Real-time health monitoring of semiconductor manufacturing equipment

Abstract

A system and method for real-time semiconductor manufacturing cluster tool health status monitoring is provided through in-situ sensors. In one method embodiment, an operating procedure for evacuating / ventilating a load lock (LL) and the LL door facing a vacuum transfer module (VTM) and an equipment front end module (EFEM), a sensor installation location and operating procedure, and a data flow and analysis procedure are proposed. The sensor provides real-time data and simultaneously monitors airborne particle contamination of the EFEM, load lock (LL), VTM, and multiple processing modules (PM) by correlating the door open / close time and ventilation / evacuation time in the load lock to particle measurement data. The method also includes providing an operation based on real-time measurement data, door status data, and the use of a machine learning algorithm to determine a recommended maintenance procedure for one of the EFEM, LL, VTM, or multiple PMs.

Description

Technical Field

[0001] The present disclosure relates to systems and methods for semiconductor manufacturing, and more particularly to monitoring the health of semiconductor manufacturing equipment for early detection of tool failures. Background Art

[0002] In the manufacture of semiconductor devices, a cluster tool having multiple process modules (PMs), workstations, and vacuum transfer modules (VTMs) is used to transfer substrates between PMs. Cluster tools are typically operated under vacuum conditions to match the pressures required for the various PMs. Substrates are typically introduced from the atmosphere into the cluster tool via an equipment front end module (EFEM) (e.g., an equipment front end module for a clean room), which may include an atmospheric transfer module (ATM). The EFEM and VTM of a cluster tool are hermetically separated by a load lock, which maintains the pressure of the EFEM or VTM during the transfer of substrates into and out of the cluster tool. The load lock is cyclically evacuated to vacuum before docking with the VTM, and vented toward the atmosphere (e.g., with nitrogen) before docking with the EFEM.

[0003] Currently, it is difficult to monitor tool health in an automated manner. First, in order to effectively utilize the space in the EFEM, VTM, and PM, there is almost no additional space in the cluster tool to place monitoring equipment. In addition, even if there is space, the monitoring equipment may not be able to operate under the vacuum conditions of the cluster tool. Therefore, there is a need in the industry to provide real-time and automated tool health monitoring for early detection of tool performance anomalies and timely prompting operators to perform maintenance to reduce tool downtime.

[0004] Against this background, the implementation plan was created. Summary of the invention

[0005] Methods and systems for real-time tool health status monitoring are presented. Embodiments include methods and systems for using in-situ sensors to monitor airborne particles, whose measurement data is used to monitor maintenance needs of semiconductor processing systems. Door status data regarding the timing of various door open / close operations in the semiconductor processing system is used to identify the source of airborne particles to provide recommendations regarding maintenance procedures.

[0006] In one embodiment, a method is provided. The method includes the following operations: when both the door facing the equipment front end module (EFEM) and the door facing the vacuum transfer module (VTM) are closed, a load lock (LL) is evacuated to a vacuum pressure to match the pressure of the VTM. The LL is docked with the VTM through the door facing the VTM and docked with the EFEM through the door facing the EFEM, wherein the VTM is docked with a plurality of processing modules (PMs), and the plurality of PMs have corresponding plurality of PM doors. The method also includes the following operations: opening the door facing the VTM so that airborne particles present in the VTM can diffuse into the LL, the airborne particles originating from the VTM or the plurality of PMs. The method also includes the following operations: closing the door facing the VTM and venting the LL to the atmosphere; and using a sensor in fluid communication with the LL, obtaining measurement data of the airborne particles diffused into the LL during the opening of the door facing the VTM. The method further includes the following operations: using a maintenance detection module, based on a quantitative analysis of the measurement data, determining a recommended maintenance procedure for one of the LL, the VTM, or the plurality of PMs.

[0007] In another embodiment, another method is proposed. The method includes the following operations: when the door of the LL facing the VTM is closed, the door facing the EFEM is opened, and the LL is docked with the EFEM through the door facing the EFEM, and is docked with the VTM through the door facing the VTM. Opening the door facing the EFEM allows airborne particles present in the EFEM to diffuse into the LL. The method also includes the following operations: using a sensor connected to the LL fluid, obtaining measurement data of the airborne particles diffused from the EFEM into the LL. The method also includes the following operations: using a maintenance detection module, based on a quantitative analysis of the measurement data, determining a recommended maintenance procedure for the LL or the EFEM.

[0008] In another embodiment, a system is provided. The system includes a LL that is docked with an EFEM through a door facing the EFEM and docked with a VTM through a door facing the VTM, wherein the VTM is configured to dock with multiple PMs, and the multiple PMs have PM doors. The system also includes a sensor that is in fluid communication with the LL for obtaining measurement data of airborne particles diffused into the LL by the EFEM or the VTM. The system also includes a valve that is arranged along a pipeline between the sensor and the LL, the valve being open when the pressure of the LL is at atmosphere, and the valve being closed when the pressure of the LL is at vacuum. The system further includes a computer module that is configured to process the measurement data received from the sensor to determine a recommended maintenance procedure for the LL, the VTM, the EFEM, or one of the multiple PMs.

[0009] Other aspects will become apparent from the following detailed description taken in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The implementation scheme of this case can be better understood by referring to the following description in conjunction with the accompanying drawings.

[0011] Figure 1 A semiconductor processing system is depicted showing sensors providing in-situ health status monitoring of its various chambers, according to one embodiment.

[0012] Figure 2 A schematic diagram depicts particle movement at various stages of a load lock (LL) and subsequent sensor detection during substrate transfer between an equipment front end module (EFEM) and a vacuum transfer module (VTM) and vice versa, according to one embodiment.

[0013] Figure 3 A table is depicted showing various steps associated with an in situ airborne particle monitoring and counting method according to one embodiment.

[0014] Figure 4 A semiconductor processing system and multiple door open times for correlating particle counts with contamination sources for maintenance monitoring are described according to one embodiment.

[0015] Figure 5 According to one embodiment, a maintenance detection module of a computer is shown that uses various door status data and various sensor data to detect that a component within a substrate processing system requires maintenance.

[0016] Figure 6 According to one embodiment, a method is shown for identifying the source of airborne particles and detecting maintenance needs for a semiconductor processing system using in-situ sensors.

[0017] Figure 7 The interface where the EFEM meets the LL is depicted according to one embodiment, along with various locations for placing sensors for tool health monitoring purposes.

[0018] Figure 8 Experimental data collected from various particle sensor locations under various conditions is shown according to one embodiment.

[0019] Fig. 9 The overall flow of a method is shown according to one embodiment for automatically determining whether a maintenance procedure is recommended for a tool based on sensor measurement data. DETAILED DESCRIPTION

[0020] The following embodiments describe systems, methods, and apparatus for monitoring the health of components of a cluster tool for semiconductor manufacturing. The embodiments herein describe solutions to address the current difficulties in real-time and automatic monitoring of tool health. The health of a cluster tool generally refers to the physical state of a cluster tool, including the physical state of any subset of its components. Tool health may also refer to whether one or more components of a tool require or will functionally or structurally benefit from maintenance, repair, cleaning, replacement, refurbishment, etc. The embodiments described herein provide data related to tool health monitoring in an automated manner by measuring airborne particles originating from various internal components of a cluster tool. The measured data of airborne particles (e.g., particle counts, particle sizes, etc.) are associated with component usage data (e.g., the number of rotations of a pivot, the total amount of time a door is open / closed) to identify one or more components that will benefit from maintenance and the nature and extent of such maintenance. The process of identifying components that will benefit from maintenance and estimating the nature and extent of maintenance may be referred to as maintenance detection. The embodiments contemplated herein use machine learning to improve the process of associating the measured data of airborne particles with the required maintenance detection.

[0021] A typical cluster tool may have tens of thousands of individual parts. A subset of these parts may require regular or irregular maintenance, including manual cleaning, "automatic cleaning", repair, refurbishment, calibration, replacement, modification, upgrade, surface reconditioning, etc. In some cases, these parts are susceptible to process conditions (e.g., plasma, electric fields, magnetic fields, high or low temperatures, high or low pressures, exposure to reactants, exposure to ultraviolet (UV) radiation, exposure to high-speed ions, etc.). These components may include, for example, substrate supports, electrodes, chamber walls, showerheads, chucks, exclusion rings, wafer contact supports, etc. In addition, some of these parts may suffer from mechanical wear caused by (e.g.) friction caused by angular or linear motion and physical deformation. These components may include bearings, motors, O-rings, gaskets, door seals, pistons, etc. These conditions or physical changes in component properties are generally referred to as degradation, deterioration, or reduction.

[0022] One of the main signs or indicators of component degradation and therefore maintenance is the release or sublimation of particles into the gaseous state. This is especially true under vacuum conditions, where the energy required for sublimation is reduced compared to atmospheric air. When a component (e.g., a seal) is repeatedly subjected to process conditions or friction or deformation, the component is subject to "wear and tear" sufficient to cause its underlying intermolecular, supramolecular, or even compositional changes. These molecular changes typically result in particle separation from an underlying monolithic part at the surface where the part contacts the open space. It is from this surface that individual particles sublimate or otherwise become airborne. As a result, airborne particles indicate wear of the part and the need for maintenance.

[0023] To date, tool monitoring (e.g., by detecting these airborne particles) has presented several challenges. From an architectural standpoint, because cluster tools are operated in clean rooms, and clean room space is valuable, cluster tools are designed to occupy as little floor space as possible. In addition, because of the dynamic nature of the temperature and pressure of the process modules (PMs), the internal volume of the cluster tool is also kept to a minimum to reduce, for example, the time and cost required to reach a specific temperature, pressure, or reactant concentration within the reaction chamber. Therefore, there is little extra space within the cluster tool to place monitoring equipment, such as optical or light-based sensors.

[0024] Furthermore, conventional and manual inspection of cluster tools (e.g., vacuum transfer modules (VTMs), load locks (LLs), atmospheric transfer modules (ATMs), and equipment front end modules (EFEMs)) is time consuming, inefficient, and expensive because doing so requires stopping any processing. Additionally, manual inspections may require some degree of tool disassembly, resulting in increased downtime. Furthermore, such manual inspections may often unnecessarily stop operations or otherwise fail to detect maintenance issues until they cause defects in substrate processing. A method and system for in-situ monitoring of tool health and detecting maintenance needs is presented, enabling operators to identify and address maintenance needs before failures occur.

[0025] Figure 1 A semiconductor processing system 100 is depicted according to one embodiment, showing sensors 101 providing in-situ health status monitoring for its various chambers. The semiconductor processing system 100 is shown to include an equipment front end module (EFEM) 102, which an operator 112 can interface with through a computer 103 in, for example, a clean room. The EFEM 102 can also interface with other semiconductor processing modules through one or more robots. In the figure, a cassette 114 carrying a plurality of substrates 116 is transferred to the EFEM 102. The ATM 104 operates under atmosphere and is used to transfer one of these substrates 116 from the cassette 114 to a load lock (LL) 106. The ATM 104 is located inside the EFEM 102 and is equipped with at least one robot 118, which is used to perform such transfers between the LL 106 and the EFEM 102.

[0026] When the ATM 104 transfers the substrate 116 to the LL 106, the door 130 facing the EFEM is opened, and the robot 118 places the substrate 116 on the support 120. Then, the door 130 facing the EFEM is closed, and then the pressure in the LL 106 is evacuated by the pump 128 to match the vacuum of the vacuum transfer module (VTM) 108. Once the LL 106 is in a vacuum state, the door 132 facing the VTM is opened, and the robot 122 of the VTM 108 transfers the substrate 116 from the support 120 to the support 124 of the process module (PM) 110. As an example, the PM 110 can perform plasma etching, chemical vapor deposition (CVD), atomic layer deposition (ALD), ion beam etching (IBE), sputtering, etc.

[0027] When processing of the substrate 116 is complete, the robot 122 of the VTM 108 retrieves the substrate 116 from the PM 110 and places it back on the support 120. With the doors 130 and 132 facing the VTM and ATM closed, the LL 106 is vented to the atmosphere through the vent 126. Once in the atmosphere, the door 130 facing the EFEM is opened so that the robot 118 can retrieve the substrate 116 from the LL 106 and place it back in the cassette 114 or other storage structure. There may be more than one substrate 116 in the LL 106 at a time. For example, when placing a substrate 116 from the VTM 108 to the LL 106, an unprocessed substrate 116 can be placed in the LL 106. Therefore, the robot 122 can place processed substrates 116 into the LL 106 and remove unprocessed substrates 116 from the LL 106 for processing in sequence.

[0028] The sensor 101 is shown associated with the semiconductor processing system 100 near the door 130 facing the EFEM, such that the sensor 101 is within the LL 106 and / or the ATM 104, or is otherwise in fluid communication therewith. Other locations for the sensor 101 are contemplated and referenced. Figure 7 102, LL 106, VTM 108, and PM 110. The sensor 101 is in communication with a computer 103 to send measurement data related to the measurement of airborne particles. In some embodiments, the sensor 101 is configured to provide particle counts, particle sizes, and particle compositions as measurement data to the computer 103. The computer 103 also communicates with the sensor 101 to indicate when the sensor 101 performs a measurement and when it does not. In some embodiments, since some measurements cannot be performed under vacuum, the sensor 101 performs the measurement in the atmosphere. In some embodiments, the sensor 101 can be a sensor array having a plurality of sensors that provide measurement data on particle counts, particle sizes, particle compositions, relative humidity (RH), temperature, pressure, oxygen concentration, and the like. The measurement data can further be used to narrow down, for example, which compartment or which components within the compartment require maintenance and the nature and extent of the maintenance.

[0029] According to one embodiment, Figure 2A schematic diagram depicting particle movement and subsequent sensor 106 detection at multiple stages 204-210 of LL 106 during substrate transfer between EFEM 102 and VTM 108 is depicted. In stage 204, particles 200 from EFEM 102 are represented as relatively large dots, while particles 202 from VTM 108 are represented as smaller dots. This is done for illustrative purposes only, and the relative sizes of particles 200 and 202 do not reflect the actual sizes of those particles. The apparent density of particles 200 and 202 also does not represent the concentration or particle count of these particles in their respective chambers.

[0030] The particle sensor 101 is shown in fluid communication with the LL 106, for example, by a pipe connected to a foreline associated with a pump. At stage 204, the door facing the EFEM (or the door facing the ATM) is shown as being open, for example, so that a substrate can be placed into the LL 106. In addition, during stage 204, particles 200 originating from the EFEM 102 can diffuse into the LL 106. Due to such diffusion, the concentration of particles 200 in the LL 106 may be increased. The change in can be estimated by the following equation:

[0031]

[0032]

[0033] The change in concentration of particles 200 in the LL 106 due to diffusion is determined by the flux J of particles 200 entering the LL 106. Furthermore, in equation (2), D is the diffusion coefficient of the particles 200, which itself may depend on a number of factors, including temperature and the size of the particles 200. Thus, at stage 204, the concentration of particles 200 in the LL 106 is time dependent. The concentration of particles originating from the EFEM 102 may be obtained by knowing, for example, the amount of time that the door to the EFEM is open and the particle counts obtained by the sensor 101.

[0034] Next, LL 106 is evacuated by a pump to a vacuum pressure similar to the vacuum of VTM 108. At stage 206, particles 200 are shown as having been pumped out of LL 106, while both the door facing the VTM and the door facing the EFEM are closed. Because LL 106 is under vacuum in stage 206, sensor 101 can be indicated as closed and can be sealed from the vacuum of LL 106 by a valve (not shown).

[0035] Next, in stage 208, the door facing the VTM is opened so that unprocessed substrates can be transferred from LL 106 to VTM 108, and processed substrates can be transferred from VTM 108 to LL 106. In addition, when the door facing the VTM is opened, particles 202 originating from VTM 108 can diffuse into LL 106. The particles 202 shown in VTM 108 may have multiple sources. For example, the particles 202 may originate from VTM 108 itself, which is caused by particle sublimation from seals, bearings, motors, bushings, contact surfaces, etc. of various components associated with VTM 108 (e.g., robot arm, door facing VTM, or PM door). In addition, the particles 202 may originate from PM itself, which is residual airborne particles diffused from PM after processing when the PM door is opened.

[0036] In addition, the concentration of particles 202 in LL 106 will depend on the time that circulation is allowed to occur, for example, the opening time of the door facing the VTM. Between stages 208 and 210, the door facing the VTM is closed, and LL 106 is ventilated to reach atmospheric pressure. In some embodiments, nitrogen is used for ventilation. In stage 210, the contents of LL 106 flow to particle sensor 101 to detect airborne particles originating from VTM 108. The concentration of particles 202 from VTM 108 can be inferred from particle counting data. The concentration of particles 202 in LL 106 may remain unchanged during ventilation, because the number of particles in the volume of LL 106 should not change due to ventilation.

[0037] Between stages 210 and 204, another cycle of evacuation and ventilation may occur to clear particles 202 from LL 106. However, in other embodiments, this step may not be necessary. For example, a door facing the EFEM may be opened to allow particles 200 to diffuse into LL 106 while also allowing particles 202 to diffuse out of LL 106. During this gas exchange, sensor 101 may be instructed to provide real-time particle count measurement data to determine whether there is a net increase or net decrease in particle count. If there is a net increase, it may indicate that the concentration of particles 200 in EFEM 102 is higher than the concentration of particles 202 in VTM 108. On the other hand, if there is a net decrease in particle count, it may indicate that the concentration of particles 202 in VTM 108 is higher than the concentration of particles 200 in EFEM 102.

[0038] According to one embodiment, Figure 3A table is depicted showing the various steps associated with the in-situ airborne particle monitoring and counting method. As described above, in step 1, a wafer (unprocessed) is moved from the EFEM to the LL when the door facing the EFEM is open and the door facing the VTM is closed. In some embodiments, step 1 may also include moving a processed wafer located in the LL to the EFEM before or after the unprocessed wafer is moved to the LL. In step 1, the sensor is open to count particles propagated from the EFEM. In step 2, the wafer is in the LL, and the doors of the EFEM and VTM are closed, while the LL pressure is evacuated to vacuum. The sensor is closed or disconnected from the fluid communication with the LL by a valve. This is because particle sensors generally cannot operate accurately at high vacuum pressures due to the absence of gas flowing through the sensor channel.

[0039] In step 3, another processed wafer may be in the VTM so that it can be transferred to the LL while the unprocessed wafer is moved from the LL to the VTM. In addition, in step 3, there will be a gas exchange between the LL and the VTM, because some particles that diffused into the LL during step 1 may still be present in smaller amounts in step 3. At the same time, the particles in the VTM diffuse into the LL. As mentioned above, these particles may have many sources, including the VTM itself or one of the multiple PMs. Between steps 3 and 4, various processing procedures may be performed on the substrate in the PM. The sequence of processing steps may generate particles, which will accumulate in the VTM during various transfer steps between these PMs. During processing, the door facing the VTM may be closed, and the LL may undergo ventilation and exhaust cycles to transfer the processed wafer from the LL to the EFEM.

[0040] In any case, step 4 is used to transfer the currently processed substrate from VTM to LL. When doing so, airborne particles from VTM that have accumulated for a period of time during one or more processing steps of PM will diffuse into LL. The rate of diffusion into LL will depend on the diffusion coefficient of the particles, which will depend on temperature, pressure, the mass of the molecule, the size of the molecule, the size of the opening of the door facing VTM, and the volume of LL. The time when the door facing VTM is opened will be recorded or obtained from the sequence data (if it is predetermined). In addition, the door opening time and the opening frequency will also be recorded to infer the particle source from particle counting and other measurement data. In addition, the speed at which the door is opened and closed can be considered, because when the door is opening or closing, the area where the particles flow through will change.

[0041] Steps 5 and 6 show that when the LL is vented to the atmosphere, airborne particles that migrate into the LL remain in the LL. The concentration of these particles may not change due to ventilation because the volume remains unchanged. Then, the airborne particles originating from the VTM accumulated during the processing steps are measured by the sensor. The cycle restarts with step 7, in which the processed wafers are transferred from the LL to the EFEM in exchange for unprocessed wafers. In step 7, when the door facing the EFEM is opened, the sensor can be opened to determine the change in particle count. At this stage, airborne particles originating from the VTM diffuse into the EFEM, and airborne particles originating from the EFEM diffuse into the LL. As a result, there may be a net increase or decrease in the number of particles due to the gas exchange in step 7. In other embodiments, the LL can be evacuated and ventilated for a cycle to remove the particles measured in step 6.

[0042] Figure 4 A semiconductor processing system 100 and various door opening times are depicted according to one embodiment, which are used to associate particle counts 101-1 with contamination sources for maintenance detection. It is shown that a substrate 116 is transferred between an EFEM 102 and a LL 106 by a robot 118 of an ATM 104, and the robot 118 has an end effector 118a. The substrate 116 is also transferred between the LL 106 and PMs 400, 402, 404, and 406 by a robot 122 having an end effector 122a. The semiconductor processing system 100 shows various doors that are opened and closed during wafer transfer and processing, which may be referred to as "door states" herein. That is, the door state data represents the timing of when various doors are opened or closed and the duration. For example, the door 130 facing the EFEM is shown as being associated with the door state 130a facing the EFEM, and the door state 130a is a function of time. In the period shown, the door state 130a facing the EFEM is shown as being in the open state twice. Typically, the door state data can be obtained via observation measurement or via sequencer programming data, which indicates when each door should be opened or closed according to programming. In addition, the speed at which the door opens or closes can be obtained from manufacturer data or via observation measurement.

[0043] In any case, the EFEM-oriented gate state 130a can be used to calculate the particle flux from the EFEM 102 into the LL 106, for example, by solving equations (1) and (2). The same is true for the VTM-oriented gate state 132a. In the figure, the semiconductor processing system 100 is part of four PMs 400, 402, 404 and 406, each of which is associated with a PM gate 401, 403, 405 and 407, which is used to interface with the VTM 108. Each PM gate 401, 403, 405 and 407 is associated with a PM gate state 401a, 403a, 405a and 407a. For example, PM 1 gate state 401a shows that PM 1 gate 401 is opened and closed four times in any period of time used. For each repetition, the amount of time that PM 1 gate 401 is open is relatively shorter than the gate open time of (for example) the VTM-oriented gate 132. PM 2 door status 403a shows that PM 2 door 403 was opened and closed twice during the period used, while PM 3 door status 405a shows that PM 3 door 405 was opened and closed three times. In addition, PM 4 door status 407a shows that PM 4 door 407 remained closed during the period.

[0044] The gate states of PM gates 401, 403, 405, and 407 can be used to calculate the relative amount of gas exchange between each of PM 400, 402, 404, and 406 and VTM 108. If the particle count 101-1 reading is obtained from airborne particles originating from VTM 108 (e.g., Figure 3 6), the particle counts may be proportionally attributed to PM 400, 402, 404, and 406, depending on the PM gate states 401a, 403a, 405a, and 407a.

[0045] The flux will depend on the amount of time any of the PM gates 401, 403, 405, and 407 are open, and the manner or frequency in which they are opened. For example, assume that PM 1 gate 401 and PM 2 gate 403 are open for the same total amount of time, as shown by PM gate states 401a and 403a. Further assume that the concentrations of a given particle within PM 400 and 402 are substantially similar, and that the relative concentrations of particles in PM 400 and 402 and VTM 108 are such that equilibrium is not reached within the gate open time shown. Despite both having the same total amount of gate open time, the degree of diffusion between PM 400 and VTM 108 may be greater than the degree of diffusion between PM 402 and VTM 108. This phenomenon can be described by Fick's first law equation (2).

[0046] When applied to the present situation, when diffusion begins after PM gates 401 and 403 are opened, the size of the flux depends on the concentration gradient between VTM108 and PM 400 and 402. Because the gradient will tend to decrease when diffusion occurs, the flux will also decrease with the time the gate is open. When the PM gate is closed, it enables the gradient to be reestablished via diffusion within the compartment. Therefore, under the conditions, even if PM gate 401 and PM gate 403 both have the same total opening time, the net diffusion from PM400 to VTM 108 may be greater than the net diffusion of PM 402. Therefore, when performing particle measurement by sensor 101, in addition to the total time the gate is open, the PM gate states 401a, 403a, 405a and 407a are also used to associate the source of a given particle.

[0047] Figure 5 A maintenance detection module 500 of a computer 103 is shown according to one embodiment, which uses various door status data 503 and various sensor data 505 to detect that components within a substrate processing system require maintenance. It is shown that the computer 103 has door status data for each of the doors facing the VTM, the doors facing the EFEM, and each PM door. The door status data 503 can be converted into flux data 501, which describes the degree of diffusion between two compartments in view of the nature and degree of the door status data 503. The flux data 501 is input into the maintenance detection module 500. In addition, sensor data 505 can be collected in real time for several characteristics. For example, one or more of particle counts, volatile organic compound (VOC) measurements, relative humidity (RH) measurements, temperature and pressure measurements, and oxygen level measurements can be obtained by the sensor 101.

[0048] This data is input into the maintenance detection module 500 for determining, for example, the characteristics of the detected particles, including particle count and size, whether the particles are organic or inorganic, etc. Based on this data, the maintenance detection module 500 can classify the source of contamination as oil leaks, seal degradation, or process chemical residues, etc. In addition, based on the door status data 503, the maintenance detection module 500 determines the proportion of particle characteristics caused by each of the various compartments (e.g., VTM, EFEM, and PM1-4). Therefore, the maintenance detection module 500 is able to determine the source of gas-suspended particle detection and the most favorable type of maintenance. For example, in Figure 5 In FIG. 5 , the maintenance detection module 500 generates a prompt 502 that part x of PM2 needs to be replaced within 10 days.

[0049] In conjunction with the machine detection module 500, it is also contemplated to use a machine learning module 504 to predict with greater accuracy and confidence the components that require maintenance and the nature and extent thereof. The machine learning module 504 may be implemented by the computer 103, or may be implemented on a remote server in communication with the computer 103. In any case, the machine learning module 504 may use a supervised learning algorithm that detects features associated with the various door status data 503 and sensor data 505 for classification by a maintenance detection model. Training data may be provided to the machine learning module 504 that provides ground truth regarding features that sets of sensor data and door status data will have when various components require replacement (e.g., due to wear, degradation, contamination, deterioration, etc.).

[0050] Any suitable machine learning algorithm may be implemented to meet the needs of the operator and the machine learning module 504. Certain algorithms may include, but are not limited to, Bayesian networks, linear regression, decision trees, neural networks, k-means clustering, and the like. Machine learning is considered "supervised" because when an operator is prompted to perform certain maintenance, the operator can then evaluate the accuracy of the machine learning prompt when performing the maintenance. The operator can mark the prompt as "correct" (true) if the prompt is accurate; if the prompt is incorrect, the operator can mark the prompt as "incorrect" (false), along with any comments indicating his or her findings on the actual maintenance need. In this way, the machine learning algorithm can learn from the operator of a specific machine, and if the machine learning algorithm is distributed across many computers 103, the machine learning algorithm can learn from thousands of semiconductor processing machines and operators. In certain embodiments, the maintenance detection module 500 and / or the machine learning module 504 can be provided as a service to the operator of the semiconductor processing system.

[0051] Figure 6According to one embodiment, a method is shown, which is used to identify the source of airborne particles and detect the maintenance requirements of semiconductor processing systems by in-situ sensors. In operation 600, the method provides evacuating the load lock containing the substrate to a first pressure to match the pressure of the VTM, and in many cases, the first pressure is a high vacuum. Once the load lock is in vacuum, operation 602 is used to open the door facing the VTM of the load lock so that the substrate can be transferred to the VTM. The door facing the VTM can then be closed. The robot arm of the VTM continues to place the substrate into one of multiple PMs for processing, including but not limited to, deposition by physical vapor deposition (PVD), chemical vapor deposition (CVD), atomic layer deposition (ALD), electrochemical deposition (ECD), plasma etching by transformer coupled plasma (TCP) or inductively coupled plasma (ICP), ion beam etching (IBE), lithography and sputtering. During these processes, the door of the PM will open and close, and any airborne particles from the PM are allowed to accumulate in the VTM. The composition of the airborne particles in the VTM will depend on the door state of the PM. PMs with longer average opening times contributed more to the composition than PMs with shorter average opening times.

[0052] When the processing of the substrate is completed, operation 606 is used to open the door facing the VTM so that the substrate can be retrieved from the VTM and arrive at the LL. During the opening of the door facing the VTM, airborne particles that have accumulated in the VTM will have the opportunity to diffuse into the LL. In some embodiments, before the door facing the VTM is opened, the contents of the VTM may have time to be homogenized. In any case, the door status data facing the VTM is obtained empirically or from the sequence generator of the computer. The sequence generator will contain data related to the number of times each door is opened and closed. The opening / closing time of the door can be estimated by prior correction or via manufacturer data.

[0053] Operation 608 is used to capture airborne particles that have diffused into the LL by closing the door facing the VTM. In addition, operation 608 is used to ventilate or re-pressurize the LL to a second pressure, and in many embodiments, the second pressure is atmospheric pressure. In addition, operation 608 is used to obtain the measurement data of the contents of the LL, and the contents of the LL should include any airborne particles accumulated in the VTM during processing and diffused into the LL during wafer transfer. In operation 610, the method processes the measurement data obtained by the sensor and the door status data to identify one or more PMs, or LLs, or VTMs that need to be maintained. In some embodiments, any airborne particles from the EFEM or VTM can be evacuated and then the LL can be ventilated to measure the LL separately. Any airborne particles in the LL may be pollution from the LL itself. These airborne particles can be measured by sensors in a manner similar to operation 608. Therefore, in operation 612, the method can generate prompts that notify the operator that one or more of the LL VTM or PMs need to be concerned. In some embodiments, the prompt generated at operation 612 may be used to perform an automated cleaning on one of the VTM, LL, or PM. In addition, the prompt in operation 612 may specify the type of maintenance that is predicted to be desired. For example, the prompt may specify O-ring degradation, oil leakage, lip seal leakage due to mechanical or chemical wear.

[0054] According to one embodiment, Figure 7The joint 700 where the EFEM 102 and the LL 106 meet is depicted, as well as various locations where sensors 101a-f are placed for real-time tool health monitoring. Within the load lock 106, an outlet 708 is shown at the lower horizontal surface 106a to connect the LL 106 to the pump 128 for evacuating the LL 106. A foreline 710 and a valve 712 are provided between the outlet 708 and the pump 128. One of the particle sensors 101a is shown connected to the foreline 710 via a line 714 having a valve 716. The pressure sensor 708 is actually connected to the LL 106 and sends a signal to the particle sensor 101a. During operation, when the pump 128 is turned on and the valve 712 is turned on, the valve 716 is closed and the particle sensor 101a is closed. In this way, only the contents from the LL 106 are evacuated by the pump 128. Subsequently, when LL 106 reaches atmosphere by venting, pressure sensor 718 may be passed through line 714 and valve 716 opened to detect the pressure in LL 106 or foreline 710. This allows the contents from LL 106 (which is at atmosphere) to propagate into particle sensor 101a. Particle sensor 101a may be equipped with a fan to cause the contents of LL 106 to flow into particle sensor 101a. In this configuration, since the airborne particles are caused to flow into particle sensor 101a, it is relatively straightforward to measure the airborne particles present in LL 106. However, airborne particles originating from EFEM 102 may take time to diffuse into LL 106 and be measured by sensor 101a.

[0055] The particle sensor 101b is shown as being located inside the LL 106 and at the upper horizontal surface 106b, but the particle sensor 101b may also be placed on the lower horizontal surface 106a or the vertical surface 106c. In this configuration, the particle sensor 101b can be used to operate when the pressure inside the LL 106 is above a certain threshold. In some embodiments, the measurement data from the particle sensor 101b may not be reliable below certain threshold pressures. In some embodiments, the particle sensor 101b will include a wireless transmission module to transmit the measurement data to a computer. Any suitable wireless transmission bandwidth or protocol can be used to achieve this purpose.

[0056] The particle sensor 101c is shown mounted to the upper surface 130a of the door 130 facing the EFEM so that it remains in the atmosphere and obtains measurement data of the EFEM 102 when the door 130 facing the EFEM is closed, and obtains measurement data of the LL 106 when the door 130 facing the EFEM is open. In some embodiments, the particle sensor 101c is directional in the sense that the opening of the particle sensor 101c faces a certain direction. In certain embodiments, the opening of the particle sensor 101c may face the LL 106 or the EFEM 102, or may have an opening facing both. In another embodiment not shown, the particle sensor may be mounted on the vertical surface 130b of the door 130 facing the EFEM. Like the particle sensor 101b, the particle sensor 101c may include a wireless transmission module to transmit data to and receive data from a computer.

[0057] The particle sensor 101d is shown mounted to the vertical outer surface 106d just above the door opening 703 of the LL 106. The location of the particle sensor 101d is such that airborne particles present in the LL 106 do not need to diffuse far before being picked up by the sensor 101d. The particle sensor 101e is in fluid communication with the EFEM 102 through an outlet 701 and a valve 705. The outlet 701 may be defined from the upper horizontal surface 102a of the EFEM 102, or may be defined from the lower horizontal surface 102b. The particle sensor 101e may be equipped with a fan to cause airborne particles to flow from the EFEM 102 and the LL 106 into the particle sensor 101e.

[0058] The particle sensor 101f may be placed inside the EFEM 102 at the mouth of the balancing port 707 of the balancing channel 702. The balancing channel 702 ensures that the pressures between the LL 106 and the EFEM 102 are similar before the door 130 facing the EFEM is opened. The particle sensor 101f may be connected to the balancing port 707 for ease of use.

[0059] Figure 8Experimental data collected from various particle sensor locations under various conditions are shown according to one embodiment. The data points are for particles with a diameter of 300 nm or greater for a laser-based particle counter. Measurement period 800 is performed at a location within a clean room and results in a relatively high particle count. Subsequent measurement periods 802 and 804 show that particle counts within the EFEM and in front of the LL (e.g., sensors 101c and 101d) are generally significantly lower than particle counts within the clean room. When the door of the clean LL facing the EFEM is cyclically opened and closed, measurement period 806 shows an increase in particle counts (compared to measurement period 804 when the door of the LL facing the EFEM is closed). Part of the increase in particle counts may be due to increased air flow around the particle sensor, which is caused by the door opening and closing, and the exchange of gases between the LL and the EFEM. In measurement period 808, the particle sensor is held in front of the contaminated LL while the door facing the EFEM is cyclically opened and closed. The measured particle counts are at least twice as high as in the previous measurement period 806, indicating that contaminated LL is detectable using experimental methods.

[0060] According to one embodiment, Fig. 9 The overall flow of the method is shown, which is used to automatically determine whether a maintenance procedure is recommended based on sensor measurement data. In operation 900, when the door facing the EFEM and the door facing the VTM are all closed, the LL is evacuated to a vacuum pressure to match the pressure of the VTM. The LL is docked with the EFEM through the door facing the EFEM, and is docked with the VTM through the door facing the VTM. In operation 902, the door facing the VTM is opened so that the airborne particles in the VTM are diffused into the LL. Airborne particles are usually derived from one or more of the VTM or multiple PMs, but in theory some may be derived from the LL and the EFEM. In operation 904, the door facing the VTM is closed so that the LL can be ventilated toward the atmosphere. This is usually done using nitrogen, but other gases can be used. Without first opening the door facing the EFEM, airborne particles can be measured by a sensor that is in fluid communication with the LL. In this way, it will be known that the measured airborne particles are from the VTM rather than from the EFEM. In other embodiments, before using the sensor to measure, the door facing the EFEM can be opened. In operation 908, the maintenance detection module is used to determine a recommended maintenance procedure for one of the LL, VTM, or multiple PMs, or some combination thereof. The maintenance procedure recommendation is based on the measurement data and may also be based on the door status data. The maintenance detection module may also identify which of the LL, VTM, or multiple PMs should be recommended for a maintenance procedure. Consider performing a maintenance procedure at each pump down and vent cycle. Fig. 9The method shown in the figure can be used to achieve real-time monitoring of the health status of the tool. In other embodiments, evacuation and ventilation can be performed intermittently, for example after every 5, 10, 20, 50 or 100 or more evacuation and ventilation cycles.

[0061] Although the foregoing embodiments have been described in detail for the purpose of clear understanding, it is apparent that certain changes and modifications may be made within the scope of the appended claims. Therefore, these embodiments should be regarded as illustrative rather than restrictive, and these embodiments are not limited to the details set forth herein, but may be modified within the scope of the appended claims and equivalents.

Claims

1. A tool health status detection method, comprising: When the door facing the equipment front end module and the door facing the vacuum transfer module are both closed, the load lock is evacuated to a vacuum pressure to match the pressure of the vacuum transfer module, the load lock is docked with the vacuum transfer module through the door facing the vacuum transfer module, and is docked with the equipment front end module through the door facing the equipment front end module, wherein the vacuum transfer module is docked with a plurality of processing modules, and the plurality of processing modules have a corresponding plurality of processing module doors; Opening the door facing the vacuum transfer module so that airborne particles existing in the vacuum transfer module can diffuse into the load lock, the airborne particles originating from the vacuum transfer module or some of the plurality of process modules; closing the door facing the vacuum transfer module and venting the load lock to atmosphere; obtaining, using a sensor in fluid communication with the load lock, measurement data of the airborne particles diffused into the load lock during the opening of the door facing the vacuum transfer module; as well as Based on the quantitative analysis of the measurement data, a recommended maintenance procedure is determined for one of the load lock, the vacuum transfer module, or the plurality of process modules.

2. The method according to claim 1, further comprising: obtaining door status data regarding opening and closing times of the door facing the vacuum transfer module and the plurality of process module doors; Wherein, the determining to recommend the maintenance procedure is further based on the door status data.

3. The method according to claim 1, further comprising: recommending said maintenance procedure based on said determining, generating a prompt for display to an operator; The prompt enables the operator to perform the maintenance procedure on the vacuum transfer module, the load lock, or one of the plurality of process modules.

4. The method of claim 1, wherein the sensor comprises a laser-based particle sensor and the measurement data comprises particle counts, particle size, or particle composition.

5. The method of claim 1 , wherein said evacuating said load lock, said opening said door facing said vacuum transfer module, and said closing said door facing said vacuum transfer module and venting said load lock occur as part of normal operation of a tool having said load lock, said front end module, said vacuum transfer module, and said plurality of process modules.

6. The method of claim 1, wherein said determining a recommended maintenance procedure comprises communicating with a machine learning module, said machine learning module extracting features from said measurement data for classification by a classification model for said determining to recommend said maintenance procedure.

7. The method of claim 1, wherein the recommended maintenance procedure includes one or more of manual cleaning, automatic cleaning, replacement, recalibration, repair, or refurbishment.

8. The method of claim 1, wherein the measurement data comprises one or more of particle counts, particle size, volatile organic compound levels, oxygen levels, temperature, pressure, or relative humidity.

9. The method of claim 2, wherein a maintenance detection module uses the door status to calculate a ratio of airborne particles originating from the vacuum transfer module and each of the plurality of process modules to identify which of the vacuum transfer module, the load lock, or the plurality of process modules requires the maintenance procedure.

10. A tool health status detection method, comprising: When the door of the load lock facing the vacuum transfer module is closed, the door facing the equipment front end module is opened, the load lock is docked with the equipment front end module through the door facing the equipment front end module, and is docked with the vacuum transfer module through the door facing the vacuum transfer module, and the opening of the door facing the equipment front end module allows the airborne particles existing in the equipment front end module to diffuse into the load lock; obtaining measurement data of the airborne particles diffused from the equipment front end module into the load lock using a sensor in fluid communication with the load lock; as well as Based on the quantitative analysis of the measurement data, a recommended maintenance procedure for the load lock or the equipment front end module is determined.

11. The method of claim 10, wherein prior to said opening of said door facing the equipment front end module, said load lock is evacuated and subsequently vented.

12. The method according to claim 10, further comprising: Obtaining door status data regarding the opening and closing times of the door facing the equipment front end module; Wherein, the determining to recommend the maintenance procedure is further based on the door status data.

13. The method according to claim 10, further comprising: recommending said maintenance procedure based on said determining, generating a prompt for display to an operator; The prompt enables the operator to perform the maintenance procedure on the load lock or the vacuum transfer module.

14. The method of claim 10, wherein the sensor comprises a laser-based particle sensor, the measurement data comprises a particle count, and wherein the determining to recommend the maintenance procedure comprises: The particle count is detected to be above a threshold.

15. The method of claim 10, wherein the recommended maintenance procedure includes one or more of manual cleaning, automatic cleaning, replacement, recalibration, repair, or refurbishment.

16. The method of claim 10, wherein the measurement data comprises one or more of particle counts, particle size, volatile organic compound levels, oxygen levels, temperature, pressure, or relative humidity.

17. A tool health status detection system, comprising: a load lock docked with the equipment front end module through a door facing the equipment front end module and docked with the vacuum transfer module through a door facing the vacuum transfer module, wherein the vacuum transfer module is configured to dock with a plurality of process modules having process module doors; a sensor in fluid communication with the load lock for obtaining measurement data of airborne particles diffused into the load lock from the tool front end module or the vacuum transfer module; a valve disposed along the line between the sensor and the load lock, the valve being open when the pressure of the load lock is at atmosphere and the valve being closed when the pressure of the load lock is at vacuum; as well as A computer module is configured to process the measurement data received from the sensor to determine a recommended maintenance procedure for the load lock, the vacuum transfer module, the tool front end module, or one of the plurality of processing modules.

18. The system of claim 17, wherein the computer module is further configured to process door status data regarding opening and closing times of the door facing the equipment front end module, the door facing the vacuum transfer module, and the plurality of process module doors for use in determining to recommend the maintenance procedure for one of the load lock, the vacuum transfer module, the equipment front end module, or the plurality of process modules.

19. The system of claim 18, wherein the computer module identifies which of the load lock, the vacuum transfer module, the tool front end module, or which of the plurality of processing modules requires the maintenance procedure to be performed.

20. The system of claim 17, wherein the sensor comprises a laser-based particle sensor, the measurement data comprises a particle count, and wherein the determining to recommend the maintenance procedure comprises detecting that the particle count is above a threshold.

Images