Dynamic process control in semiconductor manufacturing

Abstract

Methods and systems are provided for dynamic process control in substrate processing, such as in semiconductor manufacturing applications. Some example systems and methods are provided for advanced monitoring and machine learning in atomic layer deposition (ALD) processes. Some examples also relate to dynamic process control and monitoring for chamber parameter matching and gas line fill time.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This patent application claims priority to U.S. Provisional Patent Application Serial No. 62 / 885,667, filed on August 12, 2019 by Kumar et al., and entitled “Dynamic Process Control In Semiconductor Manufacturing,” the entire disclosure of which is incorporated herein by reference. Technical Field

[0003] The present disclosure relates generally to dynamic process control in substrate processing, and in some examples, to systems and methods for advanced monitoring and machine learning in atomic layer deposition (ALD) processes. Some examples also relate to dynamic process control and monitoring of chamber matching and gas line fill time. Background Art

[0004] Currently, many, if not most, parameters associated with substrate processing chambers are monitored to operate around component set points. For example, a mass flow controller (MFC) flow rate or chamber pressure may include a certain error range. Typically, an upper or lower parameter limit can be set to a value or a percentage to accommodate this error. For example, in an ALD process, the opening and closing times of a valve can be monitored and these times can be taken into account in the monitored parameters accordingly.

[0005] However, current process monitoring methods are generally only suitable for detecting relatively extensive or severe failures of a process chamber or its components. Such coarse detection may be acceptable in steady-state or single-step situations, such as in chemical vapor deposition (CVD) or plasma enhanced chemical vapor deposition (PECVD) processes, but its use or application in multi-step processes such as ALD where chamber conditions can change within milliseconds is limited.

[0006] The background description provided here is for the purpose of generally presenting the context of the present disclosure. The work of the presently designated inventors is neither explicitly nor implicitly admitted to be prior art against the present disclosure to the extent that it is described in this background section and in aspects of the specification that were not determined to be prior art at the time the application was filed. Summary of the invention

[0007] The present disclosure generally relates to dynamic process control in semiconductor substrate manufacturing systems. In some examples, systems and methods for advanced monitoring and machine learning in ALD processes are provided. Some examples also relate to dynamic process control and monitoring of chamber matching and gas line fill times. Some examples are described in the context of semiconductor processing, but may be equally applicable to substrate processing outside of this context, such as for metals and dielectrics, such as photomasks.

[0008] Some examples monitor consecutive ALD cycles individually and match each step of the ALD cycle using curve fitting or defined error margins. In some examples, each ALD cycle is rendered repeatable and a time reference can be defined to monitor the repeatability of different variables. In some examples, consecutive ALD cycles of different measured variables can be compared. Exemplary variables can include chamber pressure, precursor delivery pressure (or precursor manifold pressure), radio frequency (RF) energy reflection, forward power, and burst sweep pressure. Other monitored variables and parameters are possible.

[0009] Some examples provide tool warnings or error messages based on deviations from monitored values. Some examples include algorithms and software used to achieve these goals.

[0010] In an exemplary embodiment, a system for monitoring a process cycle in an atomic layer deposition (ALD) semiconductor manufacturing process is provided. An exemplary system includes a processing chamber for an ALD manufacturing process; one or more controllers configured to perform process monitoring operations, the operations including: defining a datum time reference for an ALD cycle based on a repetitive action in the manufacturing process; accessing a golden curve including a series of parameter values ​​for a series of data points in a cycle time increment based on the reference time; accessing a variability or tolerance range for each data point in the golden curve; collecting parameter data based on the cycle time increment of a cycle in the ALD manufacturing process; dynamically monitoring whether a parameter value in the parameter data falls within the variability or tolerance margin of the data point; and adjusting the manufacturing process to match the parameter value in a subsequent cycle with the associated parameter value in the golden curve based on a determination that the parameter value falls outside the variability or tolerance margin.

[0011] In some examples, the repetitive action that forms the basis of the reference time basis includes opening or closing a designated valve that supplies the process chamber. In some examples, the parameter data is collected at regular intervals based on a collection frequency, and the collection frequency is in the range of 0-1 Hz, or 1-10 Hz, or 10-100 Hz, or 100-1000 Hz.

[0012] In some examples, the regular intervals are based on trigger points in the ALD fabrication process, each trigger point defining or based on a point in time in a step in the ALD fabrication process.

[0013] In some examples, the operations further include comparing parameter data collected at the trigger point with corresponding sets of parameter data in the golden curve. In some examples, the parameter data includes parameter values ​​associated with one or more of precursor manifold pressure, sweep pressure, transition manifold pressure, chamber pressure, gas flow, RF reflected power, and RF forward power.

[0014] In another example, an exemplary system includes: a processing chamber for the manufacturing process; and one or more controllers configured to perform process monitoring operations, the operations including: identifying a parameter of the manufacturing process; generating a first curve of parameter values ​​including a first parameter value based on a first cycle of the manufacturing process; identifying a second value of the parameter for a second cycle of the manufacturing process based on the generated curve for the first cycle; and adjusting the manufacturing process to match the first parameter value with the second parameter value.

[0015] In some examples, the operations further include generating a second curve including a plurality of parameter values ​​derived from the second cycle of the manufacturing process; and curve fitting the second curve of parameter values ​​to the first curve of parameter values.

[0016] In some examples, the curve fitting operation includes: fitting a series of curves based on parameter values ​​of third and subsequent cycles of the manufacturing process or derived from third and subsequent cycles of the manufacturing process to the first or second curve to generate a golden curve defining a set of golden parameter values ​​for the manufacturing process. In some examples, each cycle of the manufacturing process includes multiple steps in an ALD process; and the operation also includes: matching the parameter values ​​in each step of the ALD process with the parameter values ​​in the set of golden parameter values.

[0017] In some examples, the golden curve includes golden parameter values ​​for each step in the ALD process. In some examples, the identified parameter is associated with a control variable of the manufacturing process; and wherein the operations further include: directly or indirectly using the first and second parameter values ​​to identify a match between a value of the control variable in the second loop and a value of the control variable in the first loop, and adjusting the manufacturing process to match the first parameter value with the second parameter value.

[0018] In another example, a self-learning system for monitoring process steps in a semiconductor manufacturing cycle is provided. The exemplary system includes a processing chamber for implementing a semiconductor manufacturing process including a series of repeated semiconductor manufacturing cycles, each cycle including multiple process steps; and one or more controllers associated with the processing chamber, the controllers configured to perform process monitoring operations, the operations including: generating a set of golden reference parameter values ​​for each step in the series of repeated cycles based on parameter data collected from the processing chamber; generating a machine learning model based on the set of golden reference values; and matching the parameter value in the second cycle in the series of repeated cycles with the corresponding parameter value in the first cycle in the series of repeated cycles using the machine learning model.

[0019] In some examples, the semiconductor manufacturing process is an ALD process; and the steps of each cycle in a series of repeated cycles include sequential steps including dosing, sweeping, converting, and sweeping steps.

[0020] In some examples, the operations further include: generating a golden curve including parameter value data for each sequential step; and using the golden curve as training data for the machine learning model.

[0021] In some examples, the operations further include repeating and matching each loop in the series of repeated loops by matching parameter values ​​in each sequential step of the second loop with corresponding parameter values ​​in each sequential step of the first loop based on a machine learning model.

[0022] In some examples, the parameters include one or more of a precursor manifold pressure, a purge pressure, a chamber pressure, a gas flow rate, a chamber temperature, an RF reflected power, and an RF forward power.

[0023] In some examples, the operations further include: collecting performance data from the process chamber; identifying drift in the performance data and generating performance drift data; and incorporating the drift data into training data for the machine learning model.

[0024] In another example, a system for monitoring a process cycle in a semiconductor manufacturing process is provided. Here, an exemplary system includes: a process chamber for the manufacturing process; and one or more controllers configured to perform process monitoring operations, the operations including identifying a parameter of the manufacturing process; and generating a first curve of parameter values ​​including a first parameter value based on a first cycle of the manufacturing process, the parameter value including a gas line fill time of a line supplying the process chamber.

[0025] In some examples, the operations further include: calculating the gas line fill time based on a pressure ramp time between opening of a valve in a line supplying the process chamber and determination of a constant pressure increase thereafter. In some examples, the operations further include: identifying a second value of the parameter for a second cycle of the manufacturing process based on the generated curve for the first cycle; generating a second curve including a plurality of parameter values ​​derived from the second cycle of the manufacturing process; and curve fitting the second curve of parameter values ​​to the first curve of parameter values.

[0026] In some examples, the curve fitting operation includes: fitting a series of curves based on parameter values ​​of third and subsequent cycles of the manufacturing process or derived from third and subsequent cycles of the manufacturing process to the first or second curve to generate a golden curve defining a set of golden parameter values ​​for the manufacturing process. In some examples, each cycle of the manufacturing process includes multiple steps in an ALD process; and the operation also includes: matching the parameter values ​​in each step of the ALD process with the parameter values ​​in the set of golden parameter values.

[0027] In some examples, the golden curve includes golden parameter values ​​for each step in the ALD process.

[0028] In some examples, the identified parameter is associated with a control variable of the manufacturing process; and the operation also includes: directly or indirectly using the first and second parameter values ​​to identify a match between a value of the control variable in the second loop and a value of the control variable in the first loop, and adjusting the manufacturing process so that the first parameter value matches the second parameter value. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Some embodiments are shown by way of example and not limitation in the figures of the accompanying drawings:

[0030] Figure 1 is a schematic diagram of a reaction chamber according to some examples in which some examples of the methods of the present disclosure can be employed.

[0031] Figure 2A-2C Aspects of a general method for monitoring room parameters according to an exemplary embodiment are shown.

[0032] Figure 3 Graph 300 depicts an exemplary curve fit of thirty (30) ALD cycles according to an exemplary embodiment.

[0033] Figure 4 Depicted is a graph including tolerance margins according to an exemplary embodiment.

[0034] Figure 5 A flowchart is depicted of exemplary operations in a method according to an illustrative embodiment.

[0035] Figure 6 A table of exemplary steps and parameters in an ALD cycle according to exemplary embodiments is included.

[0036] Figure 7 A table of exemplary golden and modified values ​​for process control in an ALD cycle according to exemplary embodiments is included.

[0037] Figure 8 An exemplary matching operation in a matching method according to an exemplary embodiment is depicted.

[0038] Fig. 9 Depicted are exemplary chamber-to-chamber differences according to exemplary embodiments.

[0039] Fig.10 Depicted is a table including sets of parameters for adjusting precursor manifold pressure in an ALD cycle according to an exemplary embodiment.

[0040] Fig.11 An exemplary matching operation in a matching method according to an exemplary embodiment is depicted.

[0041] Fig.12 Depicted are exemplary control operations in a method for dynamic process control in a chamber according to exemplary embodiments.

[0042] Fig.13 Depicted is a table including examples of parameters that may be modified to match different control variables according to an exemplary embodiment.

[0043] Fig.14 Depicted are steps in an ALD cycle according to an exemplary embodiment.

[0044] Fig.15 Operations in a self-learning monitoring method are depicted according to an exemplary embodiment.

[0045] Fig.16 Operations in a data collection method according to an exemplary embodiment are depicted.

[0046] Fig.17 An arrangement of valves according to an exemplary embodiment is depicted.

[0047] Fig.18 is a schematic diagram of a general turning arrangement for a gas manifold according to an exemplary embodiment.

[0048] Fig.19 is a schematic diagram of a local site diverting arrangement for a gas manifold according to an exemplary embodiment.

[0049] Fig. 20 Aspects of a method for determining a gas line fill time according to an exemplary embodiment are shown.

[0050] Fig.21 Aspects of a method for determining a gas decay (or residence) time according to an exemplary embodiment are depicted.

[0051] Figure 22-25A flowchart is depicted of exemplary operations in a method according to an illustrative embodiment.

[0052] Fig.26 is a block diagram illustrating an example of a machine upon which one or more exemplary embodiments may be implemented or by which one or more exemplary embodiments may be controlled. DETAILED DESCRIPTION

[0053] The following description contains systems, methods, techniques, instruction sequences, and computing machine program products that implement the illustrative embodiments of the present disclosure. In the following description, for the purpose of illustration, many specific details are described to provide a complete understanding of the exemplary embodiments. However, it will be apparent to those skilled in the art that the present disclosure can be practiced without these specific details.

[0054] Portions of the disclosure of this patent document contain material that is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure, as it appears in the Patent and Trademark Office patent files or records, but otherwise reserves all copyright rights whatsoever. The following statement applies to any data described below and in the accompanying drawings that constitute a part of this document: Copyright Lam Research Corporation, 2019-2020, All Rights Reserved.

[0055] Reference now Figure 1 , an example of a plasma-based processing chamber is shown. The present subject matter can be used for various semiconductor manufacturing and substrate processing operations, but in the examples shown, a plasma-based processing chamber is described in the context of a plasma-enhanced or free radical-enhanced CVD or ALD operation. Those skilled in the art will also recognize that other types of ALD processing techniques are known (e.g., thermal-based ALD operations) and can be combined with non-plasma-based processing chambers. An ALD tool is a special type of CVD processing system in which an ALD reaction occurs between two or more chemical substances. The two or more chemical substances are referred to as precursor gases and are used to form a thin film deposition of material on a substrate (e.g., a silicon substrate used in the semiconductor industry). The precursor gases are sequentially introduced into the atomic layer deposition processing chamber and react with the substrate surface to form a deposited layer. Typically, the substrate repeatedly interacts with the precursor to slowly deposit one or more material film layers that are increasingly thicker on the substrate. In certain applications, a variety of precursor gases can be used to form one or more thin films of various types in a substrate manufacturing process.

[0056] Figure 1A plasma-based processing chamber 101 is shown in which a showerhead 103 (which may be a showerhead electrode) and a substrate support assembly 107 are arranged. The substrate support assembly 107 may include a pedestal, such as described in more detail below. Generally, the substrate support assembly 107 seeks to provide a substantially isothermal surface and may serve as a heating element and heat sink for the substrate 105. The substrate support assembly 107 may include an electrostatic chuck (ESC) including a heating element to assist in processing the substrate 105 as described above. The substrate 105 may include a substrate comprising an elemental semiconductor (e.g., silicon or germanium), a substrate comprising a composite element (e.g., gallium arsenide (GaAs) or gallium nitride (GaN)), or a variety of other substrate types (including conductive, semiconductive, and non-conductive substrates).

[0057] In operation, a substrate 105 is loaded onto the substrate support assembly 107 through the loading port 109. A gas line 113 can supply one or more process gases (e.g., precursor gases) to the showerhead 103. The showerhead 103, in turn, delivers the one or more process gases into the plasma-based processing chamber 101. A gas source 111 (e.g., one or more precursor gas ampoules) that supplies the one or more process gases is coupled to the gas line 113. In some examples, an RF power source 115 is coupled to the showerhead 103. In other examples, the power source is coupled to the substrate support assembly 107 or the ESC.

[0058] A point of use (POU) and manifold combination (not shown) controls one or more process gases into the plasma-based processing chamber 101 before entering the showerhead 103 and downstream gas lines 113. In the case of a plasma-based processing chamber 101 for depositing thin films in a plasma enhanced ALD (PEALD) operation, the precursor gases may be mixed in the showerhead 103.

[0059] In operation, the plasma-based processing chamber 101 is evacuated by a vacuum pump 117. RF power is capacitively coupled between the showerhead 103 and a lower electrode (not explicitly shown) contained in or on the substrate support assembly 107. The substrate support assembly 107 is typically provided with two or more RF frequencies. For example, in various embodiments, the RF frequency can be selected from at least one of about 1 MHz, 2 MHz, 13.56 MHz, 27 MHz, 60 MHz and other desired frequencies. Coils for blocking or partially blocking specific RF frequencies can be designed as needed. Therefore, the specific frequencies discussed here are provided only for ease of understanding. RF power is used to excite one or more processing gases into a plasma in the space between the substrate 105 and the showerhead 103. Plasma can help deposit various layers (not shown) on the substrate 105. In other applications, plasma can be used to etch device features into various layers on the substrate 105. RF power is coupled at least through the substrate-support assembly 107. The substrate-support assembly 107 may have a heater ( Figure 1 Detailed design of the plasma based processing chamber 101 may vary.

[0060] In some examples, within a given total cycle, ALD can be considered a multi-step process (e.g., primarily four steps), including dosing, sweeping, conversion, and sweeping steps performed within the process chamber. Other cycles and steps are possible. Unlike some PECVD processes, where parameters such as gas flow rates, chamber pressure, and RF values ​​are kept constant throughout the deposition process (cycle), in ALD, these process parameters (and others) can vary with each step within the entire cycle or even in consecutive cycles.

[0061] Unmonitored parameter changes during different steps can mask fluctuations in several key control variables. For example, chamber pressure typically seeks to be maintained at a constant value or set pressure. Fluctuations in gas flow to the chamber during different steps in a multi-step process are typically controlled by a throttle valve that moves continuously seeking to maintain the set pressure. However, constantly changing gas flows and moving throttle valves can be subject to inherent feedback delays, which often result in poor control in ALD processes, for example.

[0062] Similarly, during processing, when a precursor gas, such as argon (Ar), is admitted to the chamber or directed to a bypass, the precursor manifold pressure in the gas line supplying the precursor gas almost always fluctuates due to surges or delayed changes in gas flow. In other words, a given gas line pressure can vary within a range, depending on whether the gas is flowing to the chamber (e.g., creating a sudden sweep pressure) or to a diverter (e.g., creating a restricted flow pressure). These fluctuations are noticeable and unwelcome in many, if not all, ALD cycles. Parameter deviations from setpoints can indicate chamber processing problems and can affect substrate and film properties.

[0063] refer to Figure 2A-2C , conventional methods of monitoring chamber parameters may include, by way of example only, monitoring chamber pressure ( Figure 2A ), precursor manifold pressure ( Figure 2B ), and the sweep pressure near the set point 202 ( Figure 2C ), and an error band 204 set between an upper limit 206 and a lower limit 208. Error band 204 is typically relatively large and in fact so large that it cannot detect smaller, potentially significant fluctuations that occur during an ALD cycle. Therefore, conventional efforts based on set points and error bands cannot accurately or truly monitor more detailed aspects or chamber conditions during an ALD cycle. Today's substrate manufacturers seek more accurate, controllable, and increasingly detailed chamber control to create high depth and width nano-sized structures and semiconductor devices on substrates.

[0064] In this regard, reference is now made to Figure 3 . In some examples, curve fitting is used to monitor and match the successive dosing, sweeping, converting, and sweeping (again) steps within an entire deposition or etch cycle. Some examples include defined error ranges configured for each specific step, rather than a general error band defined for a complete cycle of the type described above. In some examples, each step in an ALD cycle (e.g., each of the four steps described above) is rendered repeatable based on a curve fit or defined step-specific tolerance margins. A time reference may be defined to monitor the repeatability of different variables.

[0065] In some examples, a frame of reference for process parameter monitoring is established for subsequent steps or cycles by an earlier or initial step or cycle, as opposed to, for example, being established by conventional set points and error bands. In fact, some examples are independent of (i.e., ignore) conventional set points or error bands and operate based on repeating earlier cycles or steps in a given substrate manufacturing process. For example, parameters in an ALD cycle can be adjusted through trial and experimentation to produce a desired structure on a substrate. Without necessarily knowing or identifying what the absolute values ​​of those successful parameters are, examples of the present disclosure allow a processing chamber to be simply configured to repeat successful cycles based on data derived from a curve fit or defined by step-specific tolerance margins.

[0066] Figure 3 A graph 300 of example curve fits for thirty (30) ALD cycles is depicted. Chamber pressure (y-axis) is monitored at time increments of 100 milliseconds (ms) over the duration of each ALD cycle. As shown, the elapsed time between the start and end of each ALD cycle, where chamber pressure data is collected at each time increment over the duration, is in the range of about one second or less. During the course of successive steps (e.g., dosing, sweeping, conversion, and sweeping) within each ALD cycle, a set of curves 306 representing chamber pressure fluctuate between data peaks 302 and valleys 304. In the illustrated example, the beginning of each ALD cycle begins with a dosing step.

[0067] The curve fitting operation is performed for each curve 306 in the set of curves such that successive or subsequent ALD cycles repeat earlier cycles. Each step in the ALD cycle may also repeat the previous step. In this regard, a tolerance margin for a particular step may be established for one (or more) time increments associated with that step. The tolerance margin may be based on the time increments in the ALD cycle. Figure 4 400 of the type depicted in FIG. Here, for each time increment, the variability (or tolerance margin) of the chamber pressure may be observed. For example, a relatively wide (or loose) tolerance margin 402 may correspond to a tolerance margin 404 of the chamber pressure. Figure 3 In contrast, for example, a relatively narrow (or tight) tolerance margin 404 may correspond to a relatively wide set of chamber pressures represented by the curves set at 0.4 time increments. Figure 3 A relatively narrow (or even overlapping) set of chamber pressures represented by curves set at 0.1 time increments of 0.1 of the ALD cycle. Variability or tolerance margins can be defined for each time increment during a step or ALD cycle. One or more curves 306 within a set of curves can also be defined independently or based on tolerance margins. For example, a set of one or more curves 306 and / or step-specific tolerance margins can be established and used for step and cycle monitoring and repeatability during ALD cycles in substrate processing operations.

[0068] For example, Figure 5 A flow chart depicts exemplary operations of a monitoring method 500 for a step or cycle in an ALD process. In some examples, operation 502 includes defining a time reference point (start) of an ALD cycle (e.g., the opening of a certain dosing valve). In some examples, operation 502 may also include defining a "golden curve" or golden value for a series of data points (e.g., time increments). In some examples, operation 502 also includes defining a variability or tolerance margin for each data point (time increment).

[0069] In some examples, operation 504 includes collecting parameter data. In some examples, the parameter data is collected at regular intervals (eg, at so-called trigger points) based on a collection frequency (eg, 1 Hz, 10 Hz, 100 Hz, or 1000 Hz).

[0070] In some examples, operation 506 may include data comparison, such as comparing data of various parameters, such as chamber pressure, precursor manifold pressure, sweep pressure, RF reflected power, and RF forward power. The parameter values ​​at a particular time increment (trigger point) are compared to a golden curve or golden value. In some examples, the data comparison may be performed in real time at the end of a given cycle or step, or at the end of a given substrate manufacturing process. It may be user defined or based on process needs or optimization.

[0071] In some examples, operation 508 includes generating an algorithm that enables repeatability of steps and cycles. Exemplary algorithms may include (or be based on) one or more of curve fitting, standard deviation from a golden value, or minimum or maximum range in variability or tolerance margin. Other algorithmic factors are possible.

[0072] In some examples, operation 510 includes reporting (e.g., generating an output of a curve fit, or generating a variability report). For example, in some examples, operation 510 may include identifying and / or taking corrective actions based on the variability report to adjust the process parameters or bring the process parameters within the set of fitted curves or tolerances. For example, if the process parameters fall within the set of fitted curves or tolerances, some examples may include taking no action. Tool or chamber warnings may be generated (or not generated) accordingly.

[0073] Process parameter fluctuations can also make process control and monitoring in other areas difficult. For example, differences between substrates (or batches) may be caused by the accumulation of chamber heat during substrate processing. Differences between tools may be caused by differences in pump efficiency. Traditionally, the main efforts to control changes have focused on monitoring the performance of individual devices. Exemplary devices and their associated parameters can include MFC flow rates, where the device error limit is set to 1% of the flow rate. During substrate processing, the MFC flow rate is monitored to operate within this limit. Further devices and parameters can include valve control timing (for example, ALD valve control timing that is monitored to operate within an opening time of 50ms and a closing time of 70ms). In another example, the valve can be set to switch between open and closed positions at 25ms. In other examples, the susceptor temperature can be controlled using thermocouples to monitor deviations within a set range. RF power control can include monitoring of forward and reflected power. These devices are often affected by inherent performance or response limitations, which can lead to poor chamber control and random or fluctuating chamber conditions. Inadequate monitoring limits and the diversity of process factors and equipment limits may lead to such adverse effects.

[0074] In addition, reference Figure 6 , unlike most PECVD processes, where parameters such as gas flow, pressure, and RF power are typically kept constant during the entire deposition process, some parameters in each cycle are continuously varied during the ALD process (cycle). Figure 6 Table 600 in FIG. 6 shows exemplary steps and associated parameters in an ALD cycle. As shown, these steps may include dosing, post-dosing purge, RF power application, and purge.

[0075] Some examples herein seek to address such challenges and include matching of measured parameters rather than monitoring of equipment parameters. For example, some examples are thus configured to be able to match equipment and / or chamber performance for a given set of processing steps or cycles and across a set of processing chambers in a processing tool. For example, pressure fluctuations or pressures generated during each ALD step can be matched cycle by cycle or across tools and chambers by adjusting one or more process parameters occurring in the various steps of a cycle. For example, a given pressure set point in a single step (or cycle) in a substrate manufacturing process can be matched (or repeated) across steps or cycles using dynamic and real-time adjustment of gas flows.

[0076] Similarly, in some examples, different levels of precursor manifold pressure affect the deposition rate (depR) and may be a key variable in chamber matching. The precursor manifold pressure is in turn affected by one or more exemplary factors (e.g., precursor gas (e.g., Ar) plug flow, precursor flow (or ampoule temperature), pumping efficiency through the diverter, and chamber outlet or diversion timing). As described above, current technology controls and monitors equipment set points, such as MFC flow, valve opening and closing time, RF generator power, etc., and attempts to control chamber parameters to set points. Variability between substrates, batches, and tools is controlled by using equipment parameters and monitoring them. It is assumed that matching comes from control of equipment parameters. However, the actual conditions in the processing chamber or gas pipeline are not controlled or matched within the time limits imposed by the ALD steps and cycles.

[0077] Some examples include dynamic process control and monitoring to achieve chamber matching. Examples may include matching of measured parameters rather than matching equipment parameters to achieve matched chamber performance. In some examples, for example, a "golden curve" or golden value may represent a set of desired parameter values ​​that are used to form a given or desired substrate in a chamber. A chamber may be operated with control variables set to "golden" conditions or values. In some examples, a "golden curve" may be based on parameter data derived from one step or cycle, which is then used as a benchmark or reference point to repeat the step or cycle, or to match parameter values ​​or control variables across, for example, chambers, substrates, and tools.

[0078] An exemplary set of golden values ​​for relevant steps in the ALD cycle appears in this paper Figure 7 700. The pressure and gas flow rows include two values, respectively: a "golden" value (labeled simply as pressure or flow) and a "modified" pressure or flow. In some examples, during operation, the pressure in each step is modified so that the monitored pressure curve matches the golden value of the golden chamber. Similarly, the gas flow in each step can be modified to match the golden flow value or the golden pressure value. Adjustment of one parameter can result in a direct change in that parameter or an indirect change in another parameter that is associated with the first parameter (e.g., a change in chamber gas flow that affects a correlated change in chamber gas flow pressure).

[0079] In some examples, pressure fluctuations or pressures present in the first step of an ALD cycle are matched in subsequent steps (or cycles) by adjusting the pressure set point. Thus, in some examples, the process is controlled to match earlier process parameters as a criterion, rather than by adjusting to given equipment parameters. In some examples, the matching pressure adjustments are performed in separate steps or through the cycle. In some examples, pressure matching is performed directly or indirectly, for example, by matching earlier gas flows.

[0080] Figure 8 808. An exemplary matching operation in matching method 800 is shown in FIG. 802. In graph 802, it will be seen that the tool A pressure curve 804 in the first step or cycle does not match the tool B pressure curve 806 in the second step or cycle. Their respective value curves do not match and are offset relative to each other. In the example shown, it can be said that the tool B pressure values ​​in curve 806 lag behind the tool A pressure values ​​in curve 804. In some examples, the chamber parameters (e.g., gas flow or pressure parameters) are adjusted so that the two curves 804 and 806 are substantially matched as shown in graph 808. Accurately matched steps or cycles help ensure repeatability of chamber conditions and consistency of substrate manufacturing, thereby improving the creation accuracy of substrate structures and semiconductor devices.

[0081] In other examples, the precursor manifold pressure profiles for two different chambers during an ALD cycle may not match even if all relevant equipment on the tool is operating within specifications. Chamber-to-chamber variability may be caused by pumping efficiency differences, valve timing differences in the precursor manifold pressure and chamber valves, gas line temperatures, etc. Exemplary chamber-to-chamber variability is Fig. 9 900, the differences in pressure of the chambers can be observed during time period 902. In graph 904, the differences in pressures of the various chambers are discernible. The pressure curves 906 and 908 are shown offset (mismatched) from each other with respect to time and pressure magnitude.

[0082] Some examples address the issue of chamber-to-chamber matching. In this regard, a set of parameters for adjusting the precursor manifold pressure during an ALD cycle is presented in the present document. Fig.10 1000 in Table 1000. Exemplary parameters include precursor gas flow (e.g., argon Ar), precursor flow, ampoule inlet and outlet valve control timing, and chamber inlet and outlet valve control timing. Exemplary parameters may have a golden value or golden curve associated with each of them. Other parameters are possible.

[0083] Control or influence over the precursor manifold pressure may be imparted by making certain adjustments to certain parameters, such as noted in the notes section of table 1000. For example, argon plug flow may affect the magnitude of the precursor manifold pressure and may shift the pressure curve in a vertical direction. In other words, if the precursor manifold pressure curve established in a second or subsequent step (or cycle) is to match a corresponding curve established (or generated by) in an earlier step or cycle, and appears to be "lower" than the earlier curve in the matching graph (e.g., graph 802), then the pressure curve in subsequent cycles (third and fourth cycles, etc.) may be corrected to shift upward by adjusting the argon precursor plug flow accordingly. Thus, as opposed to being based on an inflexible set point or error band, the conditions in the processing chamber may be configured to remain substantially constant by monitoring parameter values ​​derived from earlier steps or cycles.

[0084] Fig.11 1100. Here, in order to establish matched precursor manifold pressures 1102 and 1104 in each step or cycle of the ALD process, changes in parameters such as plug flow (each ALD cycle), ampoule temperature, and valve control timing are utilized. In turn, depending on this, the matched precursor manifold pressures can match deposition rates from chamber to chamber.

[0085] Exemplary operations in a method for dynamic process control may include selecting a control variable (e.g., pressure) to match, determining a parameter to change to modify the control variable, determining a control increment for the parameter, and calculating a deviation of the control variable from a golden curve. Fig.12, an exemplary control operation in a method 1200 for dynamic process control in a chamber may include: at operation 1202, identifying a control variable (e.g., pressure) and establishing a golden curve or golden value for it; at operation 1204, identifying a first adjustment parameter that can affect the value of the identified control variable; at operation 1206, determining a magnitude or step increment of the identified parameter of the control variable; at operation 1208, selecting a parameter increment that has the greatest potential impact on the control variable; at operation 1210, determining whether the control variable is within specification; at operation 1212, if yes, repeating operations 1206 and 1208, or if no, applying the selected parameter increment at operation 1212; at operation 1214, determining whether the control variable is within specification after applying the selected parameter increment. If yes, the method 1200 includes repeating operations 1206 and 1208. If not, the exemplary method 1200 includes, at operation 1216, identifying a second adjustment parameter (e.g., a gas flow setting) that may affect the value of the identified control variable, and repeating the control operations outlined above for the second adjustment parameter at operations 1218 through 1226. Once the selected increments in the first and second adjustment parameters have established a golden value in the identified control variable, at 1218, the chamber is optimized for pressure matching.

[0086] Some examples include methods for determining deviations from a golden curve in dynamic process control. Exemplary operations in the method may include defining a time reference point at the beginning of an ALD cycle (e.g., a digital output signal for opening a valve to dose a chamber). Each data point collected thereafter at a specified interval (e.g., 1 ms, 10 ms, 50 ms, 100 ms) is compared to the corresponding data in the golden curve. The deviation or error limit from the golden curve can be pre-defined based on experiments or based on user specifications. Once a parameter, such as chamber pressure, is optimized for a particular measurement, the effect of the parameter change on other controlled variables (e.g., precursor manifold pressure) is examined and adjusted if necessary. Fig.13 Table 1300 in provides examples of parameters that can be modified to match different control variables.

[0087] Some examples herein enable dynamic parameter control to match actual measurements, such as chamber pressure, gas line pressure, temperature, delivered RF power to achieve matching between substrates, batches, and chambers. Some examples include optimizing parameters to minimize measured deviations from a golden curve or golden value. The optimized parameters may vary from chamber to chamber. The optimized parameters may be different at different accumulations to account for other factors (e.g., drift). Some examples include software functionality for performing parameter optimization for process control (described more fully below). The present process control method can be performed periodically or intermittently, such as during continuous operation or during tool startup or planned maintenance. Parameter optimization can be performed in assigned priorities and, in some examples, can be based on the control variables that have the greatest impact on a given process.

[0088] In further aspects, some examples include self-learning techniques for advanced monitoring processes (e.g., particularly ALD and CVD processes). These techniques attempt to address monitoring problems that may arise when using traditional methods to control traditional equipment. For example, MFCs are typically monitored near a set point within a set error range. Valve timing controls typically monitor the opening and closing times of ALD valves, which may switch between open and closed states in a period as short as 50 milliseconds in a given cycle or process. RF generation is typically measured and controlled based on forward and reflected power, which in turn is monitored within an error band. Chamber and other temperatures are controlled and monitored near a set point within an error band / percentage.

[0089] Conventional efforts to monitor process (as opposed to equipment) parameters involve monitoring chamber pressure around a set point with a set error band. The error band is typically set large enough to ignore the inherent fluctuations that occur during an ALD cycle. Therefore, these efforts do not truly monitor more detailed aspects or chamber conditions during an ALD cycle. Detailed and deep chamber control is increasingly being used to create high aspect ratio nanoscale substrate structures and semiconductor devices. In addition, precursor manifold and burst sweep pressures are also typically monitored around the band. Conventional error bands are typically set too wide to capture smaller fluctuations during an ALD cycle, leading to similar challenges as described above.

[0090] In terms of the RF power delivered, a one-time check of the RF power only checks whether the RF is turned on or off after the RF excitation. The voltage-current (VI) sensor monitors the RF power during the plasma "on" step and monitors only at a frequency higher than the RF power (e.g., 1kHz). Therefore, in a broad sense, the current method is based on limit or error band settings. It is a passive or "inflexible" monitoring method that is usually based on limited data. Typically, the monitoring frequency band is very wide, which cannot handle or even solve the strict process control problems in the increasingly demanding requirements of today's semiconductor manufacturing. Among other shortcomings, the same monitoring band applies to all tools, and there is no tool-to-tool modification or customization. Typically, any customization is performed manually on an ad hoc basis. Little or no substrate-to-substrate or tool-to-tool performance or comparison (whether considering accumulation after preventive maintenance or after hardware replacement) is usually performed in traditional technology.

[0091] As mentioned above, the ALD process can be considered a multi-step process. Fig.14 , a typical ALD cycle 1400 includes four major steps: dosing 1402, sweep 1404, conversion 1406, and sweep 1408. In some current examples, each step in the ALD cycle, as well as subsequent cycles in a given ALD process, is monitored separately for different variables. Curve fitting or defined error margins in a "smart" self-learning monitoring process are used to match the monitored variables in each ALD step and / or cycle. Each ALD cycle is repeatable, and parameters such as chamber pressure, precursor manifold pressure, temperature, etc. can be monitored to achieve repeatability (and make it repeatable) for each cycle or step.

[0092] Operations 1502-1522 in the exemplary self-learning monitoring method 1500 are Fig.15 The method 155 includes: at operation 1502, defining one or more parameters to monitor, such as valve timing, VI sensor, RF forward, RF reflection, pressure (such as chamber, precursor manifold or burst pressure), and other parameters. Initial data can be collected to compare with the starting value and define the basic advantages, and may include a golden value or curve. For example, the golden curve can be as follows Figure 3As shown. Operation 1504 includes data analysis and recording of aspects such as mean, standard deviation, percentage out of control (OOC), and other statistical values. Operation 1504 may also include updating the performance data with new data (e.g., derived statistical values). Operation 1506 may include monitoring chamber performance based on deviations from one or more statistical values ​​(e.g., average performance). Operation 1508 may include determining whether the deviation is within equipment limits. If so, the previous operation, such as operation 1504, may be repeated. Operation 1510 may include determining whether the deviation is outside equipment limits. If not, operation 1512 includes issuing a user review warning. Upon review, if feasible, the earlier operations in the method may be repeated as shown. If not, operation 1522 includes issuing a tool alert. Operation 1514 may include saving the tracker data at a specified interval (which may be user-defined). Operation 1516 includes comparing the performance data to the tracker data to identify data drafts. Based on the comparison, operation 1518 includes identifying whether there is drift in the data. If so, operation 1520 includes issuing a report to obtain user input and corrective action.

[0093] An exemplary self-learning monitoring method may include acquiring data from multiple systems to define in-specification chamber performance for chamber-to-chamber and tool-to-tool performance matching. Fig.16 16 shows exemplary operations 1602-1612 in a data collection method 1600. At operation 1602, a plurality of systems (e.g., process chambers) are identified. At operation 1604, the monitoring system retrieves actual performance or golden curve values ​​of parameters from selected modules in the identified process chambers (or tools). A retrieval frequency is identified. For example, the retrieval frequency can be user-defined, daily, based on tracker data, or weekly.

[0094] In operation 1606, statistical process control (SPC) measurements from each chamber or tool are obtained. Operation 1608 includes comparing and analyzing the performance from chamber to chamber (or tool to tool) for each relevant parameter and defining average performance and standard deviation values ​​for each chamber (e.g., chambers 1-3 in the view). Operation 1608 can include generating outlier performance alerts based on user-defined criteria (e.g., 3σ) and establishing correlations with the obtained SPC data. Operation 1610 includes performing an overall chamber to chamber (or tool to tool) comparison and analysis, and in operation 1612, issuing tool alerts if appropriate.

[0095] In some examples, the intelligent self-learning monitoring system includes machine learning components that create a monitoring process based on a machine learning model. These components may include data preprocessing components. The preprocessing components receive data from, for example, a process chamber or a set of golden values ​​or curves (e.g., Figure 3A preprocessing component preprocesses the training data, including, for example, applying a MapReduce function or similar function to the training data. A feature extraction component can then be used to extract multiple features (e.g., process parameters) from the preprocessed training data and feed these features into a machine learning algorithm. The extracted features can be related to one or more of the above-mentioned control variables. In some examples, the machine learning algorithm learns weights assigned to each feature and applies these weights to the function. The function and the learned weights can be included in or constitute the machine learning model discussed above. One or more machine learning models are stored in a file system and retrieved when it is needed to perform chamber performance analysis or process monitoring.

[0096] Machine learning algorithms can be selected from many different potential supervised or unsupervised machine learning algorithms. Examples of supervised machine learning algorithms include artificial neural networks, Bayesian networks, instance-based learning, support vector machines, random forests, linear classifiers, quadratic classifiers, k-nearest neighbors, decision trees, and hidden Markov models. Examples of unsupervised machine learning algorithms include expectation maximization algorithms, vector quantization, and information bottleneck methods. In an exemplary embodiment, a binary logistic regression model is used. Binary logistic regression processes the observations of dependent variables with only two possible types of situations. Logistic regression is used to predict the probability that one situation or another situation is true based on the value of the independent variable (predictor variable). In a further exemplary embodiment, a boosted tree gradient descent process is used for machine learning.

[0097] Functions contained in the machine learning model can be evaluated at runtime to produce a process match score. The match score is a prediction of the likelihood that attempting a match condition in multiple systems will result in a successful match based on evaluating various parameters and applying feature weights learned by the machine learning algorithm to the features. In some examples, the predicted match can include mixed results, or include the output of a parameter adjustment with an increased confidence that the parameter adjustment will result in a match between chambers or between tools.

[0098] In a further example, a system and method including hardware and software is provided to determine the gas line fill time. In some exemplary deposition or etching systems, the gas is delivered from a gas box and includes several valves and filters between the gas source and the deposition or etching chamber. Typically, MFCs, MFC inlet and outlet valves, filters, and chamber inlet valves are used in this regard. The travel time of the gas from a particular valve to the chamber can be defined as the gas line fill time.

[0099] In some examples, the gas line fill time may be affected by certain factors. These factors may include MFC response and ramp time, which in some examples requires 1 to 3 seconds to bring the associated gas flow to within + / -2% of the set point. Fill time factors may also include valve opening time. In some examples, the valve opening time is in the range of a few milliseconds, including pneumatic delays and other delays. Some examples include valve opening times for pneumatic valves in the range of <100ms. Other factors may include the conductivity of the gas line, including valve conductivity, or the pressure drop at the filter. Fundamentally, the gas velocity depends on the pressure differential, which in turn is affected by or determined by the conductivity of the gas line. For deposition and etching processes that can provide sufficient time for a given gas line fill, any delay in gas line filling will not significantly affect the deposition or etching process, if at all.

[0100] As further described above, the ALD process may include multiple steps that result in the formation of a film on a substrate surface. These steps may include a dosing step for the precursor gas molecules to adhere to the substrate surface, a post-dosing purge to remove excess precursor gas from the chamber, the application of RF power to energize the plasma and convert the adsorbed gas molecule monolayer on the surface into a thin film, and an RF purge to remove reaction byproducts. The gases used in these steps typically come from various manifolds.

[0101] Currently, gas line fill time is not measured or monitored. Advance frequency (kHz) monitoring can be used to monitor valve open and close times. ALD valves may include an optical sensor that detects the position of the valve diaphragm. Valve open and close times are monitored by the time difference of the command (digital input, DI) to the pneumatic group that operates the valve and the readback of the optical sensor (digital output, DO) that senses the movement of the valve diaphragm.

[0102] For example, refer to Fig.17In the arrangement of valves 1700 shown in, in a deposition or etching system, gases are typically delivered from a gas box, and several valves and filters may be provided between one or more gas sources and a deposition / etching chamber. Typically, the valves include MFCs, MFC inlet and outlet valves, filters, and chamber inlet valves. The travel time of the gas from a particular valve to the chamber may be defined as the gas line filling time. The gas line filling time depends on the MFC response and ramp time. In some examples, the valve typically takes 1-3 seconds to bring the gas flow rate within + / -2% of the set point value. The valve opening time typically occurs in the range of a few milliseconds. Including compressed dry air delays and other delays, the valve opening time of a pneumatic valve may be less than 100 milliseconds. The valve conductance may include the pressure drop at the filter. Fundamentally, the gas velocity depends on the pressure difference, which in turn is controlled by the conductance of the gas line. For deposition and etching processes that allow sufficient time for gas line filling, this delay in gas line filling may be irrelevant to the deposition / etching process, but in some cases, the gas line filling delay is unacceptable. The need for shorter ALD cycle times to achieve high throughput (and deep substrate formation capabilities) means that substrate processing operations cannot wait for MFC ramp times or delays. Therefore, in some examples, gas line fill times are determined to rule out potential causes of delays or instabilities therein. Based on the fill times so determined, the MFCs supplying the relevant gases are configured to operate in a continuous or consistent manner.

[0103] The presence of a flow diverter (which allows gas to bypass the chamber when not needed, for example to an exhaust pipe or so-called foreline) can also affect the gas line fill time. Fig.18 is a schematic diagram of a general diverter arrangement 1800 for a gas manifold. Gas is diverted away from the chamber and typically results in a relatively long time to fill the gas lines. Fig.19 is a schematic diagram of a single station diverter arrangement 1900 for a gas manifold. Gases are diverted closer to the chamber, typically resulting in a relatively shorter time to fill the gas lines.

[0104] In an ALD process, a single step time may include the gas line fill time from the nearest outlet valve to the process chamber. Therefore, some exemplary embodiments seek to address the complex and critical need of being able to provide consistent, measurable gas line fill times from tool to tool.

[0105] Some exemplary embodiments automatically measure the gas line fill time at various times, such as at startup, after preventive maintenance, or at regular intervals. The current measurement is automatically compared to the previous measurement, and any changes in the gas line fill time are reported. More generally, the measurement can also be compared to gas line fill times derived from other tools to determine tool-to-tool differences and the suitability or status of a tool.

[0106] Some examples use a chamber pressure gauge to measure or monitor the gas line filling time. For example, in some examples, the time difference between the opening of the manifold outlet valve and the increase in chamber pressure is used to measure or calculate the gas line filling time. Some exemplary methods include establishing a base pressure or a constant pressure in a processing chamber supplied by a gas pump. The method includes closing a throttle valve and / or a diverter valve of the pump to isolate the chamber from the pump. The gas flow is set to be diverted from an applicable manifold or gas line seeking to measure the line filling time. The method also includes opening the gas manifold outlet valve and closing the diverter valve and measuring the increase in chamber pressure. Typically, there is some initial delay before the chamber pressure begins to increase. The pressure gradient curve used in some examples can be used to calculate the gas line filling time or delay. In other words, the initial delay of the pressure gradient represents the gas line filling time to reach the chamber. The calculated delay (filling time) can be taken into account in the control and monitoring system and method, including the intake algorithm, to enhance the stable operation and chamber matching operation of the substrate processing chamber.

[0107] Reference now Fig. 20 Graph 2000 in describes an example method for determining gas line fill time. The graph includes line 2002, which represents the movement (closing or opening) of a gas valve that supplies or removes gas from a processing chamber. In this case, the exemplary valve is a diverter valve, as shown in the legend. The diverter valve controls gas in a line that bypasses the chamber. If the diverter valve is closed, gas is not diverted from the chamber, but can enter the chamber. Gas enters the chamber and the pressure inside the chamber increases. Typically, the closing of the diverter valve coincides with the opening of the chamber supply valve to allow supply into the chamber. In any case, in this example, the gas line fill time of the diverter valve is being determined.

[0108] Line 2002 represents the physical closure of the diverter valve from an open (diverter) position at position "1" on the y-axis to a closed position at "0" on the same axis. When the diverter valve is closed, the pressure in the chamber will increase as described above. This increase is represented by pressure line 2004 in graph 2000. After a period of time, the slope of line 2004 becomes constant, indicating a uniform or stable increase in gas pressure in response to a constant or stable entry of gas into the chamber. The slope of line 2004 will increase as the gas flow increases. The gas line fill time is determined by extrapolating the slope of line 2004 until it intersects the x-axis. The intersection is represented by position 2006 in graph 2000. The gas line fill time is the time period represented by timeline 2008, which extends between intersection 2006 and the final closure of the diverter valve represented by position 2010. In other words, there is a ramp-up or delay time (ie, gas line fill time) between the time the diverter valve closes (ie, allowing gas to fully enter or fully fill the chamber) and the point at which the pressure in the chamber increases at a constant rate.

[0109] In some embodiments, the gas decay time is identified. Closure of the gas outlet valve does not mean that the gas flow to the chamber is immediately stopped. In addition to delays caused by other components, there may be delays in indicating the valve is closed and delays in the valve physically reaching full closure. Even when the valve is closed, gas that is already in the gas line downstream of the valve continues to enter the chamber.

[0110] In some examples, a method for determining a gas decay (or residence) time may include one or more of the following operations. Initially, the chamber is implemented at a base pressure or a constant pressure. The chamber is supplied by a throttle valve and a slit valve to control the entry of gas from a supply pump. The supply throttle valve and slit valve of the supply chamber are closed to isolate the chamber from the supply pump. The gas flow is set to be diverted from the manifold or gas line seeking to measure the gas decay time. The gas manifold outlet valve is opened and the diverter valve is closed. The increase in chamber pressure is measured. The diverter valve is then opened and the gas outlet valve is closed. The method then includes measuring the time when the chamber pressure is maximized or stabilized. The time of this maximization or stabilization period is a measure of the time required for the gas to stop flowing into the chamber after the valve closing instruction is sent to the outlet valve (i.e., the gas decay or residence time).

[0111] Graph 2100 depicting exemplary gas line fill times and exemplary gas decay times is shown in FIG. Fig.21 . Line 2102 in the graph represents the opening and closing of the valve supplying the process chamber. The valve is closed during the time period indicated by 2104. The valve opens and allows gas to enter the chamber during the time period 2106, thereby forming a pressure in the chamber. The valve is closed again at 2108. The chamber pressure is measured by the chamber pressure gauge and represented by the pressure line 2110. As described above, the gas line filling time can be determined based on the time period or increment indicated at 2112 (i.e., the period between the intersection of the extrapolation of the pressure slope with the x-axis and the moment when the valve is opened indicated at 2102). The gas decay time is determined by the time it takes to reach stability after the valve is closed at 2108. The gas decay time can be more clearly seen in the expanded view of 2114 (80ms in this example).

[0112] Therefore, some examples provide methods for measuring gas line filling time using chamber pressure and methods for measuring gas decay (or residence) time using chamber pressure. These values ​​can be integrated into dynamic monitoring processes and software for automatic parameter measurement, chamber control and matching techniques. Filling and decay (residence) values ​​may be important for monitoring tool-to-tool (or chamber-to-chamber) variability and daily drift of gas line filling or decay over time. Such methods can be applied to product lines that use multi-step procedures (or cycles) for deposition or etching and require very fast steps and cycle times. Faster steps and cycle times can make gas line filling and decay times critical. Embodiments of the present invention enable their measurement and monitoring.

[0113] Exemplary embodiments may include methods. Fig. 22 , a method 2200 for monitoring a processing cycle in an ALD semiconductor manufacturing process includes: at operation 2202, defining a reference time base for an ALD cycle based on a repetitive operation in the manufacturing process; at operation 2204, accessing a golden curve including a series of parameter values ​​for a series of data points in a cycle time increment based on the reference time; at operation 2206, accessing a variability or tolerance margin for each data point in the golden curve; at operation 2208, collecting parameter data based on the cycle time increment of the cycle in the ALD manufacturing process; at operation 2210, dynamically monitoring whether a parameter value in the parameter data falls within the variability or tolerance margin at the data point; at operation 2212, based on determining that the parameter value falls outside the variability or tolerance margin, adjusting the manufacturing process so that the parameter value in a subsequent cycle matches the relevant parameter value in the golden curve.

[0114] In some examples, the repetitive action that forms the basis of the reference time basis includes opening or closing a designated valve that supplies the process chamber.

[0115] In some examples, parameter data is collected at regular intervals based on a collection frequency, the collection frequency being in the range of 0-1 Hz, or 1-10 Hz, or 10-100 Hz, or 100-1000 Hz.

[0116] In some examples, the regular intervals are based on trigger points in the ALD fabrication process, each trigger point defining or based on a point in time in a step in the ALD fabrication process.

[0117] In some examples, the operations further include comparing parameter data collected at the trigger point with corresponding sets of parameter data in a golden curve.

[0118] In some examples, the parameter data includes parameter values ​​associated with one or more of precursor manifold pressure, sweep pressure, switch manifold pressure, chamber pressure, gas flow rate, RF reflected power, and RF forward power.

[0119] refer to Fig.23 , a method 2300 for monitoring a processing loop in a semiconductor manufacturing process includes: at operation 2302, identifying a parameter of the manufacturing process; at operation 2304, generating a first curve of parameter values ​​including a first parameter value based on a first loop of the manufacturing process; at operation 2306, identifying a second value of the parameter for a second loop of the manufacturing process based on the curve generated for the first loop; and at operation 2308, adjusting the manufacturing process to match the first parameter value with the second parameter value.

[0120] In some examples, the operations further include: generating a second curve including a plurality of parameter values ​​derived from a second cycle of the manufacturing process; and curve fitting the second curve of parameter values ​​to the first curve of parameter values.

[0121] In some examples, the curve fitting operation includes fitting a curve based on a series of parameter values ​​of or derived from third and subsequent cycles of the manufacturing process to the first curve or the second curve to generate a golden curve for the manufacturing process that defines a set of golden parameter values.

[0122] In some examples, each cycle of the manufacturing process includes multiple steps in the ALD process; the operations further include: matching parameter values ​​in each step of the ALD process with parameter values ​​in the set of golden parameter values.

[0123] In some examples, the golden curve includes golden parameter values ​​for each step in the ALD process.

[0124] In some examples, the identified parameter is associated with a control variable of a manufacturing process; and the operation also includes: directly or indirectly using the first and second parameter values ​​to identify a match between a value of the control variable in the second loop and a value of the control variable in the first loop; and adjusting the manufacturing process to match the first parameter value to the second parameter value.

[0125] refer to Fig.24 , a machine learning method 2400 for monitoring processing steps in a semiconductor manufacturing cycle includes: at operation 2402, generating a set of golden reference parameter values ​​for each step in a series of repeated cycles based on parameter data collected from a processing chamber; at operation 2404, generating a machine learning model based on the set of golden reference values; and, at operation 2406, using the machine learning model to match the parameter value in the second cycle in the series of repeated cycles with the corresponding parameter value in the first cycle in the series of repeated cycles.

[0126] In some examples, the semiconductor manufacturing process is an ALD process; and the steps of each cycle in a series of repeated cycles include sequential steps including dosing, sweeping, converting, and sweeping steps.

[0127] In some examples, the operations further include: generating a golden curve including parameter value data for each sequential step; and using the golden curve as training data for the machine learning model.

[0128] In some examples, the operations further include repeating and matching each loop in a series of repeated loops by matching parameter values ​​in each sequential step of the second loop with corresponding parameter values ​​in each sequential step of the first loop based on the machine learning model.

[0129] In some examples, the parameters include one or more of precursor manifold pressure, purge pressure, chamber pressure, gas flow rate, chamber temperature, RF reflected power, and RF forward power.

[0130] In some examples, the operations further include: collecting performance data from the processing chamber; identifying drift in the performance data and generating performance drift data; and incorporating the drift data into training data for the machine learning model.

[0131] refer to Fig.25 , a method 2500 for monitoring a processing cycle in a semiconductor manufacturing process includes: at operation 2502, identifying a parameter of the manufacturing process; and at operation 2504, generating a first curve of parameter values ​​including a first parameter value based on a first cycle of the manufacturing process, wherein the parameter value includes a gas pipeline fill time for a pipeline supplying a processing chamber.

[0132] In some examples, the operations further include calculating a gas line fill time based on a pressure ramp time between opening of a valve in a line supplying the process chamber and determining a constant pressure increase thereafter.

[0133] In some examples, the operations further include: identifying a second value of the parameter for a second cycle of the manufacturing process based on the curve generated for the first cycle; generating a second curve including a plurality of parameter values ​​derived from the second cycle of the manufacturing process; and curve fitting the second curve of parameter values ​​to the first curve of parameter values.

[0134] In some examples, the curve fitting operation includes fitting a series of curves based on parameter values ​​of or derived from third and subsequent cycles of the manufacturing process to the first or second curve to generate a golden curve defining a set of golden parameter values ​​for the manufacturing process.

[0135] In some examples, each cycle of the manufacturing process includes multiple steps in the ALD process; and wherein the operations further include: matching parameter values ​​in each step of the ALD process with parameter values ​​in the set of golden parameter values.

[0136] In some examples, the golden curve includes golden parameter values ​​for each step in the ALD process.

[0137] In some examples, the identified parameter is associated with a control variable of a manufacturing process; and the operation also includes: directly or indirectly using the first and second parameter values ​​to identify a match between a value of the control variable in the second loop and a value of the control variable in the first loop, and adjusting the manufacturing process to match the first parameter value to the second parameter value.

[0138] Fig.26 is a block diagram illustrating an example of a machine 2600 by which one or more of the exemplary process embodiments described herein may be controlled. In alternative embodiments, the machine 2600 may operate as a standalone device, or may be connected (e.g., networked) to other machines. In a networked deployment, the machine 2600 may operate in the capacity of a server machine, a client machine, or both in a server-client network environment. In one example, the machine 2600 may be used as a peer machine in a peer-to-peer (P2P) network (or other distributed network) environment. In addition, although only a single controller 2600 is shown, the term "machine" should also be construed to include any collection of machines (controllers) that individually or jointly execute a set (or multiple sets) of instructions to perform any one or more of the methods described herein, such as via cloud computing, software as a service (SaaS), or other computer cluster configuration. In some examples, and with reference to Fig.26 , the non-transitory machine-readable medium includes instructions 2624 that, when read by the machine 2600, cause the controller to control operations in the method including at least the non-limiting example operations summarized above and described herein.

[0139] The examples described herein may include logic, or multiple components or mechanisms, or may be operated by logic, multiple components or mechanisms. A circuit system is a collection of circuits implemented in a tangible entity including hardware (e.g., simple circuits, gates, logic, etc.). Circuit system components may be flexible over time and with basic hardware variability. A circuit system includes components that can perform specified operations individually or in combination when operating. In one example, the hardware of the circuit system can be designed in a fixed and immutable manner to perform specific operations (e.g., hard wiring). In one example, the hardware of the circuit system may include variable connection entity components (e.g., execution units, transistors, simple circuits, etc.), which include computer-readable media that are modified by physical means (e.g., magnetically, electrically, by a movable setting of a constant mass particle, etc.) to encode instructions for specific operations. When connecting entity components, the basic electrical properties of the hardware components are changed (e.g., from an insulator to a conductor, and vice versa). Instructions enable embedded hardware (e.g., an execution unit or a loading mechanism) to generate components of the circuit system in hardware via variable connections to perform parts of specific operations when operating. Thus, when the device is operating, the computer-readable medium is communicatively coupled to other components of the circuit system. In one example, any of the physical components can be used in more than one component in more than one circuit system. For example, in operation, an execution unit can be used in a first circuit of a first circuit system at one point in time, and reused by a second circuit of the first circuit system, or by a third circuit of the second circuit system at a different time.

[0140] The machine (e.g., computer system) 2600 may include a hardware processor 2602 (e.g., a central processing unit (CPU), a hardware processor core, or any combination thereof), a graphics processing unit (GPU) 2632, a main memory 2604, and a static memory 2606, some or all of which may communicate with each other via an interconnect (e.g., a bus) 2608. The machine 2600 may also include a display device 2610, an alphanumeric input device 2612 (e.g., a keyboard), and a user interface (UI) navigation device 2614 (e.g., a mouse). In one example, the display device 2610, the alphanumeric input device 2612, and the UI navigation device 2614 may be a touch screen display. The machine 2600 may additionally include a mass storage device (e.g., a drive unit) 2616, a signal generating device 2618 (e.g., a speaker), a network interface device 2620, and one or more sensors 2630, such as a global positioning system (GPS) sensor, a compass, an accelerometer, or another sensor. The machine 2600 may include an output controller 2628 (e.g., a serial (e.g., Universal Serial Bus (USB)), parallel, or other wired or wireless (e.g., infrared (IR), near field communication (NFC), etc.) connection) to communicate with or control one or more peripheral devices (e.g., a printer, a card reader, etc.).

[0141] The mass storage device 2616 may include a machine-readable medium 2622 on which one or more sets of data structures or instructions 2624 (e.g., software) may be stored that implement or be used by any one or more of the techniques or functions described herein. The instructions 2624 may also reside, as shown, completely or at least partially within the main memory 2604, within the static memory 2606, within the hardware processor 2602, or within the GPU 2632 during their execution by the machine 2600. In one example, one or any combination of the hardware processor 2602, the GPU 2632, the main memory 2604, the static memory 2606, or the mass storage device 2616 may constitute the machine-readable medium 2622.

[0142] Although the machine-readable medium 2622 is shown as a single medium, the term "machine-readable medium" may include a single medium or multiple media (eg, a centralized or distributed database and / or associated caches and servers) configured to store one or more instructions 2624.

[0143] The term "machine-readable medium" may include any medium capable of storing, encoding, or carrying instructions 2624 for execution by the machine 2600 and causing the machine 2600 to perform any one or more of the techniques of the present disclosure; or any medium capable of storing, encoding, or carrying data structures used by or related to such instructions 2624. Non-limiting examples of machine-readable media may include solid-state memory and optical and magnetic media. In one example, a large amount of machine-readable media includes a machine-readable medium 2622 having a plurality of particles having a constant mass (e.g., rest mass). Therefore, the large amount of machine-readable media is not a transient propagating signal. Specific examples of large amounts of machine-readable media may include non-volatile memory, such as semiconductor memory devices (e.g., electronically programmable read-only memory (EPROM), electronically erasable programmable read-only memory (EEPROM), and flash memory devices; magnetic disks, such as internal hard disks and removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks. The instructions 2624 may be further sent or received via the network interface device 2620 using a transmission medium through a communication network 2626.

[0144] Although examples have been described with reference to specific exemplary embodiments or methods, it is apparent that various modifications and changes may be made to these embodiments without departing from the broader scope of the embodiments. Therefore, the description and the drawings are to be regarded as illustrative rather than restrictive. The drawings constituting a part of this article show specific embodiments in an illustrative (but not limiting) manner in which the subject matter may be practiced. The illustrated embodiments are described in sufficient detail to enable those skilled in the art to practice the teachings disclosed herein. Other embodiments may be used and derived therefrom so that structural and logical substitutions and changes may be made without departing from the scope of the present disclosure. Therefore, this specific embodiment is not to be regarded as restrictive, and the scope of the various embodiments is limited only by the appended claims, together with the full scope of the equivalent schemes granted by these claims.

[0145] These embodiments of the subject matter of the present invention may be referred to herein individually and / or collectively by the term "invention", which is merely for convenience and is not intended to voluntarily limit the scope of the present application to any single invention or inventive concept (if more than one invention or inventive concept is in fact disclosed). Therefore, although specific embodiments are shown and described herein, it should be understood that any configuration calculated to achieve the same purpose can replace the specific embodiments shown. The present disclosure is intended to cover any and all adjustments or variations of the various embodiments. After reading the above description, the combination of the above embodiments and other embodiments not specifically described herein will be apparent to those skilled in the art.

Claims

1. A system for monitoring a process cycle in a manufacturing process, the system comprising: a processing chamber for the manufacturing process; as well as One or more controllers configured to perform process monitoring operations comprising: identifying parameters of the manufacturing process; generating a first curve of parameter values ​​including first parameter values ​​based on a first cycle of the manufacturing process, the parameter values ​​including a gas line fill time for a line supplying the process chamber; calculating the gas line fill time based on a pressure ramp time between opening of a valve in a line supplying the process chamber and determination of a constant pressure increase thereafter; and identifying a second value of the parameter for a second cycle of the manufacturing process based on the generated curve for the first cycle; generating a second curve comprising a plurality of parameter values ​​derived from the second cycle of the manufacturing process; and Performing curve fitting on the second curve of the parameter value and the first curve of the parameter value, wherein the curve fitting operation comprises: A series of curves based on parameter values ​​of or derived from third and subsequent cycles of the manufacturing process are fit to the first or second curve to generate a golden curve defining a set of golden parameter values ​​for the manufacturing process.

2. The system according to claim 1, wherein: Each cycle of the manufacturing process includes multiple steps in the ALD process; as well as The operations also include: Parameter values ​​in each step of the ALD process are matched with parameter values ​​in the set of golden parameter values.

3. The system according to claim 1, wherein: The golden curve includes golden parameter values ​​for each step in the ALD process.

4. The system according to claim 2, wherein: The identified parameters are associated with control variables of the manufacturing process; as well as The operations also include: directly or indirectly using the first and second parameter values ​​to identify a match between a value of a control variable in the second loop and a value of a control variable in the first loop, and The manufacturing process is adjusted to match the first parameter value to the second parameter value.

5. A method for monitoring a process cycle in a manufacturing process, the method comprising: identifying parameters of the manufacturing process; generating a first curve of parameter values ​​including first parameter values ​​based on a first cycle of the manufacturing process, the parameter values ​​including a gas line fill time of a line supplying a process chamber; calculating the gas line fill time based on a pressure ramp time between opening of a valve in a line supplying the process chamber and determination of a constant pressure increase thereafter; as well as identifying a second value of the parameter for a second cycle of the manufacturing process based on the generated curve for the first cycle; generating a second curve comprising a plurality of parameter values ​​derived from the second cycle of the manufacturing process; as well as Performing curve fitting on the second curve of the parameter value and the first curve of the parameter value, wherein the curve fitting operation comprises: A series of curves based on parameter values ​​of or derived from third and subsequent cycles of the manufacturing process are fit to the first or second curve to generate a golden curve defining a set of golden parameter values ​​for the manufacturing process.

6. The method according to claim 5, wherein: Each cycle of the manufacturing process includes multiple steps in the ALD process; as well as The operations also include: Parameter values ​​in each step of the ALD process are matched with parameter values ​​in the set of golden parameter values.

7. The method according to claim 5, wherein: The golden curve includes golden parameter values ​​for each step in the ALD process.

8. The method according to claim 6, wherein: The identified parameters are associated with control variables of the manufacturing process; as well as The operations also include: directly or indirectly using the first and second parameter values ​​to identify a match between a value of a control variable in the second loop and a value of a control variable in the first loop, and The manufacturing process is adjusted to match the first parameter value to the second parameter value.

9. A machine-readable medium comprising instructions, which when read by a machine cause the machine to perform operations comprising: Identify parameters of the manufacturing process; generating a first curve of parameter values ​​including first parameter values ​​based on a first cycle of the manufacturing process, the parameter values ​​including a gas line fill time of a line supplying a process chamber; calculating the gas line fill time based on a pressure ramp time between opening of a valve in a line supplying the process chamber and determination of a constant pressure increase thereafter; as well as identifying a second value of the parameter for a second cycle of the manufacturing process based on the generated curve for the first cycle; generating a second curve comprising a plurality of parameter values ​​derived from the second cycle of the manufacturing process; as well as Performing curve fitting on the second curve of the parameter value and the first curve of the parameter value, wherein the curve fitting operation comprises: A series of curves based on parameter values ​​of or derived from third and subsequent cycles of the manufacturing process are fit to the first or second curve to generate a golden curve defining a set of golden parameter values ​​for the manufacturing process.

10. The medium according to claim 9, wherein Each cycle of the manufacturing process includes multiple steps in the ALD process; as well as The operations also include: Parameter values ​​in each step of the ALD process are matched with parameter values ​​in the set of golden parameter values.

11. The medium according to claim 9, wherein The golden curve includes golden parameter values ​​for each step in the ALD process.

12. The medium according to claim 9, wherein: The identified parameters are associated with control variables of the manufacturing process; as well as The operations also include: directly or indirectly using the first and second parameter values ​​to identify a match between a value of a control variable in the second loop and a value of a control variable in the first loop, and The manufacturing process is adjusted to match the first parameter value to the second parameter value.

Images