TWI297506B - Method and system for virtual metrology prediction for quality control in semiconductor manufacture - Google Patents

Description

1297506 IX. Description of the Invention: TECHNICAL FIELD OF THE INVENTION The present invention relates to a quality control method for semiconductor manufacturing, and more particularly to a virtual measurement estimation method and system for quality control of semiconductor manufacturing. [Prior Art] Statistical Quality Control (SQC) is a 'technology to maintain and improve product quality, and Statistical Process Control (SPC) is one of the main tools, it focuses on Analysis of data during the manufacturing process to determine the cause of product variation. Statistical quality control consists of two main components, statistical process control and sampling acceptance criteria. Statistical process control includes the basic theory of quality control (QC) processing and probability and statistics and its applications. Spc is the use of process operation variables to predict the production variables or the quality variables of the product • Sexual monitoring, and from the process operation variables to the safety / quality problems, there is a certain time behind, so how can be in the shortest The problem of predicting quality variables in time is one of the important factors to consider when evaluating the advantages and disadvantages of spp related methods. In addition, the product allows for a number of differences in the production process, but the difference is more appropriate to control, and the degree of quality is f (f), so that it can be driven within the scope of =. The quality control (QC) is based on the manufacture of this, the measurement of the number. The statistical analysis of the e-graphics of the official system of whether an abnormal phenomenon, or from the big buckle

Client's Docket No· : PT.AP-663 TT^s Docket No : 0532-A40826-TW / Final / Alex Chen 1297506 In the product, several samples are taken, the characteristics are checked, and the data are analyzed to determine whether the product is qualified or not. Need to be disposed of. Traditional measurement methods may lead to errors in judgment due to incorrect or incomplete information collected. Therefore, the present invention proposes a virtual measurement prediction method applied to semiconductor manufacturing quality control, combined with real-time statistical process control (SPC) and data storage system to effectively control the virtual measurement prediction mode ( The effectiveness of the Virtual Metrology Prediction Model) avoids the loss of yield caused by other related systems (such as the control system) or mechanisms (for example, the Hold Lot). SUMMARY OF THE INVENTION Based on the above objects, an embodiment of the present invention discloses a virtual measurement estimation system, including a process machine, a measurement machine, a manufacturing execution system, a machine control system, and a machine equipment automation. Program, < failure detection and eight types of systems, a virtual measurement engine, a data repository, and a control module. • The process machine performs a process on a wafer in batches. The measuring machine 斟-曰 ^ 曰曰 circle performs a measuring operation in batches and generates an actual measured value. The manufacturing execution system takes the actual measurement from the measuring machine. The machine control system and the machine equipment automation program sample the wafer lot according to a predetermined sampling frequency. The failure detection and classification system generates failure detection and classification data after the wafer is completed in batches. The virtual measurement engine obtains the failure detection and classification data from the failure value measurement and classification system, and generates an estimated measurement value accordingly. The data storage obtains the actual anti-measurement value from the manufacturing execution system and obtains the estimated measurement value from the virtual measurement engine, and

Client's Docket No. : ΡΤ.ΑΡ-663 N TT5s Docket No : 0532-A40826-TW / Final /Alex Chen 1297506 The estimated value performs a mathematical operation with an actual measured value to obtain a residual value' and The residual value is transmitted to the control module. The control module obtains the estimated measurement value from the seven virtual measurement engine and obtains the residual value from the data storage, and determines whether the residual value is excessively large according to the real-time statistical process control rule, and if so, A first control command is issued to stop the operation of the virtual measurement model of one of the virtual measurement engines. The embodiment of the invention further discloses a virtual measurement estimation method. Develop a sampling frequency for a wafer batch. When the batch of wafers is processed in a process machine, the wafer is sampled using a machine control system and a machine equipment automation program. When the batch of wafers completes the process, the failure detection and the type of negative materials are transmitted to a virtual test engine. The virtual 1 test engine generates an estimated measurement value based on the current failure detection and classification data and transmits it to a data storage. A mathematical operation is performed on the estimated measured value and an actual measured value to obtain a residual value, and the residual value is transmitted to a control module. The control module determines whether the residual value has changed excessively according to the real-time statistical process control rules. Right is 'and then a brother-and-control command is issued to stop the operation of the virtual measurement model of the virtual volume. [Embodiment] The objects, features, and advantages of the present invention will become more apparent from the following detailed description of the preferred embodiments. The present specification provides various embodiments to illustrate the technical features of various embodiments of the present invention. The arrangement of the various elements in the embodiments is for illustrative purposes and is not intended to limit the invention. In the embodiment, the parts of the drawings are repeated for the sake of simplicity of explanation, and do not mean different realities.

Client’s Docket No. : PT.AP-663 XT’s Docket No : 0532-A40826-TW / Final / Alex Chen 1297506 The correlation between the examples. Embodiments of the present invention disclose a virtual measurement and prediction method and system for quality control of semiconductor manufacturing. In order to obtain detailed process performance data, a large number of high-resolution system variables are required, that is, the status variable identification (SVID) of each machine, for example, temperature, pressure, Gas flow I and so on. Using a failure detection and classification (Fault Detection and Classification, hereinafter referred to as fdc) • The system collects the SVID of each machine (that is, FDC data), and then transmits the obtained FDC data to the present invention. The virtual metrology engine (Virtual Metrology Engine) is processed to produce virtual measurement results, as shown in Figure 1. Virtual Metrology (VM) is the application of Computational Intelligence to instantly estimate the quality of the items processed on the production machine. If the accuracy of the estimation is in line with the requirements of Advanced Process/Equipment Control (APC/AEC), this virtual measurement can replace the traditional measurement machine. Therefore, the virtual measurement engine of the present invention is a system with a built-in estimation model, and when the FDC data is input into the virtual measurement engine, a virtual measurement result can be generated. The estimated model includes (which is used to estimate the film thickness 1 = 1 (Thickness)), 〇〇 = / (~12, "., ', Xiao) (which is used to estimate the line width (Critical Dimension, CD)) and so on, where Xi is represented as SVID and ADI is after development development wafer inspection (After Development Inspection)

Client’s Docket No· : ΡΤ.ΑΡ-663 TT5s Docket No : 0532-A40826-TW / Final / Alex Chen 9 1297506 Figure 2 shows the architecture of a traditional virtual measurement and estimation system. The conventional virtual measurement and prediction system 200 includes a manufacturing execution system (MES) 210, a machine control system, and a control system (Taiwan). 220, a control system/device automatic program (TCS/EAP) Process machine 230, a measurement machine 240, an FDC system 250, a virtual measurement engine 260, and a batch control system 270.

The machine control system and the machine equipment automation program 220 transmit the relevant data of the process machine 230 and the measurement machine 240 to the FDC system 250, and the FDC system 250 collects the machine data to form failure failure detection and classification data. • Then transferred to the virtual measurement engine 260. After the virtual measurement engine 260 analyzes the fail-over detection and classification data, it can generate virtual quality control (QC) data and feed back to the batch control system 270 to perform subsequent processing, or according to the prediction result. A standard value is generated, and when the standard value is exceeded during execution of the process, a pause batch command is issued to the manufacturing execution system 210 to suspend processing of the current wafer lot. However, the 'conventional' machine's set value may change after the Maintenance Maintenance (PM). For example, the original temperature is set at 250 ° C and may become 260 ° C or 270 ° C after maintenance, causing the virtual measurement engine 260 to fail to operate or the predicted result is inaccurate because it does not have the correct set value. Therefore, the present invention proposes a new architecture to avoid the above problems. Fig. 3 is a schematic view showing the architecture of the virtual measurement estimation system of the embodiment of the present invention.

Client's Docket No. · PT.AP-663 TT^ Docket No : 0532-A40826-TW/Final/Alex Chen 10 1297506 The virtual measurement estimation system 300 of the embodiment of the present invention includes a manufacturing execution system 310 and a machine Control system and machine equipment automation program 320, a process machine 330, a measurement machine 34, an FDC system 350, a virtual measurement engine 360, a data warehouse 380, and an online real-time (RT) statistics The process control monitoring application technology module (hereinafter referred to as RT-SPC module) 390, wherein the thick solid line indicates the wafer-related control operation, and the broken line indicates the control related to the virtual measurement model (Virtuai Metrol〇gy Model). operating. The difference from the virtual measurement estimation system 200 is that when the virtual measurement engine 360 obtains the FDC data (FDCD) from the FDC system 350, the estimated measurement value generated based on the FDC data (ie, the virtual qC data (Virtual) QC Data 'VQCD)) is simultaneously transmitted to the data repository 380 and the RT-SPC module 390. The RT-SPC module 390 determines whether to suspend the batch based on the estimated measurement value and the instantaneous statistical process control (RT-SPC) and the rule. If so, the RT-SPC module 390 issues a suspend batch command to the manufacturing execution system 310 to suspend processing of the current wafer lot. In order to prevent the occurrence of inaccurate predictions in the virtual measurement and prediction system, the following prevention mechanisms are provided. First, the virtual s-measurement model is to develop the sampling frequency of the wafer. When the wafer is in the process of the processing machine 330, the wafer is sampled by the machine control system and the machine equipment automation program 320, and then the fdc data is transmitted by the FDC system 35 to the virtual measurement during the wafer completion process. The engine 360° virtual measurement engine 36 generates an estimated measurement (ie, VQCD) based on the fDc data and transmits it to the data repository 38〇. on the other hand,

Clienfs Docket No. : PT.AP-663 TT?s Docket No : 0532-A40826-TW/Final/Alex Chen 11 1297506 After the wafer is measured by the measuring machine 340, the measuring machine 340 will measure the actual amount. The measurements are transmitted to the data repository 38 via the manufacturing execution system 310. Data warehousing 380 provides the function of Data Mart, which subtracts the estimated measured value from the actual measured value (ie, performs a mathematical or logical operation), and the result of the subtraction is the residual value (Residual). And the residual value is transmitted to the RT_SPC module 39Q. The RT_SPC module 390 determines whether the residual value has changed excessively according to the RT-SPC rule. If so, a control command is issued to the virtual measurement engine 360 to stop the operation of the virtual measurement model, and another control command is issued to the manufacturing execution system 310 to return the sampling frequency to the initial value. In addition, if the wafer is not sampled, the control operations associated with the virtual measurement model will not be performed. The definition of the RT_SPC rule in the embodiment of the present invention is as follows. Referring to Annex 1 and Annex 2, an embodiment of the present invention defines an upper control limit (UCL) and a lower control limit (Lower c〇ntr〇1).

Limit’ LCL), when the control point (Control P〇int) is not between the upper control limit and the lower control limit, indicating that the residual value is too large and the estimated measurement value is completely inaccurate, an alarm will be issued. Referring to Annex 1, when the bounce of two of the three control points exceeds one sample (Sigma, Shift GAP), the change in the residual value is too large, so a warning is issued. Referring to Annex 2, when the eight control points are on the same side of the control limit (c〇ntr〇1 Limit, CL), it indicates that the estimated measured value is always higher or lower than the actual measured value, so a warning will be issued. . Figure 4 is a schematic view showing a virtual measurement estimation system according to another embodiment of the present invention.

Client's Docket No. : ΡΤ.ΑΡ-663 TT,s Docket No : 0532-A40826-TW / Final / Alex Chen 12 1297506 & Compared with Figure 3, Figure 4 is more than one person - advanced process control disk batch - The person controls the line 400. #Advanced process control and batch control system for precise batch control management of wafers. Figure 5 is a flow chart showing the steps of the virtual measurement estimation method according to the embodiment of the present invention. ^ One hundred first, develop a sampling frequency of one wafer batch (step SU). When the batch of wafers is in the process of executing the process, the batch of wafers is sampled by a machine control system Luzer and the machine equipment automation program (step / Sl2), and then the process is completed in the batch of wafers. The FDC data is transmitted to a virtual measurement engine (step S13). The virtual measurement engine generates an estimated measurement value based on the fdc data and transmits it to a data repository and an RT_spc module (step S14). On the other hand, the batch of wafers produces an actual measured value after the measurement is completed on a measuring machine (step S21), and the actual measured value is transmitted to the data storage (step S22). Calculating the estimated measured value by subtracting the actual measured value from the one of the data storage and storing the estimated measured value to take a residual value (step S31), and transmitting the residual value to the RT- SPC module (step S32). The rt_spc module determines whether the residual value has changed excessively according to the RT_spc rule (step S33). If yes, a control command is issued to the virtual measurement engine to stop the operation of the virtual measurement model (step S34), and another control command is issued to a manufacturing execution system to restore the sampling frequency (Sampling Rate). It becomes an initial value (step S35). In addition, if the wafer is not sampled, the control operations associated with the virtual measurement model are not performed. On the other hand, RT-SPC modules are based on estimated measurements and RT_spc gauges.

Client’s Docket No. : PT.AP-663 TT^ Docket No : 0532-A40826-TW / Final / Alex Chen 1297506

Then, it is judged whether or not the batch is to be suspended (step S41). If so, the RT-SPC loose group issues a pause batch command to the manufacturing execution system (step S42), and the manufacturing execution system suspends processing of the current wafer lot (step S43). The virtual 1 measurement and prediction method applied to the quality control of semiconductor manufacturing according to the embodiment of the present invention, combined with the real-time statistical process control and data storage system, uses the residual value as the control object, and cooperates with the spc rule to effectively control the virtual measurement estimation. The effectiveness of the model avoids the loss of yield due to misjudgment by other related systems (eg, batch control systems) or mechanisms (eg, pause batches). Although the present invention has been disclosed in the above preferred embodiments, it is not intended to limit the invention, and any person skilled in the art can make various changes and movements within the scope of the present invention. The scope defined in the patent application is subject to change. Show 5 only [Simplified illustration of the diagram] Figure 1 shows a schematic diagram of the virtual measurement engine. Brother 2 shows the point of the traditional virtual measurement and prediction system. 4: The figure shows the virtual measuring frame of the embodiment of the present invention. FIG. 4 is a schematic view of the structure of the present invention. 〖Predicting the system Figure 5 is a flow chart showing the steps of the virtual quantity pre-implementation in the embodiment of the present invention. The method of the top estimation method [Main component symbol description]

Client's Docket No. : PT.AP-663 TT's Docket No : 0532-A40826-TW/Final/Alex Chen 14 1297506 200, 300~Virtual Engine Estimation System 210, 310~ Manufacturing Execution System 220, 320~ Machine Control System And machine equipment automation programs 230, 330 to process machine 240, 340 - measuring machine 250, 350 ~ failure detection and classification system 260, 360 ~ virtual measurement engine 270 ~ batch control system 380 ~ data storage 390 ~ Online real-time statistical process control monitoring application technology module 400 ~ advanced process control and batch control system

Client’s Docket No. : PT.AP-663 TT^ Docket No : 0532-A40826-TW/Final/Alex Chen

Claims (1)

1297506 X. Patent application scope: 1. A virtual measurement and estimation system, comprising: a processing machine, which executes a process in batches on a wafer; a measuring machine, which performs a measurement on a wafer in batches Operating and generating an actual measured value; a manufacturing execution system that obtains the actual measured value from the measuring machine; a machine control system and a machine equipment automation program that processes the crystal according to a predetermined sampling frequency Sampling in a batch; a failure detection and classification system that generates failure detection and classification data after the wafer is batch-processed; a virtual measurement engine that obtains from the failure detection and classification system Failure detection and classification data, and based on which a predicted measurement value is generated; a data storage, which obtains the actual measurement value from the manufacturing execution system and obtains the estimated measurement value from the virtual measurement engine, Performing a mathematical operation on the estimated measured value and the actual measured value to obtain a residual value; and a control module, obtaining the residual value from the data storage, according to the instant The statistical process control rule determines whether the residual value has changed too much, and if so, issues a first control command to stop the operation of the virtual measurement model of the virtual measurement engine. 2. The virtual measurement estimation system according to claim 1, wherein the control module further issues a second control command to the manufacturing execution system to restore the sampling frequency to an initial value. 3. For the virtual measurement estimation system described in item 2 of the patent application, Client's Docket No. : PT.AP-663 TT^ Docket No : 0532-A40826-TW / Final / Alex Chen ! 6 1297506 where The control module determines whether to suspend the wafer batch according to the estimated measurement value and the instant statistical process control rule, and if so, issues a pause batch command to the manufacturing execution system to suspend the wafer batch - Reason. 4. The virtual measurement estimation system according to claim 3, wherein the control module is an on-line real-time statistical process control monitoring application technology module. 5. If the virtual measurement prediction system described in item 1 of the patent application is applied, • the data storage uses a data supermarket function to subtract the estimated measurement value from the actual measurement value to obtain the residue. Difference. 6. A virtual measurement estimation method, comprising the steps of: formulating a sampling frequency of a wafer batch; and using a machine control system and machine equipment automation when the batch of wafers executes a process on a processing machine The program samples the batch of wafers; transmits the fail-fail detection and classification data to a virtual measurement engine when the batch of wafers completes the process, and measures the batch of wafers by using a measuring machine* to generate An actual measurement value is transmitted to the data storage, and the virtual placement engine generates an estimated measurement value according to the failure failure detection and classification tribute and transmits the data to a data storage; Performing a mathematical operation on the estimated measured value and the actual measured value to obtain a residual value; transmitting the residual value to a control module; the control module determining whether the residual value is based on an instant statistical process control rule Oversized; Clienfs Docket No. : PT.AP-663 TT^ Docket No : 0532-A40826-TW/Final/Alex Chen 17 1297506 If yes, issue a first control command to stop one of the virtual measurement engines Intended to measure the amount of operation of the model. 7. The virtual measurement estimation method of claim 6, further comprising issuing a second control command to the manufacturing execution system to restore the sampling frequency to an initial value. 8. The virtual measurement estimation method according to claim 7, further comprising the following steps: the control module determines whether to suspend the wafer according to the estimated measurement value and the instant statistical process control rule Batch; if so, the regulatory module issues a pause batch command to the manufacturing execution system; and the manufacturing execution system suspends processing of the wafer batch. 9. The virtual measurement estimation method according to item 8 of the patent application scope, wherein the control module is an on-line real-time statistical process control monitoring application technology module. 10. The virtual measurement estimation method according to claim 6, wherein the data storage uses a data supermarket function to subtract the estimated measurement value from the actual measurement value to obtain the residual value. . Client’s Docket No. ·· PT.AP-663 TT?s Docket No : 0532-A40826-TW /Final / Alex Chen 18