Advanced process control method and system

Abstract

This invention proposes an advanced process control method, which includes capturing product history information within a time interval of a process cavity. The product history information includes the etching rate of various products and product groups in an etching process in the process cavity and the corresponding radio frequency power time of the process cavity. Based on the product history information, the method derives the etching rate curve of each product and product group in relation to the radio frequency power time and finds a nonlinear tangent point of the etching rate curve. The method also includes reading information of a process product and obtaining the coefficients and weights of the etching rate model equation applicable to the process product.

Description

Technical Field

[0001] This invention is generally related to an advanced process control method, and more specifically, to an advanced process control method and system based on etch rate. Prior Technology

[0002] Faced with fierce competition to produce higher-quality products, chip manufacturers have widely invested in building Advanced Process Control (APC) system architectures to provide optimal production decisions. APC typically consists of several engineering systems: Run-to-Run Control (R2R), Fault Detection and Classification (FDC), Overall Equipment Efficiency (OEE), and e-Diagnostic technologies. Its implementation primarily involves collecting massive amounts of process-related data for statistical analysis to reduce process variation, improve equipment operating efficiency, reduce human intervention adjustments, thereby increasing capacity and improving IC component yield. To achieve this goal, a large number of mathematical, statistical, and physical techniques are applied and deployed within the APC architecture.

[0003] As semiconductor device pattern sizes continue to shrink and foundries shift between different product types and linewidth generations, maintaining process variations within acceptable limits becomes extremely difficult. For example, process challenges may include decreased tooling productivity, increased operator interference, lower yields, and increased rework rates, all of which can lead to increased costs. While advanced process control (APC) is widely used to reduce process variations, most APCs cannot control or adjust parameters affected by multiple process steps. In some applications, noise from semiconductor manufacturing can influence and interfere with the design and implementation of APCs. This noise can originate from preceding steps, current operating conditions, and the operation of the APC itself.

[0004] For example, the condition of a machine after maintenance (or preventative maintenance, PM) and related process performance will vary depending on the specific machine used. For instance, for etching machines, different products and different RF power times will cause variations in the etching rate. In this context, how to address the differences between etching machines and perform process control on different types of products / linewidth generations while maintaining stable product yield and quality is a crucial issue. This is especially true as the semiconductor industry moves towards a low-volume, high-variety foundry model. Engineers need to frequently intervene after machine maintenance to perform pilot runs on various products and confirm product quality before continuing APC control. This undoubtedly overburdens engineers and impacts production cycles and costs. Therefore, although advanced process control technologies exist for various purposes, they still cannot meet all practical needs. Those skilled in the art must continue to research and improve existing advanced process control methods, particularly for etching machines and factors related to etching rate. Summary of the Invention

[0005] In view of the problems encountered by the aforementioned prior art, the present invention proposes an advanced process control method, characterized by finding the nonlinear cut point of a product or product group through the etching rate curve of the machine, and then determining the target etching rate of the current product based on the nonlinear cut point, the RF power time of the machine / cavity, and the pattern density of the product. In this way, process control of different types of products / linewidth generations can be carried out according to the differences in etching machines, and product yield and quality stability can be maintained.

[0006] One aspect of this invention is to provide an advanced process control method, comprising: providing a process cavity; reading product history information of the process cavity within a time interval, the product history information including the etching rate of various products and product groups in an etching process in the process cavity and the corresponding radio frequency power time of the process cavity; deriving the etching rate curve of each product and product group relative to the radio frequency power time based on the product history information, and finding a nonlinear tangent point of the etching rate curve; providing a process product; and reading information of the process product and obtaining coefficients and weights of an etching rate model equation applicable to the process product, wherein the step of setting the etching rate model equation includes: determining whether the difference in etching rate of the process cavity before and after preventive maintenance is greater than or equal to a first set value; determining whether the current radio frequency power time of the process cavity is less than the nonlinear tangent point; and determining whether there is product history information of the same product or product group within a valid time period.

[0007] Another aspect of the present invention is to provide an advanced process control system for controlling a process cavity, comprising: a memory for storing product history information of the process cavity within a time interval, the product history information including etch rate curves of various products and product groups undergoing an etching process in the process cavity, and pattern densities of the products and product groups, wherein the etch rate curve is a curve relating the etch rate of each product and product group to the radio frequency power time of the corresponding process cavity; and an advanced process control module for controlling the process cavity according to a plurality of instructions, the instructions being used to perform the following steps: calculating the slope between each data point in the etch rate curve and the slope difference before and after the data point; finding the data point with the largest slope difference as a nonlinear tangent point; and controlling the process cavity to process a wafer according to the radio frequency power time, the nonlinear tangent point, and the pattern density.

[0008] These and other objects of the present invention should become more apparent to the reader after reading the detailed description of the preferred embodiments, which are illustrated in various figures and drawings below. Simple Explanation of the Diagram

[0009] This specification includes accompanying drawings, which form part of this specification, to provide the reader with a further understanding of the embodiments of the invention. These drawings depict some embodiments of the invention and, together with the description herein, illustrate its principles. In these drawings: Figure 1 is a functional block diagram illustrating various types of devices in the system of the present invention; Figure 2 is a trend graph of the etching rate of the etching machine / cavity versus the radio frequency power time (RF time) according to Embodiment 1 of the present invention; Figure 3 shows the relationship between the etching rate of the etching machine / cavity and the product pattern density according to Embodiment 1 of the present invention; Figure 4 is a block diagram illustrating the process of determining the nonlinear tangent point of the etching rate of an etching cavity by the APC system according to an embodiment of the present invention. Figure 5 is a block diagram of the APC system's condition monitoring process for an etching machine / cavity according to an embodiment of the present invention; and Figure 6 is a flowchart of an APC control method according to an embodiment of the present invention. It should be noted that all illustrations in this specification are for illustrative purposes only. For clarity and ease of explanation, the size and scale of the components in the illustrations may be exaggerated or reduced. Generally, the same reference symbols in the illustrations are used to indicate corresponding or similar component features in modified or different embodiments. Implementation

[0010] This invention relates to the manufacture of semiconductor devices, and particularly to process control in semiconductor device manufacturing. It should be noted that the several embodiments provided are merely examples illustrating the broad concept of this invention, and those skilled in the art can readily apply the concepts disclosed herein to other methods or apparatuses. Furthermore, the methods and apparatus of this invention include several common structures and processes. Since these common structures and processes are well-known to those skilled in the art, they are only discussed in general terms. In addition, several element symbols are used repeatedly in this invention; this repetition is for simplification and does not represent a relationship between different features and / or steps of the invention.

[0011] Figure 1 is a functional block diagram illustrating various types of apparatus 10 in the system of the present invention. Apparatus 10 is a partial semiconductor manufacturing system. In this embodiment, apparatus 10 is used to etch a plurality of semiconductor wafers, one of which is represented by component symbol 11 (wafer 11). Apparatus 10 includes a computer 12, an equipment block (etching machine 13), and a metrology machine 14. The etching machine 13 is a common device in the semiconductor manufacturing field, such as etching machines manufactured by companies like Lam Research, AMAT, and TEL. It includes an etching reaction chamber (not shown), and the semiconductor wafer 11 is disposed within the etching reaction chamber for the etching process. The wafers described below are all evaluated and inspected using the metrology machine 14 to assess their process-related results.

[0012] In Figure 1, although physical measurements of the wafer are performed in the measurement equipment 14, virtual measurement techniques, common in the art, can also be used to obtain measurement results in advance without actually performing physical measurements. Depending on factors such as process characteristics and product density, virtual measurement can reduce or replace real or physical measurements for a period of time.

[0013] Referring again to Figure 1. The computer 12 includes computer hardware, which may be common, commercial, or other suitable computer hardware. The computer hardware of the computer 12 includes a processor 21 and memory 22. The memory 22 is used to store the computer program 23 executed by the processor 21, enabling the computer 12 to control the etching machine 13. Regarding the control of the etching machine 13, the computer 12 includes etching process information to represent the process characteristics of the etching process performed by the etching machine 13. These process characteristics are represented by component symbols 36 (process characteristics 36) and include the process recipe for processing the wafer. For the etching process performed by the etching machine 13, the process characteristics 36 typically include temperature, pressure, chemical gases, and theoretically or practically measured etching rates of the electron-excited reaction. Some process characteristics 36 are provided with an Advanced Process Control (APC) module 46 with up to multiple inputs and a real-time estimation module 61, and the Advanced Process Control module 46 is used to control the etching machine 13. Technically, the advanced process control module 46 is part of program 23, but for better illustration, the advanced process control module 46 is shown separately in Figure 1. The advanced process control module 46 is configured in a manner common in the art, and will not be described further here.

[0014] Referring again to Figure 1. As previously described, after the etching machine 13 etches the wafer, the metrology machine 14 then evaluates all or part of the wafer 11. The metrology information obtained from the evaluation is provided back to the computer 12 via paths 51 and 52. Specifically, part of the metrology information via path 51 is provided directly back to the advanced process control module 46, and another part of the metrology information via path 52 is provided to the real-time estimation module 61. Technically, the real-time estimation module 61 is part of program 23. For simplicity, the etching process performed by the etching machine 13 of this invention is used to etch trenches into the material layer on the wafer 11. The trench depth is one of the parameters measured by the metrology machine 14, which is related to the current etching rate of the process. Typically, the system selects a specific process feature 36 to achieve the target trench depth. However, the actual trench depth and etching rate of the etching process are affected by the actual process conditions. For example, the characteristics of the components in the etching machine or the circuit pattern design (such as pattern density) of the wafer being etched can affect trench formation. Furthermore, the state of the etching reaction chamber also affects the etching process. For instance, the saturation of the chemical gas within the etching reaction chamber accumulates over time and gradually changes the etching rate under the original control conditions. On the other hand, after preventative maintenance (PM), the etching rate of the etching reaction chamber becomes less stable due to changes in the gas saturation within the chamber. The measurement unit 14 is used to estimate the trenches formed on the wafer 11 by the etching process performed by the etching machine 13. The measurement information, including the measured trench depth, is fed back to the computer 12 via path 52. The real-time estimation module 61 can make a real-time estimate of the actual etching rate provided by the etching machine 13 based on the measured trench depth, and use this actual etching rate to calculate the adjusted etching time for the etching process performed by the etching machine 13. Then, it provides the updated etching time to the advanced process control module 46 via path 66.

[0015] Please now refer to Figure 2, which is a trend graph of the etching rate versus radio frequency (RF) time for an etching machine or etching chamber according to Embodiment 1 of the present invention. In an etching machine, RF power time generally refers to a time parameter related to radio frequency (RF) signals, particularly the time related to the RF power drive during plasma etching. Plasma etching utilizes RF energy to excite gas molecules to generate plasma, which is used to etch the material surface; therefore, RF power time can represent the continuous operating time of the etching machine or etching chamber. This figure uses a preventive maintenance (PM) cycle (approximately 650 hours) of the etching chamber as an example to illustrate the relationship between RF power time and etching rate for wafers of different product groups (such as A, B, … F). Each point in the figure represents the etching rate data obtained from a single etching formulation treatment of a wafer of a certain product group. The product group referred to here can be classified based on the semiconductor manufacturing process used, such as various types of wafers and related products manufactured through different process technologies, different linewidth generations or versions, and / or various applications. These products are typically used in electronic products, such as mobile phones, computers, and automotive electronics, and may have the same or similar properties or characteristics in semiconductor manufacturing.

[0016] In this embodiment, as shown in Figure 2, different product groups will exhibit different etching rates during the etching process in an etching chamber, which may be related to the product's process generation, version, or pattern density. Generally, the same product or product group will have a steady-state target etching rate, which can be expressed in nanometers per second (nm / s). Furthermore, and more importantly, the etching rate of the etching machine will gradually increase from its lowest point in the range of RF power time less than 25 hours, and then tend to stabilize after that. This etching rate trend may be due to over-cleaning or improper component calibration of the etching machine immediately after preventative maintenance, or certain components of the machine (such as filters, vacuum pumps, etc.) not yet reaching their optimal state, causing some parameters of the machine to deviate from their optimal operating state, thereby affecting the etching effect and leading to a decrease in the etching rate. In other embodiments, the etching rate of the etching machine may also decrease from its highest point and then stabilize, or stabilize before increasing or decreasing, depending on the type of machine and / or the etching process performed, and is not limited thereto. The time it takes for the RF power etching rate to stabilize can vary depending on the type of equipment and / or the etching process being performed, and is not limited to the 25 hours illustrated. As this figure shows, to ensure process stability after equipment return and to improve production efficiency, we must develop solutions to address these etching rate characteristics of the etching equipment.

[0017] Please now refer to Figure 3, which shows the relationship between the etching rate of the etching machine / cavity and the pattern density of the product according to an embodiment of the present invention. To avoid the aforementioned variable of RF power time that affects the etching rate, this figure only uses data points where the RF power time is greater than 450 hours and the etching rate has stabilized. As shown in the figure, generally speaking, the etching rate of the etching machine gradually decreases as the pattern density of the product increases, exhibiting a linear distribution. This is because the higher the pattern density, the more material layer needs to be removed by the etching reaction, and therefore the etching rate will naturally be lower. As can be seen from this figure, in addition to the aforementioned RF power time of the machine / cavity, the pattern density of the product is also one of the factors affecting the change in etching rate, and the relationship between pattern density and etching rate changes will also be different for different linewidth generations G2 / G1. Especially in today's semiconductor industry, which is moving towards a small-batch, high-variety foundry model, various products have their own pattern density and the resulting etching rate targets and characteristics. Engineers need to frequently intervene after machine maintenance to perform trial production (Pi-Run) on various products and confirm product quality before continuing with APC control. This will undoubtedly overburden engineers and affect product cycle and production costs.

[0018] To address the aforementioned variations in etching rate, the APC model of this invention specifically switches between different algorithms for different RF power time periods and different product pattern densities. In an embodiment, for the RF power time period of less than 25 hours, the etching rate formula used by the APC model for the corresponding etching process or etching reaction chamber can be expressed by the following equation 1: (Equation 1)

[0019] In Equation 1, ERt represents the etching rate of the product to be processed, ERt-1 represents the etching rate of the most recently shipped product of the same type, and fEQ represents the RF power time compensation factor, which can be expressed by the following Equation 2: (Equation 2)

[0020] Where RFTlatest is the current RF power time of the etching cavity, RFTt-1 is the RF power time of the etching cavity when it was last shipped for the same product / product group, and αRF is the weight of the RF power time compensation factor on the etching rate (e.g., 0.01). From Equations 1 and 2, it can be seen that in this embodiment, for the RF power time range of less than 25 hours, the etching rate of the current product will gradually increase with the increase of the RF power time.

[0021] The above etching rate formula only considers the impact and compensation of RF power time on the etching rate. For different product specifications and pattern densities, such as linewidth generations G2 / G1 and product versions, the above etching rate formula must also include a compensation term for the pattern density. The etching rate of the corresponding etching process or etching reaction chamber can be expressed by the following formula 3: (Equation 3)

[0022] In Equation 3, ERt represents the etching rate of the product to be processed, ERt-1 represents the etching rate of the most recently shipped product group, and fPD represents the pattern density compensation factor, where fPD can be expressed by the following Equation 4: (Equation 4)

[0023] Where Dt is the pattern density of the product to be manufactured, Dt-1 is the pattern density of the most recently shipped product group, and αx is the weight of the pattern density compensation factor on the pattern density, as shown in Figure 3, which represents the slope of the etching rate against the pattern density. The αx values ​​for linewidth generations G2 / G1 are -12.3 and -10.5, respectively. As can be seen from Equations 3 and 4, in this embodiment, for RF power times less than 25 hours and for different product specifications and pattern densities, the etching rate of the current product gradually increases with increasing RF power time and decreasing pattern density.

[0024] On the other hand, in this embodiment of the invention, for the RF power time greater than or equal to 25 hours and for the same product, since the etching rate of the cavity has tended to stabilize as shown in Figure 2, and there are no variables such as product pattern density affecting it, the etching rate can be regarded as the average etching rate of all n batches of products, without batch distinction, as shown in Equation 5: (Equation 5)

[0025] For the RF power duration of 25 hours or more, and for different product specifications and pattern densities, the etching rate formula only needs to include a compensation factor for the product pattern density. The etching rate of the corresponding etching process or etching reaction chamber can be expressed by the following formula 6: (Equation 6)

[0026] Where ERt is the current product's etching rate, ERt-1 is the previous etching rate of the same group of products, and fPD is the pattern density compensation factor.

[0027] As can be seen from the descriptions of Equations 1 to 6 above, this invention specifically switches between different algorithms for different RF power time segments and different product specifications and pattern densities to conform to the actual etching process or etching rate of the etching reaction chamber. This allows the APC system to accurately find the optimal solution and take appropriate corresponding actions, thereby achieving process stability and maintaining product yield. It should be noted that the above embodiments are for illustrative purposes only. The actual etching rate formula may vary due to different etching machines, etching formulations, and / or products, and is not limited thereto. The following embodiments will specifically illustrate the control method flow of the system based on the above etching rate model.

[0028] Please now refer to Figure 4, which is a block diagram of the APC system's determination of the nonlinear cut-off point of the etching rate of an etching cavity according to an embodiment of the present invention. In the present invention, the nonlinear cut-off point can determine the RF power time point at which the etching rate of an etching cavity changes characteristics, such as the RF power time required for the etching rate to stabilize.

[0029] First, in step A1, the system retrieves the historical information of the product to be controlled by APC. This historical information may include process history and manufacturing data, such as batch number and manufacturing records, a unique identifier for each wafer or batch, including production line, process conditions, process recipes performed or planned, equipment usage status, or abnormal conditions. More specifically, the historical information must include the etch rate measured after the product has undergone the etching process in the corresponding etching machine / cavity, and the RF power time during the etching process in that etching machine / cavity. In one embodiment, the historical information of products shipped within 1-2 preventive maintenance (PM) cycles of an etching machine can be retrieved. This information can then be used to create a trend graph of etch rate versus RF power time, as shown in Figure 2, along with related curve fitting formulas.

[0030] Next, in step A2, a curve fitting formula for the etching rate versus RF power time is calculated based on the captured product history information. The choice of fitting formula may depend on the characteristics of the data. In this embodiment of the invention, based on the data shown in Figure 2, linear fitting is preferred but not limited to reflect the characteristics of the etching rate of an etching machine / cavity changing with RF power time.

[0031] After deriving the curve fitting formula for etching rate versus RF power time, step A3 involves calculating the slope between points on the curve based on this formula, and then calculating the difference in slope before and after each point. Since a larger difference in slope before and after a specific point on a curve indicates a significant change in the curve's characteristics at that point, the APC model can determine which point on the curve is the nonlinear tangent point based on these differences. In a preferred embodiment of the invention, this nonlinear tangent point is the point at which the etching rate of the etching machine / cavity tends to stabilize.

[0032] Finally, in step A4, the point with the largest difference in slope before and after is filtered out as the nonlinear tangent point of the RF power time in the APC control model. Taking Figure 2 as an example, the nonlinear tangent point of the process equipment / cavity is located at the time point when the RF power time is 25 hours. The etching rate of the etching equipment / cavity in the interval before this time point (i.e., the RF power time is less than 25 hours) will gradually increase from the lowest point and tend to stabilize in the subsequent RF power time. By determining the nonlinear tangent point of the RF power time, the APC system can adapt different etching rate formulas, model coefficients, weights, parameter values, and judgment rules to various characteristic segments of the RF power time of an etching cavity (such as the interval of a preventive maintenance cycle) to achieve the purpose of efficient APC control as intended by this invention.

[0033] Please now refer to Figure 5, which is a block diagram of the APC system's condition monitoring process for an etching machine / cavity according to an embodiment of the present invention. In this invention, this condition monitoring process can determine whether a machine can directly enter the APC control process after preventive maintenance.

[0034] First, in step B1, for an etching machine / cavity, a similarity analysis of the old and new etching rate versus RF power time curves is performed after preventative maintenance (PM) of the machine. In this embodiment, the old etching rate versus RF power time curve can be the etching rate versus RF power time curve of the machine before the PM, as obtained using the curve fitting formula obtained in step A2 above. The new etching rate versus RF power time curve can be the etching rate versus RF power time curve of the machine after the PM, and its data points can be obtained from the return test of the machine after PM or from the data of the shipped products. This similarity analysis can use any feasible analysis method, such as Euclidean distance or Pearson correlation coefficient, and is not limited thereto.

[0035] Next, in step B2, if the similarity analysis shows that the similarity between the old and new etching rates and the RF power time curves is greater than a set value, such as greater than 90%, then the two curves are considered to be roughly matched, and the machine can directly enter the subsequent APC control process (step B3). On the other hand, if the similarity analysis shows that the similarity between the old and new etching rates and the RF power time curves is not greater than a set value, such as not greater than 90%, then the two curves are considered to be mismatched. This reflects that the etching rate characteristics of the etching machine have changed significantly after PM, and the nonlinear tangent point judgment process shown in Figure 4 needs to be repeated to obtain a new trend graph of etching rate versus RF power time and the relevant curve fitting formula (step B4).

[0036] As can be understood from the above process, in the embodiments of the present invention, the fitting and similarity analysis of the etching rate to the RF power time curve can determine the condition of the etching machine / cavity in terms of etching rate, so that the APC control system can respond accordingly, such as re-judging the cut-off point or directly entering the subsequent APC control process. Therefore, it can avoid the condition or process variation of the etching machine / cavity caused by PM, and maintain the product yield and quality stability.

[0037] Please now refer to Figure 6, which is a flowchart of an APC control method according to an embodiment of the present invention. It will specifically explain the implementation details of the relevant steps and features of setting the etch rate model equation in the previous embodiment of the present invention.

[0038] First, in step C1, the system receives historical information about the etching machine / cavity. This historical information may include quality control related data, such as measurement and inspection data from the previous process of the product or machine (e.g., etching rate, linewidth, film thickness, electrical properties, etc.), analysis data after quality problems were discovered, and data related to the machine's previous production process control, such as parameter data of the previous process (e.g., temperature, pressure, gas flow rate, power, voltage / current, chemical composition, etc.), machine data of the previous process (e.g., RF power time, maintenance records, calibration data, etc.), and statistical quality control related data (e.g., process fluctuations, average values, standard deviations, etc.). In step C1, the system also receives the setting parameters for this process control, such as the formula, temperature, pressure, gas flow rate, processing steps and time, power, chemicals, operating conditions and modes of the process to be performed, for subsequent judgment.

[0039] Next, in step C2, the system inputs product information into the classifier model to retrieve the corresponding model coefficients, compensation factors, and their related weights. This product information may include product identification information (such as wafer number, product model, batch number, and product version) and process-related information (such as the process recipe, equipment number, process parameters, batch processing records, etc.). In this embodiment, the system can classify products into various product groups based on this product information. The model coefficients and weights of each product group are obtained by reading the product information of the batch wafers and finding the corresponding values ​​according to a pre-established numerical table, then substituting them into the formula for calculation. In the APC control method of this invention, product groups represent products with similar etch rate curves or characteristics, and these products can all be applied to the specific APC model, etch rate formula, and compensation factor proposed in this invention.

[0040] Next, in step C3, based on the received quality control data, it is determined whether the difference between the etching rate of the first batch of the product in the Pi Run test after preventive maintenance (PM) and the etching rate of the last batch of the product before preventive maintenance is greater than or equal to a set value (e.g., greater than or equal to 0.5 nm / s). If the etching rate difference is greater than or equal to the set value, it indicates that the etching rate of the etching machine has changed significantly after preventive maintenance, and it is necessary to determine the RF power time segment (step C9). On the other hand, if the etching rate difference is less than the set value, it indicates that the etching rate of the etching machine remains largely unchanged after preventive maintenance, and subsequent processes can directly refer to historical information of the same product or product group (step C5 or step C7).

[0041] When the difference in etching rate after preventative maintenance of the etching machine in step C3 is less than the set value, the process proceeds to step C4. The system checks whether there is historical information for the same product within a preset valid time period. This valid time period is, for example, 7 days. The historical information for this product includes the etching rate of the same product most recently shipped by the etching machine. If historical information for the same product is found within the preset valid time period, the parameter values ​​for the process to be performed can be directly calculated using this historical information (step C5). If no historical information for the same product is found within the preset valid time period, the system will further check whether there is historical information for a group of the same product within the preset valid time period (step C6).

[0042] When historical information of the same product is detected within a preset valid time in step C4, the process proceeds to step C5, where the system inputs the historical information of the same product into the APC model to calculate its parameter values. Specifically, in this embodiment, since the etching rate of the etching machine remains approximately unchanged after the PM returns (step C3), and there is historical information of the same product within the valid time, the etching rate of the machine in this step can be directly applied to the etching rate of the same product, that is, the average etching rate of the most recently shipped product of the same product, as shown in the following formula 7: (Equation 7)

[0043] On the other hand, if no historical information for the same product is found within the preset valid time in determination step C4, the process proceeds to determination step C6, where the system further checks whether there is historical information for the same product group within the preset valid time. If historical information for the same product group is found within the preset valid time, the parameter values ​​for the process to be performed can be directly calculated using this historical information (step C7). If no historical information for the same product group is found within the preset valid time, the system will notify the engineer to intervene manually (step C8).

[0044] When historical information of the same product group is detected within the preset valid time in step C6, the process proceeds to step C7, where the system inputs the historical information of the same product group into the APC model to calculate its parameter values. Specifically, in this embodiment, since the etching rate of the etching machine remains approximately unchanged after the PM returns (step C3), and there is historical information of the same product group within the valid time (step C6), the etching rate of the machine in this step can be directly applied to the etching rate of the same product group, that is, the average etching rate of one product in the most recently shipped product group, and a compensation factor fPD (Equation 4) for different product pattern densities is added, as shown in Equation 8 below: (Equation 8)

[0045] When step C6 detects no historical information for the same product group within the preset valid time, the process proceeds to step C8. Since there is no historical information for the same product or product group as a reference, the system will notify the engineer to intervene manually. For example, the engineer can manually input the etching rate parameter value for the product based on experience, or conduct a controlled trial production of the product to confirm that the relevant quality control tests are satisfactory before proceeding with the shipment.

[0046] Returning to decision step C9, when the difference in etching rate before and after PM on the etching machine is greater than or equal to a set value, the system will further determine whether the current RF power time of the machine is less than the nonlinear cutoff point (e.g., 25 hours) obtained in step A4. If the system determines that the current RF power time of the machine is less than the nonlinear cutoff point, it means that the current etching rate of the machine is not yet stable, and the subsequent APC model needs to consider RF power time compensation. The process will then proceed to decision step C10 to search for product historical data. On the other hand, if the system determines that the current RF power time of the machine is not less than the nonlinear cutoff point, it means that the current etching rate of the machine is stabilizing, and the subsequent APC model does not need to consider RF power time compensation. The process will then proceed to the aforementioned decision step C4 to search for relevant product historical data.

[0047] When the system determines in step C9 that the current RF power time of the etching machine is less than the nonlinear cut-off point, the process proceeds to step C10. The system checks whether there is historical information for the same product within a preset valid time period. This valid time period is, for example, 7 days. The historical information for the product may include the etching rate of the same product most recently shipped by the etching machine. If historical information for the same product is found within the preset valid time period, the parameter values ​​of the process to be performed can be directly calculated using the historical information for the same product, and then RF power time compensation is added (step C11). If there is no historical information for the same product within the preset valid time period, the system will further check whether there is historical information for the same product group within the preset valid time period (step C12).

[0048] When historical information of the same product is detected within the preset valid time in step C10, the process proceeds to step C11, where the system inputs the historical information of the same product into the APC model to calculate its parameter values. Specifically, in this embodiment, since the etching rate of the etching machine changes significantly after the PM returns (step C3), and the etching rate of the machine has not yet stabilized (i.e., step C9, the RF power time of the machine is less than the nonlinear tangent point), and there is historical information of the same product within the valid time, the etching rate of the machine in this step can be directly applied to the etching rate of the same product, that is, the etching rate of the most recently shipped product of the same product (ERt-1), and then the compensation factor fEQ of the RF power time (Equation 2) is added, as shown in Equation 9 below: (Equation 9)

[0049] On the other hand, if no historical information for the same product is found within the preset valid time in determination step C10, the process proceeds to determination step C12, where the system further checks whether there is historical information for the same product group within the preset valid time. If historical information for the same product group is found within the preset valid time, the parameter values ​​for the process to be performed can be directly calculated using this historical information, and then compensation for RF power time and different product pattern densities is added (step C13). If no historical information for the same product group is found within the preset valid time, the system will notify the engineer to intervene manually (step C8).

[0050] When historical information of the same product group is detected within the preset valid time in step C12, the process proceeds to step C13, where the system inputs the historical information of the same product group into the APC model to calculate its parameter values. Specifically, in this embodiment, since the etching rate of the etching machine changes significantly after the PM returns (step C3), and the etching rate of the machine has not yet stabilized (i.e., step C9, the RF power time of the machine is less than the nonlinear tangent point), and there is historical information of the same product group within the valid time, the etching rate of the machine in this step can be directly applied to the etching rate of the same product group, that is, the etching rate of the most recently shipped product group (ERt-1), and then the compensation factor fPD (Equation 4) for different product pattern densities and the compensation factor fEQ (Equation 2) for RF power time are added, as shown in the following Equation 10: (Equation 10)

[0051] When step C12 detects no historical information for the same product group within the preset valid time, the process proceeds to step C8. Since there is no historical information for the same product or product group as a reference, the system will notify the engineer to intervene manually. For example, the engineer can manually input the etching rate parameter value for the product based on experience, or conduct a controlled trial production of the product to confirm that the relevant quality control tests are satisfactory before proceeding with the shipment. According to the above embodiments of the present invention, it can be understood that the present invention determines whether the machine condition has changed by comparing the similarity of the etching rate curves before and after machine maintenance, and finds the nonlinear cut point of the product or product group based on these etching rate curves. Then, the target etching rate of the current product can be determined based on the nonlinear cut point, the RF power time of the machine / cavity, and the pattern density of the product. In this way, process control can be carried out across products according to different machine conditions, and the product yield and quality stability can be maintained. There is no need for frequent intervention by engineers after machine maintenance and for trial production of various products. This can shorten the production cycle and cost of the product, which is a novel and progressive invention. The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made in accordance with the claims of the present invention shall be covered by the present invention.

[0052] 10: Device 11: Wafer 12: Computer 13: Etching machine 14: Measuring equipment 21: Processor 22: Memory 23: Program 36: Process characteristics 46: Advanced Process Control (APC) Module 51: Approach 52: Approach 61: Real-time estimation module 66: Approach A1~A4: Steps B1~B4: Steps C1~C13: Steps

Claims

1. An advanced process control method, comprising: providing a process cavity; reading product history information of the process cavity within a time interval, the product history information including the etch rate of various products and product groups in an etch process in the process cavity and the corresponding radio frequency power time of the process cavity; deriving an etch rate curve of the etch rate of each product and product group relative to the radio frequency power time based on the product history information, and finding a nonlinear tangent point of the etch rate curve; providing a process product; and reading information of the process product and obtaining coefficients and weights of an etch rate model equation applicable to the process product, wherein the step of setting the etch rate model equation includes: determining whether the difference in the etch rate of the process cavity before and after preventive maintenance is greater than or equal to a first set value; determining whether the current radio frequency power time of the process cavity is less than the nonlinear tangent point; and determining whether there is product history information of the same product or product group within a valid time period.

2. The advanced process control method as described in claim 1, wherein the step of finding the nonlinear tangent point of the etch rate curve comprises: calculating the slope between each data point in the etch rate curve and the slope difference before and after the data points; and finding the data point with the largest slope difference as the nonlinear tangent point of the etch rate curve.

3. The advanced process control method as described in claim 2 further includes performing a similarity analysis on the etching rate curves before and after a preventive maintenance of the process cavity, and when the obtained similarity is less than or equal to a second set value, re-performing the step of finding the nonlinear tangent point.

4. The advanced process control method as described in claim 1, wherein when the difference in etch rate before and after preventive maintenance is less than the first set value, or the current RF power time of the process cavity is greater than or equal to the nonlinear cut point, the etch rate model equation of the process product can be set as the average etch rate of the same product: , where is the target etch rate of the process product, and is the etch rate of the most recently shipped product of the same product.

5. The advanced process control method as described in claim 1, wherein when the difference in etch rate before and after preventive maintenance is less than the first set value, or the current RF power time of the process cavity is greater than or equal to the nonlinear tangent point, the etch rate model equation of the process product can be set as the average etch rate of the same product group plus a pattern density compensation factor: , where is the target etch rate of the process product, is the etch rate of the most recently shipped product group, is the pattern density compensation factor, is the pattern density of the process product, is the pattern density of the most recently shipped product group, and is the relevant weight of the pattern density compensation factor on the pattern density.

6. The advanced process control method as described in claim 1, wherein when the difference in etch rate before and after preventive maintenance is greater than or equal to the first set value, and the current RF power time of the process cavity is less than the nonlinear tangent point, the etch rate model equation of the process product can be set as the etch rate of the same product plus an RF power time compensation factor: , where is the target etch rate of the process product, is the etch rate of the most recently shipped same product, is the RF power time compensation factor, is the current RF power time of the process cavity, is the RF power time of the etched cavity when the same product was most recently shipped to the process cavity, and is the relevant weight of the RF power time compensation factor for the etch rate.

7. The advanced process control method as described in claim 1, wherein when the difference in etch rate before and after preventive maintenance is greater than or equal to the first set value, and the current RF power time of the process cavity is less than the nonlinear tangent point, the etch rate model equation of the process product can be set as the etch rate of the same product group plus a pattern density compensation factor and an RF power time compensation factor: , , where is the target etch rate of the process product, is the etch rate of the most recently shipped product group of the same product group, is the pattern density compensation factor, is the RF power time compensation factor, is the pattern density of the process product, is the pattern density of the most recently shipped product group of the same product group, is the weight of the pattern density compensation factor on the pattern density, is the current RF power time of the process cavity, is the RF power time of the etch cavity when the same product group was most recently shipped to the process cavity, and is the weight of the RF power time compensation factor on the etch rate.

8. An advanced process control system for controlling a process cavity, comprising: A memory for storing product history information of the process cavity within a time interval, the product history information including etch rate curves of various products and product groups undergoing an etching process in the process cavity, and the pattern density of these products and product groups, wherein the etch rate curve is a curve showing the relationship between the etch rate of each product and product group and the corresponding RF power time of the process cavity; and an advanced process control module for controlling the process cavity according to a plurality of instructions, the instructions being used to perform the following steps: calculating the slope between each data point in the etch rate curve and the slope difference before and after these data points; finding the data point with the largest slope difference as a non-linear tangent point; and controlling the process cavity to process a wafer according to the RF power time, the non-linear tangent point, and the pattern density.

Images