-02-07 2021-09-10 锋睿领创(珠海)科技有限公司 Semiconductor bonding pose compensation repairing method and device and storage medium
US12504364B2 (en) * 2021-03-03 2025-12-23 Applied Materials, Inc. In-situ monitoring to label training spectra for machine learning system for spectrographic monitoring
US11931853B2 (en) * 2021-03-05 2024-03-19 Applied Materials, Inc. Control of processing parameters for substrate polishing with angularly distributed zones using cost function
US12593646B2 (en) 2021-04-28 2026-03-31 Tel Manufacturing And Engineering Of America, Inc. Automated fault detection in microfabrication
US11791184B2 (en) 2021-07-26 2023-10-17 Samsung Electronics Co., Ltd. Semiconductor fabrication process and method of optimizing the same
JP7689937B2 (en) * 2022-05-27 2025-06-09 株式会社荏原製作所 Method for estimating polishing pad life and polishing apparatus
CN117079737B (en) * 2023-10-17 2024-08-02 深圳市永霖科技有限公司 Polishing solution prediction method and device based on component analysis
CN117067096B (en) * 2023-10-18 2023-12-15 苏州博宏源机械制造有限公司 Automatic control system and method for double-sided grinding and polishing equipment based on parameter optimization
CN117697613B (en) * 2023-12-22 2024-05-07 上海强华实业股份有限公司 Identification and grinding control method and system applied to quartz product grinding equipment
US20260051038A1 (en) * 2024-08-19 2026-02-19 Globalwafers Co., Ltd. Systems and methods for enhanced wafer manufacturing
CN119761276B (en) * 2025-03-03 2025-07-01 全芯智造技术有限公司 Simulation method, electronic device and storage medium for semiconductor device

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7001243B1 (en) * 2003-06-27 2006-02-21 Lam Research Corporation Neural network control of chemical mechanical planarization
US7037172B1 (en) * 1999-04-01 2006-05-02 Beaver Creek Concepts Inc Advanced wafer planarizing
US20070224915A1 (en) * 2005-08-22 2007-09-27 David Jeffrey D Substrate thickness measuring during polishing
US20100129939A1 (en) * 2008-11-26 2010-05-27 Applied Materials, Inc. Using optical metrology for within wafer feed forward process control
KR20120123493A (en) * 2010-01-29 2012-11-08 도쿄엘렉트론가부시키가이샤 Method and system for self-learning and self-improving a semiconductor manufacturing tool

Family Cites Families (40)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0465632A (en) 1990-07-04 1992-03-02 Toshio Fukuda measuring device
US6776692B1 (en) 1999-07-09 2004-08-17 Applied Materials Inc. Closed-loop control of wafer polishing in a chemical mechanical polishing system
US6439964B1 (en) 1999-10-12 2002-08-27 Applied Materials, Inc. Method of controlling a polishing machine
US6544103B1 (en) 2000-11-28 2003-04-08 Speedfam-Ipec Corporation Method to determine optimum geometry of a multizone carrier
US6540591B1 (en) 2001-04-18 2003-04-01 Alexander J. Pasadyn Method and apparatus for post-polish thickness and uniformity control
US7698012B2 (en) 2001-06-19 2010-04-13 Applied Materials, Inc. Dynamic metrology schemes and sampling schemes for advanced process control in semiconductor processing
US7160739B2 (en) 2001-06-19 2007-01-09 Applied Materials, Inc. Feedback control of a chemical mechanical polishing device providing manipulation of removal rate profiles
US6594024B1 (en) * 2001-06-21 2003-07-15 Advanced Micro Devices, Inc. Monitor CMP process using scatterometry
US6857947B2 (en) 2002-01-17 2005-02-22 Asm Nutool, Inc Advanced chemical mechanical polishing system with smart endpoint detection
US7112270B2 (en) * 2002-09-16 2006-09-26 Applied Materials, Inc. Algorithm for real-time process control of electro-polishing
TWI238754B (en) 2002-11-07 2005-09-01 Ebara Tech Inc Vertically adjustable chemical mechanical polishing head having a pivot mechanism and method for use thereof
US20050070205A1 (en) 2003-09-30 2005-03-31 Speedfam-Ipec Corporation Integrated pressure control system for workpiece carrier
US6932671B1 (en) 2004-05-05 2005-08-23 Novellus Systems, Inc. Method for controlling a chemical mechanical polishing (CMP) operation
EP1758711B1 (en) 2004-06-21 2013-08-07 Ebara Corporation Polishing apparatus and polishing method
US7150673B2 (en) 2004-07-09 2006-12-19 Ebara Corporation Method for estimating polishing profile or polishing amount, polishing method and polishing apparatus
JP4689367B2 (en) 2004-07-09 2011-05-25 株式会社荏原製作所 Method for predicting polishing profile or polishing amount, polishing method and polishing apparatus
JPWO2006106790A1 (en) 2005-04-01 2008-09-11 株式会社ニコン Polishing apparatus, semiconductor device manufacturing method using the polishing apparatus, and semiconductor device manufactured by the semiconductor device manufacturing method
US7184853B2 (en) * 2005-05-18 2007-02-27 Infineon Technologies Richmond, Lp Lithography method and system with correction of overlay offset errors caused by wafer processing
US20070143234A1 (en) * 2005-12-16 2007-06-21 Taiwan Semiconductor Manufacturing Company, Ltd. Method and system for intelligent model-based optical proximity correction (OPC)
US7623978B2 (en) * 2006-03-30 2009-11-24 Tokyo Electron Limited Damage assessment of a wafer using optical metrology
US7115017B1 (en) 2006-03-31 2006-10-03 Novellus Systems, Inc. Methods for controlling the pressures of adjustable pressure zones of a work piece carrier during chemical mechanical planarization
US20070249071A1 (en) * 2006-04-21 2007-10-25 Lei Lian Neural Network Methods and Apparatuses for Monitoring Substrate Processing
JP5283506B2 (en) 2006-09-12 2013-09-04 株式会社荏原製作所 Polishing apparatus and polishing method
US7699688B2 (en) 2006-11-22 2010-04-20 Applied Materials, Inc. Carrier ring for carrier head
US7667858B2 (en) * 2007-01-12 2010-02-23 Tokyo Electron Limited Automated process control using optical metrology and a correlation between profile models and key profile shape variables
US8577820B2 (en) * 2011-03-04 2013-11-05 Tokyo Electron Limited Accurate and fast neural network training for library-based critical dimension (CD) metrology
US8774958B2 (en) * 2011-04-29 2014-07-08 Applied Materials, Inc. Selection of polishing parameters to generate removal profile
CN202264145U (en) 2011-09-14 2012-06-06 桂林电子科技大学 Artificial neural network-based automatic grinder grinding wheel dresser
US8838422B2 (en) 2011-12-11 2014-09-16 Tokyo Electron Limited Process control using ray tracing-based libraries and machine learning systems
US9754846B2 (en) 2014-06-23 2017-09-05 Applied Materials, Inc. Inductive monitoring of conductive trench depth
US11209804B2 (en) * 2014-11-11 2021-12-28 Applied Materials, Inc. Intelligent processing tools
US10430719B2 (en) * 2014-11-25 2019-10-01 Stream Mosaic, Inc. Process control techniques for semiconductor manufacturing processes
KR102521159B1 (en) * 2014-11-25 2023-04-13 피디에프 솔루션즈, 인코포레이티드 Improved process control techniques for semiconductor manufacturing processes
KR102437268B1 (en) 2015-02-25 2022-08-29 주식회사 케이씨텍 Chemical mechanical polishing system
US11580375B2 (en) 2015-12-31 2023-02-14 Kla-Tencor Corp. Accelerated training of a machine learning based model for semiconductor applications
TWI807987B (en) * 2016-11-30 2023-07-01 美商應用材料股份有限公司 Spectrographic monitoring using a neural network
JP7023455B2 (en) 2017-01-23 2022-02-22 不二越機械工業株式会社 Work polishing method and work polishing equipment
TWI816620B (en) * 2017-04-21 2023-09-21 美商應用材料股份有限公司 Polishing apparatus using neural network for monitoring
TWI783037B (en) * 2017-09-25 2022-11-11 美商應用材料股份有限公司 Semiconductor fabrication using machine learning approach to generating process control parameters
US20220347813A1 (en) * 2021-04-30 2022-11-03 Applied Materials, Inc. Monitor chemical mechanical polishing process using machine learning based processing of heat images

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7037172B1 (en) * 1999-04-01 2006-05-02 Beaver Creek Concepts Inc Advanced wafer planarizing
US7001243B1 (en) * 2003-06-27 2006-02-21 Lam Research Corporation Neural network control of chemical mechanical planarization
US20070224915A1 (en) * 2005-08-22 2007-09-27 David Jeffrey D Substrate thickness measuring during polishing
US20100129939A1 (en) * 2008-11-26 2010-05-27 Applied Materials, Inc. Using optical metrology for within wafer feed forward process control
KR20120123493A (en) * 2010-01-29 2012-11-08 도쿄엘렉트론가부시키가이샤 Method and system for self-learning and self-improving a semiconductor manufacturing tool

Also Published As

Publication number Publication date
CN111133560B (en) 2023-09-19
KR102450925B1 (en) 2022-10-06
KR20200047739A (en) 2020-05-07
TWI783037B (en) 2022-11-11
CN111133560A (en) 2020-05-08
TW201921273A (en) 2019-06-01
US12524675B2 (en) 2026-01-13
WO2019060577A4 (en) 2019-05-09
JP7036908B2 (en) 2022-03-15
US20190095797A1 (en) 2019-03-28
JP2020535644A (en) 2020-12-03

Similar Documents

Publication Publication Date Title
US12524675B2 (en) Semiconductor fabrication using machine learning approach to generating process control parameters
US12020159B2 (en) Training spectrum generation for machine learning system for spectrographic monitoring
TWI903225B (en) Machine learning systems for monitoring of semiconductor processing
JP7561817B2 (en) Spectroscopic monitoring using neural networks.
US12451380B2 (en) Semiconductor fabrication using process control parameter matrix
TWI873407B (en) Method of training a neural network for spectrographic monitoring and chemical mechanical polishing system

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18859010

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2020516913

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 20207011805

Country of ref document: KR

Kind code of ref document: A

122 Ep: pct application non-entry in european phase

Ref document number: 18859010

Country of ref document: EP

Kind code of ref document: A1