Method for autotuning chamber control and process performance evaluation - Patent Application 20070122997

Abstract

Embodiments disclosed herein include a method for auto-tuning a system. In one embodiment, the method includes determining if the system is in a steady state. Thereafter, the method includes waking up the system. In one embodiment, the method includes storing process feedback measurements from the woken up system to provide a set of stored data. In one embodiment, the set of stored data is a subset of all available data generated by the woken up system. In one embodiment, the method further includes determining when the woken up system returns to a steady state and tuning the system using the set of stored data.

Description

[Technical Field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority from U.S. Patent Application No. 17 / 467,020, filed September 3, 2021, the entire contents of which are incorporated herein by reference.

[0002] Embodiments of the present disclosure relate to the field of semiconductor processing, and more particularly to autotuning of chamber controls and process performance evaluation in a semiconductor processing environment.

[0003] 2. Description of Related Art In the semiconductor industry, processing tools have grown into highly complex systems. In an effort to improve process uniformity, processing tools are equipped with numerous sensors to measure various process parameters. The sensors may be part of a control architecture. The control architecture is used to adjust actuators, thereby changing the process parameters. For example, some process parameters in semiconductor tools include, but are not limited to, temperature, pressure, RF frequency, and power. During the process of setting up such semiconductor tools, system control parameters (e.g., gain values) must be set to achieve accurate and precise control of the semiconductor tool. Currently, tuning tasks require engineers with specific skills to accomplish the task, which may be difficult to find, further extending chamber setup times.

[0004] When a chamber is operated over an extended period of time, changes in the behavior and / or performance of some hardware components are typically observed. Therefore, processes used to correct such drift include retuning the various closed-loop controllers. However, this task requires significant engineering time and results in commensurately long chamber downtime. Summary of the Invention

[0005] Embodiments disclosed herein include a method for autotuning a system. In one embodiment, the method includes determining whether the system is in a steady state. Thereafter, the method includes activating the system. In one embodiment, the method includes saving process feedback measurements from the activated system to provide a set of saved data. In one embodiment, the set of saved data is a subset of all available data generated by the activated system. In one embodiment, the method further includes determining when the activated system returns to a steady state and tuning the system using the set of saved data.

[0006] An additional embodiment includes a method for monitoring recipe performance in a semiconductor processing tool. In one embodiment, the method includes identifying a noise level in the system. After the noise level is identified, the embodiment includes storing process feedback measurements from the system to provide a set of stored data. In one embodiment, the set of stored data is a subset of all available data generated by the system. In one embodiment, the method further includes comparing the set of stored data to a set of reference data.

[0007] Embodiments may further include a semiconductor processing tool. In one embodiment, the tool comprises a chamber and a plurality of components interfacing with the chamber. In one embodiment, each of the components is controlled by a different closed-loop control system. In one embodiment, the processing tool comprises an autotuning module for tuning the closed-loop control systems. In one embodiment, the autotuning module comprises a noise estimation module that identifies noise present in each closed-loop control system. In one embodiment, the autotuning module further comprises a data collection module that stores process feedback measurements from the closed-loop control systems to provide a set of stored data for each closed-loop control system. In one embodiment, the set of stored data is a subset of all available data generated by the closed-loop control systems. [Brief explanation of the drawings]

[0008] [Figure 1A] FIG. 1 is a schematic diagram of a single loop control architecture in a semiconductor processing tool, according to an embodiment. [Figure 1B] FIG. 1B is a schematic diagram of a semiconductor processing chamber having temperature control components controlled by the control architecture of FIG. 1A, according to an embodiment. [Figure 2A] FIG. 1 is a schematic diagram of a control architecture for a semiconductor processing tool having two control loops, according to an embodiment. [Figure 2B] 2B is a schematic diagram of a processing chamber having temperature and pressure control components controlled by the control architecture of FIG. 2A, according to an embodiment. [Figure 3] FIG. 1 is a schematic diagram of a control architecture for a semiconductor processing tool having multiple control loops, according to an embodiment. [Figure 4A] 1 is a graph of a response variable over time, according to an embodiment. [Figure 4B] 1 is a graph of a response variable over time compared to known values ​​of the response variable over time, according to an embodiment. [Figure 5]FIG. 1 is a process flow diagram of a method for data collection for improved closed-loop control in a semiconductor manufacturing tool, according to an embodiment. [Figure 6] 1 is a graph illustrating geometric reduction of a response variable, according to an embodiment. [Figure 7] FIG. 1 is a process flow diagram of a method for autotuning a semiconductor tool, according to an embodiment. [Figure 8] FIG. 1 is a process flow diagram of a method for monitoring a recipe on a semiconductor tool, according to an embodiment. [Figure 9] FIG. 1 illustrates a block diagram of an exemplary computer system according to one embodiment of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0009] Autotuning of chamber controls and process performance evaluation in a semiconductor processing environment are described herein. In the following description, numerous specific details are presented to provide a thorough understanding of embodiments of the present disclosure. It will be apparent to those skilled in the art that embodiments of the present disclosure may be practiced without these specific details. In other instances, well-known aspects have not been described in detail so as not to unnecessarily obscure embodiments of the present disclosure. Furthermore, it should be understood that the various embodiments shown in the figures are illustrative representations and are not necessarily drawn to scale.

[0010] Controlling temperature and pressure in semiconductor manufacturing is crucial to achieving expected processing results. Utilizing multi-station, multi-chamber processing platforms increases processing throughput, allowing multiple types of processes to be performed sequentially on a substrate. To achieve this, these complex platforms must incorporate hundreds of sensors to monitor temperature, pressure, and flow signals throughout the platform, as well as power usage and motion of moving components. To operate these many sensors, the platform must incorporate a means to automatically control these signals to drive response variables (e.g., temperature and pressure) toward desired targets. As mentioned above, controllers must be initialized and continuously monitored to account for platform drift.

[0011] Examples of control systems in semiconductor platforms include a temperature-controlled lid, a multi-zone temperature-controlled gas distribution system, a temperature-controlled faceplate, a substrate backpressure controller, and a pressure-controlled ampule for gas supply to the process chamber. The number of such controllers within a single processing chamber can easily reach 50 or more sensors. For example, in one scenario, a single chamber may contain one lid heater, four faceplate heaters, 30 gas heater zones, four substrate backpressure controllers, 10 pressure-controlled ampules and reservoirs, and possibly multiple motorized assemblies for moving the susceptor within the processing chamber.

[0012] 1A, a schematic diagram of the structure of a control architecture 108 is shown, according to one embodiment. In one embodiment, the overall control architecture 108 includes actuators 103, input / output (I / O) devices 102, and a computing server 101. The actuators 103 provide control signals 106 to the processing environment. In one embodiment, the actuators 103 may be, but are not limited to, valves, amplifiers, silicon-controlled rectifiers (SCRs), drivers, pumps, or mass flow controllers, and are intended to interface directly with the processing hardware being controlled. In one embodiment, the I / O devices 102 receive process feedback 107. The process feedback 107 may be generated by sensors monitoring process parameters (e.g., temperature, pressure, etc.). For example, the process feedback 107 may include, but is not limited to, temperature, pressure, flow transducer measurements, or level transducer measurements.

[0013] In one embodiment, the I / O device 102 provides processing feedback 107 to the computing server 101. The computing server 101 identifies a desired response and communicates the response to the I / O device 102, which then relays the desired response 105 to the actuator 103. Regardless of the application, the computing server 101 is where the control paradigm is stored and executed. The computing server 101 communicates with the I / O device 102 via bus 104. That is, the I / O device 102 updates the computing server 101 with the latest processing feedback 107, and the computing server 101 communicates the corresponding desired response to be taken to the I / O device 102.

[0014] 1B, a schematic diagram of a processing tool 120 according to an embodiment is shown. In one embodiment, the processing tool 120 includes a chamber 121. A heated faceplate 125 has holes through which an inlet gas enters the chamber 121. A portion of the faceplate's thermal energy is transferred to a gas, which is then directed downward toward a substrate 122 placed on a substrate support 123. Due to the faceplate's proximity to the substrate 122, temperature control of the faceplate is critical to achieving desired processing results. Therefore, a temperature transducer 124 can be used to accurately measure the faceplate temperature. The measurements 107 are used by a control strategy executing on the computing server 101 of the control architecture 108, the objective of which is to maintain the faceplate temperature at a target value.

[0015] 1B, actuator 103 may be an SCR connected to an external power source and driven by a digital signal 105 provided by computing server 101 via I / O device 102. SCR actuator 103 generates a control signal 106 that controls the temperature of faceplate 124. Thus, computing server 101 determines how to manipulate digital signal 105 so that temperature 107 of faceplate 125 reaches and remains at a desired value.

[0016] Referring now to FIG. 2A, a schematic diagram of the structure of control architecture 208 is shown, according to one embodiment. In one embodiment, computing server 201 is responsible for controlling two different process feedbacks 207A and 207B using corresponding control outputs 206A and 206B. In this case, the same I / O device 202 may be used to interface computing server 201 with the rest of the control architecture (via bus 204). Computing server 201 identifies a desired response and communicates that response to I / O device 202, which then relays desired responses 205A and 205B to actuators 203A and 203B, respectively. In one embodiment, two different actuators 203A and 203B may be used to control process feedbacks 207A and 207B, respectively. Control architecture 208 of FIG. 2A is said to include two closed-loop controls. The first closed loop control is for controlling process feedback 207A, and the second closed loop control is for controlling process feedback 207B.

[0017] 2B, a schematic diagram of a processing tool 220 according to an embodiment is shown. In one embodiment, processing tool 220 is an example of a two-loop control architecture. In addition to controlling the temperature of faceplate 225, control architecture 208 also controls the backside pressure of the substrate. In this example, pressure measurement 207B is obtained using pressure gauge 227, which measures the pressure in gas line 226. Server 201 receives real-time notification of pressure feedback 207B and, in response, commands mass flow control actuator 203, which drives pump 228 in response to control output 206B.

[0018] 1B, substrate 222 rests on substrate support 223 below faceplate 225. Temperature transducer 224 provides temperature feedback 207A to control architecture 208. Server 201 receives real-time notification of temperature feedback 207A and, in response, commands temperature control actuator 203 (e.g., SCR), which issues control signal 206A that controls the temperature of faceplate 225.

[0019] In the illustrated embodiment, a pair of closed loop controls is shown. However, it should be understood that in an actual processing tool, the overall control architecture of tool 220 may include multiple closed loop controls (e.g., 40 or more closed loop controls). In one embodiment, each loop is designed to interface with a different transducer and actuator.

[0020] Referring now to FIG. 3, a schematic diagram of the structure of a control architecture 308 with multiple closed-loop controls is shown, according to an embodiment. In FIG. 3, feedback signal bus 307A-n is a bus including feedback lines from all transducers. In one embodiment, the number of actuators 303A-n is equal to the number of feedback signals 307. I / O device 302 may be communicatively coupled to each of actuators 303A-n via digital signals 305a-n. Actuators 303a-n may be coupled to the system via control signals 306a-n. As shown, a single computing server 301 may interface with each of actuators 303A-n via bus 304 coupled to I / O device 302.

[0021] It should be appreciated that as the value of n increases, so does the engineering time required to tune all of the controllers in the control architecture 308 and the setup time required to get the chamber ready to run a recipe. Additionally, the engineers responsible for tuning tasks require specific skills to accomplish the task, and such engineers may be hard to come by, further extending the chamber setup time. Such tuning is also required before the tool can be used for the first time.

[0022] When a chamber is operated over an extended period of time, changes in the behavior and / or performance of some hardware components are usually observed. To correct for such drifts, it may be sufficient to occasionally retune the various closed-loop controllers. However, this task requires a significant amount of engineering time and, proportionally, an equal amount of chamber downtime. As a precaution, or perhaps as a best practice, it may be advisable to periodically retune some of the closed-loop controllers. Such measures could be taken, for example, during an already scheduled chamber downtime. However, the significant resources required to accomplish this task may be the reason why it is rarely performed.

[0023] Accordingly, embodiments disclosed herein include a method for automatically tuning all loops within the control architecture 308, thereby reducing the required engineering time to negligible levels. Furthermore, embodiments significantly reduce the critical skill set required by a lab operator using the autotuning methods disclosed herein. Furthermore, a lab operator can schedule tuning operations to run automatically during times when the chamber is not normally operating (e.g., overnight). Furthermore, the availability of such autotuning methods simplifies the scheduling of routine maintenance, and therefore any maintenance caused by known system failures or drift.

[0024] To apply the above advantages to semiconductor manufacturing tools, the autotuning method may include the ability to use the method reliably and accurately with minimal assistance in a real-time software environment. The processes being monitored may have unknown bandwidths, and potentially the bandwidth of the processes may be of various magnitudes. Furthermore, the autotuning should have a minimal footprint in terms of the amount of server memory required. Thus, the method can be run simultaneously for multiple processes in each chamber and can run in the background while other work is being done around the tool.

[0025] It will be appreciated that autotuning methods characterized as described above are not currently used within semiconductor fabrication facilities. One reason is that third-party autotuning software is typically not permitted within the proprietary software infrastructure common in the semiconductor industry. Furthermore, similar methods impose less stringent requirements. For example, such methods are implemented on dedicated computers that run only the autotuning routine (and not other resource-intensive processes). Furthermore, existing processes are designed to be used with one closed-loop control at a time. Therefore, such methods are not compatible with semiconductor fabrication facilities and are not practical as a way to minimize chamber downtime.

[0026] In addition to autotuning capabilities, embodiments disclosed herein can also provide the ability to monitor the performance of n closed-loop controllers throughout a process recipe. In complex systems where n is a large value, it is often important to monitor the performance of each loop throughout the process recipe. This allows for timely reporting of any performance drift and identification of the loop in which the drift occurred. Identifying the drifting loop can speed the process of finding the root cause of the drift and minimize potential production waste.

[0027] General performance monitoring allows for monitoring whether process feedback (e.g., feedback 107, discussed above) exceeds some preset limit, changes faster or slower than expected, or deviates significantly from a desired value. Similarly, general performance monitoring notifies a user if controller operation 106 exceeds a predetermined limit. However, as shown in FIGS. 4A and 4B, such general performance monitoring may not identify whether the process feedback behaved as expected. For example, process feedback 431 in FIG. 4A may represent the expected process behavior during a particular process recipe. As shown in FIG. 4A, process feedback 431 falls within drift limits 432 and 433, and as a result, the process behavior is deemed acceptable. However, in FIG. 4B, process feedback 434 observed during a later run of the same process recipe appears to be shifted in time relative to the behavior 431 observed in the earlier run. Because behavior 434 falls within limits 432 and 433, general performance monitoring would deem process behavior 434 acceptable. However, given the expectations set by behavior 431, behavior 434 is less than expected.

[0028] Recipe performance monitoring described in accordance with embodiments disclosed herein can solve the above-mentioned problems. That is, recipe performance monitoring provides users with real-time information on whether each closed-loop controller is performing as expected throughout the recipe. For such performance monitoring to function within a semiconductor processing server, similar requirements as those described for the autotuning method are necessary. Recipe performance monitoring must be executable in a real-time server environment, must minimize server memory usage to enable simultaneous monitoring of multiple loops, and must run in the background while the closed-loop controllers are running. Minimizing server memory usage is critical and not easy. In fact, process recipes can span minutes to hours. Furthermore, monitored signals are refreshed very quickly. That is, the time between two consecutive updates of the same signal can be very short. Therefore, a large number of data samples per unit time are available for analysis. Therefore, the ability to optimize data in real time is crucial for recipe performance monitoring.

[0029] Some instances of recipe performance monitoring can be found in non-real-time environments. When implemented in an offline capacity, recipe monitoring is not bound by memory usage limitations. Instead, such monitoring typically operates offline on a dedicated server. The benefit of performance monitoring in a real-time environment is the ability to receive immediate notifications in response to unexpected performance in semiconductor processing tools, thereby minimizing production waste.

[0030] The embodiments disclosed herein include an autotuning method intended to simultaneously and automatically tune a large number of closed-loop controllers in a chamber, and a recipe performance monitoring method intended to monitor the performance of such controllers throughout a process recipe. These two applications are based on similar underlying methods for data mining. The data mining process has low memory usage and low execution costs, making these applications suitable for real-time execution.

[0031] In offline data analysis, the entire data set is available before the analysis is performed. Such data is available as a time series, often with data points evenly spaced in time. In contrast, in real-time data analysis, the entire data time series is not available at the time the analysis is performed; only past and present data are available. Furthermore, in real-time data analysis, it is often not possible to collect and store data at a specific time rate. This is because if the time series is too long, the system may exhaust available memory resources before the processing recipe is completed. Therefore, online data analysis is inherently challenging for the following reasons: 1) Online data analysis requires more efficient ways to determine which input data points are worth storing for future analysis; and 2) Analysis performed on past and present data may lead to erroneous results because the complete picture is not available at the time the analysis is performed.

[0032] Referring now to Figure 5, a process flow diagram of a process 540 for data mining, according to an embodiment, is shown. As shown in Figure 5, the method developed for input data point selection includes a noise estimation operation 541 followed by a data collection operation 542. During the noise estimation operation 541, the noise in the signal is estimated to determine which variations in the process feedback correspond to actual variations in the process response and which variations are instead due to measurement errors. Noise can be due to interference from external power / magnetic sources or transducer inaccuracies and can be characterized as high frequency variations, with frequency components in a range far beyond the bandwidth of the process.

[0033] In one embodiment, the noise estimation operation 541 does not need to be performed every time data needs to be collected. For example, the noise estimation operation can be performed before each data collection, once a day, once a month, once a year, or at any other predetermined period. The result of the noise estimation operation 541 is a single value that is utilized in subsequent data collection operations or a value that can be retained for future reference.

[0034] During the data collection operation 542, consecutive data are grouped into rectangular boxes 650, as shown in FIG. 6 . In one embodiment, the height D of the rectangular boxes 650 is a function of a noise estimate. In one embodiment, the rectangular boxes 650 depend on the variability of the process feedback. That is, if consecutive data points fall within a given box, the relative variation between such data points is discarded. In one embodiment, new boxes are created as a function of the noise level, the amount of variation in the process feedback, and the relative slope of such boxes. Only one data point should be stored for each box, and the remaining data points can be discarded until a new box is created. In this manner, the amount of data points collected is significantly reduced by a factor inversely proportional to the sampling rate of the raw process feedback measurements. Furthermore, the value of the process feedback at any point in time can be easily retrieved from the collected data by simple interpolation from stored adjacent data points.

[0035] It should be appreciated that the rectangular box 650 approach to data collection means that the data stored is not affected by the data sampling rate. For example, if data is acquired at a microsecond rate, a nanosecond rate, a second rate, or any rate, the total data stored includes a single value within the rectangular box 650 and the shape / slope of the rectangular box. In other words, the data recorded is a small portion of all available data. The small amount of data required allows embodiments to perform data collection and analysis in real time, without the need to run offline on a dedicated analysis server.

[0036] Referring now to FIG. 7 , a process flow diagram of a process 760 for autotuning a closed-loop controller is shown, according to an embodiment. In one embodiment, process 760 may begin with a user command to initiate process 760. In other embodiments, process 760 may begin automatically. For example, process 760 may begin after a preset period of time or after a preset number of recipe cycles have been executed. In one embodiment, the autotuner's first operation 761 is performed twice. The first operation 761 includes identifying noise in the process feedback measurements in operation 763 and evaluating whether the process feedback is at a steady state in operation 762. Operation 762 is critical for improving the accuracy of future tuning operations 767. The process feedback is assumed to be at a steady state based on an analysis of the orientation of a rectangle generated from the collected data relative to the time axis.

[0037] Once the noise has been measured and the process feedback is assessed to be at steady state, process 760 proceeds to operation 764. Operation 764 includes initiating the system. In one embodiment, an activation signal is sent to the system. This activation signal may manifest a key characteristic of the process to be controlled. The activation signal is a change in an input to the system. For example, in the case of a closed-loop controller controlling temperature, the activation signal may be an increase in power to a heater.

[0038] After initiation operation 764, processing proceeds to operation 765. Operation 765 includes data collection. Data collection operation 765 may be substantially similar to data collection operation 542 described above in connection with FIG. 5. That is, data collection operation 542 may include generating a number of rectangular boxes that substantially match the shape of the processing feedback graph. For example, a single data point within the box and the shape / slope of the box may be the only data stored. In this manner, the stored data is a subset of the generated data. The small amount of data allows for real-time processing of the data.

[0039] In one embodiment, data collection operation 765 continues until the process is deemed to be in a steady state, as indicated by operation 766. As described above, the process is deemed to be in a steady state based on an analysis of the orientation of a rectangle generated from the collected data relative to the time axis. Once the process reaches a steady state, tuning operation 767 analyzes the data collected thus far. In one embodiment, tuning operation 767 is performed with the objective of determining all of the gain values ​​for the closed-loop controller.

[0040] After tuning operation 767, process 760 proceeds to operation 768, which includes validating the tuning. The validation operation is performed by comparing collected data with an internal simulation model derived from such data. Such validation is useful for determining whether tuning results are reasonable based on previously collected data. In one embodiment, the validation operation may further include, if desired, a comparison with a physics-based model. A physics-based model is a virtual representation of the control loop that uses physics-based equations to model system behavior.

[0041] Referring now to FIG. 8 , a flow diagram of a process 870 is shown, according to an embodiment. Process 870 may include a process for monitoring a recipe as it is executed. In one embodiment, process 870 may begin with a start command that signals the beginning of a process recipe. The start command may be generated automatically at the start of a new process recipe. For example, all instances of a process recipe may be monitored by process 870. In other embodiments, recipe monitoring process 870 may monitor a subset of all iterations of a process recipe. In one embodiment, with appropriate timing of start and stop commands, the same recipe performance analysis process 870 may be applied to portions of a recipe, as desired. That is, in some embodiments, the entire duration of a process recipe may not be monitored.

[0042] In one embodiment, process 870 begins with operation 871. Operation 871 includes estimating noise. The noise estimation operation 871 may be substantially similar to the noise estimation operation 541 of FIG. 5. After the noise estimation operation 871, the process may proceed to operation 872. Operation 872 includes starting data collection 873 and stopping data collection 874. Data collection 872 may be substantially similar to data collection 542 of FIG. 5. That is, data collection process 542 may include generating multiple rectangular boxes that substantially match the shape of the process feedback graph. For example, a single data point within the box and the shape / slope of the box may be the only data stored. In this manner, the stored data is a subset of the generated data. The small amount of data allows for real-time processing of the data.

[0043] In some embodiments, the saved data is considered a "standard" data set, i.e., future analysis will be based on a comparison of the "standard" recorded performance with the currently recorded data. The "standard" data set may be repopulated after a preset number of iterations of the recipe, after a period of time (e.g., days, weeks, a year, etc.), or after maintenance of the processing tool.

[0044] In one embodiment, process 870 may proceed to operation 875. Operation 875 includes comparing the recorded data to a known reference. For example, the known reference may be a “standard” data set. In other embodiments, the known reference may be a data set from a previous iteration of the recipe. In one embodiment, process 870 then proceeds to operation 876. Operation 876 includes reporting the recipe performance status. The report may include a warning that the recipe is deviating from the “standard” data set and / or may include other statistical information about the process as output. In one embodiment, the deviation from the “standard” data set may include a change in the shape, size, and orientation of the rectangular box. Alternatively, the shape, size, and orientation of the box may be similar to the “standard” data set (e.g., similar to the embodiment shown in FIG. 4B ), except that there is a time delay in the new data.

[0045] 9 illustrates a schematic diagram of an exemplary form of a machine, computer system 900, within which a set of instructions may be executed to cause the machine to perform any one or more of the methods described herein. In alternative embodiments, the machine may be connected (e.g., networked) to other machines in a local area network (LAN), an intranet, an extranet, or the Internet. The machine may operate in the capacity of a server or a client machine in a client / server network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. The machine may be a personal computer (PC), a tablet PC, a set-top box (STB), a personal digital assistant (PDA), a mobile phone, a web appliance, a server, a network router, switch, or bridge, or any machine capable of executing a set of instructions (sequentially or otherwise) that specify operations to be performed by the machine. Furthermore, although only a single machine is shown, the term "machine" may be interpreted to include any collection of machines (e.g., computers) that individually or together execute a set (or sets) of instructions to perform any one or more of the methodologies described herein.

[0046] The exemplary computer system 900 includes a processor 902, a main memory 904 (e.g., read-only memory (ROM), flash memory, dynamic random access memory (DRAM) (such as synchronous DRAM (SDRAM) or Rambus DRAM (RDRAM))), a static memory 906 (e.g., flash memory, static random access memory (SRAM), MRAM, etc.), and a secondary memory 918 (e.g., a data storage device), which communicate with each other via a bus 930.

[0047] Processor 902 represents one or more general-purpose processing devices, such as a microprocessor, a central processing unit, or the like. More specifically, processor 902 may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor implementing other instruction sets, or a processor implementing a combination of instruction sets. Processor 902 may also be one or more special-purpose processing devices, such as an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), a network processor, or the like. Processor 902 is configured to execute processing logic 926 to perform the operations described herein.

[0048] Computer system 900 may further include a network interface device 908. Computer system 900 may include a video display unit 910 (e.g., a liquid crystal display (LCD), a light emitting diode display (LED), or a cathode ray tube (CRT)), an alphanumeric input device 912 (e.g., a keyboard), a cursor control device 914 (e.g., a mouse), and a signal generation device 916 (e.g., a speaker).

[0049] The secondary memory 918 may include a machine-accessible storage medium (or more specifically, a computer-readable storage medium) 932 having stored thereon one or more sets of instructions (e.g., software 922) that embody one or more of the methods or functions described herein. This software 922 may also reside, completely or at least partially, within the main memory 904 and / or within the processor 902 during execution by the computer system 900. The main memory 904 and the processor 902 further constitute machine-readable storage media. This software 922 may further be transmitted or received over the network 920 via the network interface device 908.

[0050] While the illustrated embodiment depicts the machine-accessible storage medium 932 as a single medium, the term "machine-readable storage medium" should be understood to include a single medium or multiple media (e.g., a centralized or distributed database, and / or associated caches and servers) that store one or more sets of instructions. Furthermore, the term "machine-readable storage medium" should be taken to include any medium capable of storing or encoding a set of instructions that are executed by a machine, causing the machine to perform any one or more of the methodologies of the present disclosure. Accordingly, the term "machine-readable storage medium" should be taken to include, but is not limited to, solid-state memory, optical media, and magnetic media.

[0051] According to one embodiment of the present disclosure, a machine-accessible storage medium has stored thereon instructions that cause a data processing system to perform a method for monitoring a process recipe using a data mining algorithm executed in real time.

[0052] Thus, a method for monitoring process recipes using data mining algorithms executed in real time is disclosed.

Claims

1. 1. A method for autotuning a system, comprising: determining whether the system is in a steady state; Varying an input to the system; storing process feedback measurements from the system whose input has changed to provide a set of stored data, the set of stored data being a subset of all available data generated by the system whose input has changed; determining when the system returns to the steady state following the change in the input; tuning the system using the set of stored data; and Including, storing the processed feedback measurements includes grouping consecutive data points into a series of rectangular boxes, and only the first data point within a given rectangular box is stored as part of the set of stored data. method.

2. The method of claim 1 further comprising identifying a noise level in a process feedback measurement before changing an input to the system.

3. A method for autotuning a system, comprising: repeating the method of claim 2, The method wherein the noise level is determined for each of the iterations.

4. A method for autotuning a system, comprising: repeating the method of claim 2, The method, wherein the noise level is determined before starting the first of the iterations and after performing a given number of the iterations or after a given period of time has elapsed after starting the first of the iterations.

5. The method of claim 1 , wherein the height of the box is a function of noise in the system.

6. The method of claim 1 , wherein the input-changed system is determined to have returned to the steady state if the rectangular box has a slope that is substantially parallel to the time axis.

7. The method of claim 1 , wherein tuning the system comprises setting a gain value of a closed-loop controller of the system.

8. 8. The method of claim 7, further comprising validating the tuning by comparing collected data to an internal simulation model, the internal simulation model being a model derived from information extracted from the stored data or from a first principles analysis of the system.

9. The method of claim 1 , wherein the system is a closed-loop control system in a semiconductor manufacturing tool.

10. The method of claim 9 , wherein the closed loop control is control of a thermal system or a pressure system.

11. The method of claim 1 , wherein the system is a multiple closed-loop control system in a semiconductor manufacturing tool.

12. 1. A method for monitoring performance of a recipe in a semiconductor processing tool, comprising: Identifying the noise level in the system; storing processed feedback measurements from the system to provide a set of stored data, the set of stored data being a subset of all available data generated by the system; comparing said set of stored data with a set of reference data; Including, storing the processed feedback measurements includes grouping consecutive data points into a series of rectangular boxes, and only a single data point within a given rectangular box is stored as part of the set of stored data. method.

13. The method of claim 12 , wherein the set of reference data is known data when the recipe has worked well.

14. The method of claim 12 , further comprising reporting a recipe performance status if the set of stored data differs from the set of reference data.

15. 15. The method of claim 14, wherein the difference between the set of stored data and the set of reference data is a time delay shift of recorded data points.

16. 1. A semiconductor processing tool comprising: Chamber, a plurality of components interfacing with the chamber, each controlled by a different closed-loop control system; and an autotuning module for tuning the closed-loop control system; The autotuning module is a noise estimation module that identifies noise present in each closed-loop control system; a data collection module that stores processed feedback measurements from the closed-loop control systems to provide a set of stored data for each closed-loop control system, the set of stored data being a subset of all available data generated by the closed-loop control systems; Equipped with storing the processed feedback measurements includes grouping consecutive data points into a series of rectangular boxes, and only a single data point within a given rectangular box is stored as part of the set of stored data. Semiconductor processing tools.

17. 17. The semiconductor processing tool of claim 16, wherein the plurality of components interfacing with the chamber comprises a temperature-controlled component and a pressure-controlled component.

Images