The semiconductor industry is entering its golden decade. Who can help domestic wafer factories seize the opportunity?

The CIM system is known as the "central brain" of wafer fab manufacturing. It integrates a series of key systems such as the production execution system MES, equipment management system EAP, statistical process control SPC, advanced process control APC, fault detection and classification FDC, and yield management system YMS. It runs through the key links of chip production execution, production operation and production control, and manages equipment and factory personnel. Since the development of the industry, the complexity of chip production processes has increased exponentially, and the process has been continuously extended, driving wafer fabs to continuously pursue technological breakthroughs and innovations in advanced processes. This also requires the CIM system to be continuously upgraded and iterated to meet and support the rapid growth of the industry.

According to the current process, the production process of a chip requires at least hundreds or even thousands of processes and hundreds of equipment. Each device can generate TB-level massive production data in just 24 hours of operation. This requires the production and manufacturing system to be able to cope with and process massive and complex data, and provide production control and support for the increasingly stringent wafer manufacturing needs. Therefore, the CIM system must first have high stability and high concurrency processing capabilities. More importantly, the system development throughout the entire production process also requires rich semiconductor know-how.

"With the increase in process complexity, wafer fabs are facing huge challenges in production management and process control. This puts higher requirements on the intelligence of CIM systems and the refinement of system functions, and the value of semiconductor CIM systems is more prominent. Taking Geely as an example, the new generation of Industry 4.0 smart factories we have built for the semiconductor industry use intelligent software to make equipment smarter and process control more sophisticated, breaking through the bottleneck of traditional automated production lines, helping mature process production lines to increase production capacity and reduce production time. For advanced processes, it can also improve product performance, optimize production and process design, etc." Xiao Changbao, chief architect of Geely and head of semiconductor software product development at Geely, told Semiconductor Industry Observer.

"GeChuang Dongzhi has been selected and certified by the Ministry of Industry and Information Technology as the "2022 Cross-Industry and Cross-Field Industrial Internet Platform". As a dual cross-industry platform originating from the semiconductor manufacturing industry in China, GeChuang Dongzhi is not only good at industrial software and system research and development in cutting-edge manufacturing industries such as semiconductors, but also one of the few industry leaders in China to provide full-stack factory integration services. The services cover digital consulting and solution design for the entire factory and multiple factories, information implementation and delivery, and subsequent operation and maintenance. It also builds a factory production system for the semiconductor industry based on a new perspective of "production-analysis-prediction". Its independently developed CIM has industry-leading core technology strength and enabling capabilities.

For wafer fabs, whether CIM, which plays a vital role, can be successfully implemented depends not only on the technical strength and maturity of the software itself, but also on the rich and comprehensive overall planning, design and implementation experience of CIM manufacturers. Especially today, when the upstream and downstream production cooperation in the industry is becoming increasingly close and the new generation of CIM even includes as many as dozens of complex systems, the "stitching" model that only focuses on the single-point requirements of MES or a few subsystems and gradually supplements and wedges into other systems or even external systems is increasingly showing the core defects of its commercial suite with high functional duplication and heavy system integration workload.

Therefore, the whole-factory capability - whether the service provider has a self-developed full-stack solution with leading technology and stable performance, and is familiar with how to perfectly connect and integrate dozens of systems with different functions and interrelated functions through forward-looking planning and precise implementation in the whole-factory, multiple-factory and even upstream and downstream collaboration of the industrial chain, has become the cornerstone for the successful implementation of CIM in wafer fabs in the next decade.

As one of the few domestic manufacturers whose self-developed products have been implemented in the upstream, midstream and downstream of the industrial chain, Gechuang Dongzhi's semiconductor customers include semiconductor materials, wafer manufacturing, packaging and testing, and semiconductor equipment. Its full-process and full-stack products and solutions for smart factories cover production operations, quality and yield improvement, equipment health, energy consumption management and other fields. It also reconstructs the technical architecture of traditional software by utilizing new technologies to better support high-concurrency, low-latency, precise and stable business scenarios, and excels in integrating a large number of complex, non-native systems.

Taking the whole-factory CIM project implemented in a domestic 8-inch factory as an example, Gechuang Dongzhi helped customers realize production management, quality control and process optimization of multiple factories and workshops from silicon wafers, wafer production, packaging testing to finished product assembly. It is the first domestic full-stack CIM case that has truly been implemented in the front, middle and back ends. By completing the data integration of multiple businesses, multiple processes and multiple systems, customers can achieve full-process traceability management from raw material pulling to finished product assembly, greatly optimizing the management efficiency of multiple factories, and the collaboration between MES and EAP has reached a leading level.

Behind the full-stack products and whole-plant capabilities is a growing semiconductor professional team of more than 700 people at Gechuang Dongzhi. Among them, a group of core technical experts such as Xiao Changbao, Xu Lei, and Huang Shengxiang have served in leading domestic and foreign semiconductor companies such as Intel, IBM, Applied Materials, TSMC, and Silan Microelectronics. They have an average industry experience of more than 15 years, and have rich experience in 6/8/12-inch wafer fab construction and CIM, MES, EAP, APC, FDC and other system construction, and have strong R&D, delivery and scale capabilities.

As more and more wafer fabs regard CIM as an important track to win in the next decade, how to fully explore and give full play to the greater value of CIM in this competition has become a question that must be answered. Although the industry generally believes that supporting the production of all links in the wafer fab is the "basic plate" of the CIM system, some leading manufacturers have also put forward more forward-looking views.

"After achieving automation and informatization, what is more valuable to the wafer fab is to make full use of the data resources continuously generated in manufacturing, realize fully automated control through big data analysis and intelligent prediction, and truly realize unmanned intelligent factories. Faced with the repeated global epidemics and increasingly severe challenges in the security of the industrial chain, this is increasingly important for 12-inch factories with a high degree of automation and increasing challenges in capacity expansion." Huang Shengxiang, a semiconductor expert at Gechuang Dongzhi, said. He has worked for industry leaders such as TSMC and has led a large number of semiconductor MES, EAP, APC, ERP and other construction projects. He is well aware of the importance of data value to wafer fabs, especially 12-inch factories.

Take the yield and quality, the core indicators that wafer and packaging and testing plants pay the most attention to besides production capacity, as an example. They are closely related to factory efficiency and profit margins. With the increase in production capacity and the optimization of processes, there are more and more factors affecting production yield, and the complexity between the various influencing factors is also increasing. There are also a large number of hidden influencing factors that are difficult to identify and judge through traditional experience. Yield and quality issues have become the main "stumbling block" that hinders semiconductor companies from achieving rapid growth. For example, the CMP (chemical mechanical polishing) process, a key process in wafer production, is highly complex and has strict requirements on the uniformity of material removal, high production efficiency and low cost. It relies on manual spot checks of film thickness at the site, and it is difficult to carry out abnormal analysis of film thickness in the same batch and different batches. Not only does the efficiency and accuracy of detection need to be improved, but it is also difficult to detect the factors that cause abnormal defects in a timely manner.

Through technological innovation, Gechuang Dongzhi has launched a rich and complete portfolio of diversified products in terms of quality, including yield management system YMS, single factor monitoring system SPC, multi-factor analysis system MFA, visual inspection system ADC and big data analysis platform, etc. The principle behind these product combinations is to collect, process and analyze massive structured and unstructured data sources, identify and monitor, and conduct intelligent analysis to fully tap into multiple influencing factors hidden in the production process, timely realize quality warning, quality traceability, and yield warning, and use big data analysis results to force process optimization and avoid unnecessary quality losses.

In response to the difficulties in CMP detection mentioned above, Gechuang Dongzhi built a big data platform and made full use of data parameters related to the equipment and process to establish a multi-factor analysis model for CMP film thickness anomalies and defect root causes, taking into account massive influencing factors and correlations, and unearthing related influencing factors that engineers could not identify in the past; in terms of improving the timeliness of discovering anomalies, virtual measurement (VM) is used to build a model to predict film thickness values ​​with high precision and intercept abnormal values ​​in real time, realizing "real-time full inspection" of characteristic values, reducing the frequency of site spot checks by 60%, giving full play to the value of data, and effectively reducing costs.

In the decades of ups and downs in the semiconductor industry, there has always been a phenomenon of focusing on hardware and neglecting software, which is also one of the important factors restricting my country's wafer fabs from seizing the mid-to-high-end market. With the unexpected chip shortage in recent years and the subsequent wave of factory construction, more and more wafer fabs have begun to realize the importance of chip production and manufacturing related software, and "soften overcome hardness" to solve the difficulties and pain points in production.

Therefore, we also look forward to more domestic software manufacturers like Gechuang Dongzhi in the future, which will have rich whole-factory capabilities and innovative technological advantages, and collaborate with more hardware manufacturers and manufacturing plants to escort China's semiconductors into the golden decade.

*Disclaimer: This article is originally written by the author. The content of the article is the author's personal opinion. Semiconductor Industry Observer reprints it only to convey a different point of view. It does not mean that Semiconductor Industry Observer agrees or supports this point of view. If you have any objections, please contact Semiconductor Industry Observer.

Today is the 3091st content shared by "Semiconductor Industry Observer" for you, welcome to follow.