Advanced Process Control (APC) techniques have matured in the semiconductor manufacturing facilities over the years. The availability and use of custom and off-the-shelf APC framework in a fab is taken for granted, as the devices and manufacturing techniques get more complex than ever before. This information is provided to communicate a high level overview of APC, with classic examples to communicate potential and the infrastructure to nurture APC.
APC system refers to the use of advanced statistical and analytical techniques to manipulate process control parameters and inputs on process tools to improve output quality. It is algorithm-based, and indicates the adjustments necessary (from one lot to the next) to reduce systematic variation. APC system adjustments focus on
(1) Feedforward: Data from preceding operations is utilized to adjust processing at subsequent operations.
(2) Feedback: Data from material already processed is used to adjust prior operations input parameters.
Primary components of an APC system:
- Tool to be controlled
- Tool Monitor/Controls
- Inspection Tool
- APC System
- APC Database
- MES System
In a generic APC system, (1) an inspection tool measures predefined inspection points on a processed wafer and feeds the results a database. The APC system (2) analyzes the results and (3) decides whether Recipe A, that processed the inspected wafer, needs to be modified to Recipe A+, to compensate for errors detected after analysis inspection results. The Recipe A+ is the result of changing the input parameters, is then (4) downloaded to the process tool. The APC engine is the workhorse of the system, analyzing output parameters, and recommending input parameter changes to the recipe, when it detects actionable signals from the inspection results. Also, there is a database associated with the system to store the large amount of data generated by the inspection tools, and for the storage of data models, versioning, etc.
Today, an APC framework is essential for rapid implementation and scaling ability to install many APC applications across multiple tools in the factory. A framework allows for a common data format, scripting language, standardized helpful algorithms, and a database that is accessible to factory-wide controllers and tools. Some companies choose to develop a custom APC framework. But, there are many APC framework solutions available in the market these days. Frameworks have assisted in the proliferation of APC systems, coupled with rapid application development and implementation with lower costs and faster realization of an expected benefit, which is to maximize revenue from every wafer processed.
SEMI – (Semiconductor Equipment and Materials International) consortium, which provides the electronics industry with stewardship and guidance, has issued the SEMI E133 – The Process Control System Standards that among other things defines communication among the components to enable run-to-run (R2R) control, fault detection (FD), fault classification (FC), fault prediction (FP), and statistical process control (SPC). It is supported by SEMI E125 and E134 on EDA (Equipment Data Acquisition) specifications that enable and improve communication between a factory’s data gathering software applications, and process and inspection tools. EDA is very important in the Photolithography and Plasma Etch modules that traditionally generate the most volume of data needed for APC system implementations.
Applications of APC:
- Photolithography: Making dynamic changes to photolithography scanner parameters, such as exposure time and dosage to maintain or improve key output parameters such as develop ‘check critical dimensions’ that are measured after plasma etching is performed on the wafer.
- CD control: Critical Dimension control is necessary for the fine patterns formed on a wafer, and this is usually measured by a CD-SEM Scanning Electron Microscope metrology tool. APC is used to measure and evaluate whether the shape and dimensions of trench formed after the photolithography and etch process is as per specifications. When deviation signals are detected by the APC engine, tool control data is fed-forward to the scanner/track equipment processing the wafer, to eliminate the deviations.
- Planarization CMP processing: To control quality of deposition (thickness, flatness, etc.) that was achieved, recipe changes are made to planarization parameters dynamically. This ensures that the CMP tool removes the right amount of deposit on each wafer, to provide a consistent and uniform flatness of the wafer surface, to the next process step.
Like other disciplines Process Control has its own terminology, and the entire scope of advanced process control in manufacturing usually involves all the aspects below.
Run-to-Run Control (R2R) describes the scenario, where in-between processing of two lots or two wafers of the same lot, an advanced process control system can implement changes in recipe parameters. The system makes continuous adjustments if necessary at the end of each processing and measurement cycle, before repeating the next cycle, and thereby minimizing drift.
Fault Detection and Classification (FDC) describes the scenario where the factory monitors data on the manufacturing tool, while simultaneously using analytical techniques to detect tool faults. And once a fault is detected, then it is classified to determine an assignable cause. This cause is then corrected before the tool is released for further processing. Fault detection and classification has another branch that is called Fault Prediction and this involves using the same data to identify parameters that can be monitored to flag the need for tool maintenance activity ahead of when the fault manifests itself.
Models: Tool processing is described in a model that is specific to its activity. The model describes all the controllable input parameters and measurable output parameters that are statistically linked in mathematical terms that reflects a desired run.
Hence, when specific measured output parameter deviates from the desired values, the related input parameters are changed to bring the process back under control. For temperature-based thermal profiles, a ‘golden’ time-series temperature profile is compared against each run to detect deviation. Advanced process control relies on mathematics, statistics, and machine learning. Commonality analysis, multivariate analysis, linear programming, time-series, and machine learning, are used extensively in advanced process control.
Leveraging APC Technology: This requires the manufacturing facility to implement an infrastructure that encourages the generation of APC project ideas, and facilitates evaluation and final implementation. Education and training augmented with a proper APC application framework, designated Process Control Engineers, user-groups, and an APC management body, are essential. This will ensure that feasibility is ascertained, and RoI is scrutinized before starting implementation.
APC Health Monitoring: In a dynamic manufacturing environment, tool upgrades and changes are constantly done both physically and at a software level. Just as SPC health monitoring is done regularly, APC health monitoring is also required by the identification and monitoring of appropriate parameters. Monthly reviews must be implemented to look at indicators such as Cpk, data integrity, number of interventions, control lag time, etc.
Contemporary trends in APC: There are several topics that dominate the discipline these days in the APC environment. They include (1) Big data and data mining, (2) Data analytics, (3) Data management, (4) Fault detection and classification, (5) Metrology and virtual metrology, (6) Model-based control, (7) Predictive technologies (8) Run2Run control (9) Sensors, and (10) Standards.
Increasing, factories are faced with the special challenges of High Mix Low Volume loadings involving shorter production runs and frequent changeovers, advanced process control strategies are more relevant than ever before. Falling device margins also needs to be addressed and advanced process control techniques helps factories improve output and quality and thereby remain profitable and competitive.