Recent advances address the looming need for complex nodes and programs, but everything is still in place.
New advances and inventions in metrology and inspection will enable chipmakers to identify and fix defects faster and more accurately than ever, which will be necessary in long-term procedural nodes and in densely packed chipsets.
These advancements will have an effect on front-end and back-end processes, offering greater accuracy and efficiency, combined with synthetic intelligence/machine learning and insights analytics. These types of innovations will be very important to meet the industry’s conversion desires, enabling deeper insights and more accurate measurements at speeds suitable for high-volume manufacturing. But gaps still need to be filled, and new ones will most likely emerge as new nodes and processes are deployed.
“As semiconductor devices become more complex, demand for high-resolution, high-precision metrology equipment increases,” says Brad Perkins, product line manager at Nordson Test.
The move to high NA EUV lithography (0. 55 NA EUV) at the 2 nm node and beyond is expected to exacerbate stochastic variability, requiring physically more powerful bottom-up metrology solutions. Traditional critical size (CD) measures alone are inadequate for the required research objective. To ensure the integrity and functionality of complex elements, it is mandatory to perform comprehensive measurements that add line edge roughness (LER), line width roughness (LWR), local edge placement error (LEPE), and local CD uniformity ( LCDU), as well as CD measurements. semiconductor devices. These measurements require complex equipment capable of capturing and analyzing small variations at the nanoscale, where even slight deviations can have a significant impact on the capacity and functionality of the device.
“Metrology is now at the forefront of performance, especially given the existing demand for DRAM and HBM,” said Hamed Sadeghian, President and CEO of Nearfield Instruments. “The next generations of HBM are reaching a level where hybrid bonding will be imperative due to increasing stack thickness. Hybrid bonding requires maximum resolutions in vertical directions to ensure that all pads and surface height relative to the dielectric remain within the nanoscale procedure windows. Therefore, the equipment used should be an order of magnitude more accurate.
To address those challenges, corporations are creating hybrid metrology systems that combine diverse measurement techniques for a comprehensive set of knowledge. The integration of scattering, electron microscopy, and/or atomic force microscopy allows for deeper investigation of critical features. In addition, AI and ML algorithms use the predictive features of those tools, allowing adjustments to be made to procedures.
“Our consumers who are migrating to more complex build nodes are desperate to know what drives their performance,” said Ronald Chaffee, senior director of application engineering at NI/Emerson Test.
Traditional defect detection, trend recognition, and quality strategies typically used spatial trend recognizers and wafer photograph-based pasrhythms to troubleshoot wafer-level problems. “However, we want to go beyond those techniques,” says Prasad Bachiraju, Senior Director of Business Development. at Onto Innovation. ” Our observations show that about 20% of wafers have systematic disorders that can limit performance, of which only about 4% are new additions. There is a pressing need for complex metrology for in-line tracking to achieve defect-free manufacturing. .
Several corporations have recently announced inventions in metrology to enable more accurate inspections, in hard-to-see areas, edge effects, and highly reflective surfaces.
Nordson has introduced its AMI SpinSAM acoustic rotary scanning formulation. The formula represents a significant replacement for classic raster scanning methods, a rotary scanning approach. Instead of moving the wafer in an x,y trend relative to a desktop lens, the wafer rotates, like a record player. This reduces wafer movement and increases inspection speed, eliminating the need to stitch symbols and achieving better symbol quality.
“We have been looking to implement this strategy for years and it is gratifying that despite everything it is successful. It’s anything we think would be incredibly beneficial,” Perkins says. “The SpinSAM is primarily designed to improve inspection speed and power, meeting the industry’s usual demand for higher product throughput and higher state-of-the-art inspection capabilities. »
At the same time, Nearfield Instruments presented a multi-head atomic force microscopy (AFM) formula called QUADRA. It is a non-destructive, high-performance metrology tool for HVM that features a new multi-miniaturized AFM head architecture. Nearfield claims that the independent, parallel multi-head scanner can deliver a hundred times greater performance than traditional single-probe AFM tools. This architecture enables accurate measurements of superior aspect ratio structures and complex three-dimensional functions, for complex memory and logical processes (3-D NAND, DRAM, HBM).
In April, Onto Innovation introduced a breakthrough in underground defect inspection generation with the launch of its Dragonfly G3 inspection formula. The new formula allows for one hundred percent inspection of wafers, focusing on subsurface defects that can lead to performance losses, such as microcracks and other hidden defects that can lead to failure of entire wafers in subsequent processing steps. The Dragonfly G3 uses a new generation of infrared (IR) combined with specially designed algorithms to detect those defects, which in the past were difficult to fix in a production environment. The new capability supports HBM, complex logic, and various specialized segments, and aims to end performance and load savings by cutting wafers and stacks from discarded chips.
More recently, researchers at the Paul Scherrer Institute announced a high-throughput X-ray tomography strategy consistent with burst ptography. This new approach can provide consistent, detailed, non-destructive perspectives of nanostructures as small as four nm in tissues such as silicon and metals at a rapid acquisition rate of 14,000 solution elements consistent with one second. Backpropagation tomographic reconstruction allows samples to be imaged up to ten times larger than the traditional field strength.
There are also other metrology technologies and techniques in semiconductor manufacturing, adding ultrasonic inspection at the wafer level, which consists of flipping the wafer to inspect it on the other side. New acoustic microscopy techniques, such as scanning acoustic microscopy (SAM) and in-flight acoustic microscopy (TOF-AM), allow the detection and characterization of very small defects, such as voids, delaminations and cracks in thin films and interfaces.
“We used to look at 80 to one hundred micron films, but with packaging built into 3D, we’re now dealing with 160 to 240 micron films, very thick films,” says Christopher Claypool, senior applications scientist at Bruker OCD. . “In TSV and microbumps, the dominant method today is white light interferometry, which provides profile data. While it has some advantages, it is slow to perform and is a focus-based strategy. “This limitation makes it difficult to measure TSV structures smaller than 4 or five microns in diameter. ”
Acoustic metrology equipment supplied with the latest generation of focal length transducers (FLTs) can emit acoustic waves with an accuracy of a few microns, allowing for detailed, non-destructive inspection of edge defects and critical stress points. This feature is especially useful for identifying small-scale images. defects that could go undetected with other inspection methods.
The progression and integration of smart sensors into metrology apparatus plays a key role in gathering enormous amounts of knowledge required for accurate measurement and quality control. These sensors are very sensitive and capable of operating in various environmental conditions, ensuring consistent operation. One of the main benefits of smart sensors is their ability to facilitate expected maintenance. By frequently monitoring the status and functionality of metrology devices, these sensors can anticipate potential failures and plan maintenance before significant downtime occurs. This capitalization capability improves appliance operation capability, reduces maintenance prices, and improves overall operating efficiency.
Smart sensors are also evolving to integrate seamlessly with metrology systems, offering real-time knowledge collection and analysis. These sensors can monitor various parameters of the production process, offering continuous feedback and allowing immediate changes to avoid defects. Smart sensors, combined with large knowledge platforms and complex knowledge analysis, enable more effective and accurate defect detection and classification.
A persistent challenge in semiconductor metrology is the identification and inspection of defects in critical stress problems, especially at the edges of silicon. For glued pads, it is on the outer ring of the pad. For chip-on-wafer packaging, it is at the edge of the chips. These edge defects are particularly problematic because they occur at higher stress issues relative to the impartial axis, making them more prone to failure. As semiconductor devices continue to rely on more complex packaging techniques, such as on-chip or wafer-level packaging, edge inspection becomes even more critical.
“When defects occur in a factory, what is needed is images that can identify and classify them,” says Onto’s Bachiraju. “So what you have to do is locate the fundamental reasons for its origin, and for that you need a complete integration of knowledge and a Big Data platform to facilitate faster analysis. “
Another major challenge in semiconductor metrology is ensuring the reliability of known chips (KGDs), especially as packaging techniques and complex chipsets prevail. Ensuring that every chip/chiplet in a stacked chip configuration is of the best quality is critical to maintaining performance. and performance, but the speed of metrology processes is a constant concern. This leads to a balance between rigor and efficiency. The industry is continually looking to develop faster machines that can handle the increasing volume and complexity of inspections without compromising accuracy. In this career, inventions in knowledge processing and research are key to achieving faster results.
“Customers need 100 percent inspection for many of those processes because of the intelligent quality of the dies, but the price is too high because the machines simply can’t run fast enough,” says Nordson’s Perkins.
Industry 4. 0 – a term introduced in Germany in 2011 to refer to the fourth commercial revolution and called intelligent production in the United States – focuses on the integration of virtual technologies such as the Internet of Things, synthetic intelligence and the analysis of great knowledge in production processes. Unlike the further revolutions driven by mechanization, electrification, and computerization, Industry 4. 0 focuses on connectivity, knowledge, and automation of production functions and efficiencies.
“The greater the integration of knowledge, the more effective the performance ramp,” says Dieter Rathei, CEO of DR Yield. “It is essential to integrate all available knowledge into the formula for effective monitoring and analysis. »
In semiconductor manufacturing, this shift toward Industry 4. 0 is particularly transformative, driven by the increasing complexity of semiconductor devices and the demand for greater precision and performance. Traditional metrology methods, which rely heavily on manual processes and limited automation, are evolving into highly interconnected systems that enable real-time knowledge sharing and decision-making across the production chain.
“There are many teams available to consolidate other types of knowledge on a single platform,” says NI’s Chaffee. “Historically, performance control systems focused on testing, while FDC or procedure systems focused on the procedure itself, without correlating the two. As brands venture into the 5, 3, and 2 nm domains, they discover that defect density alone is not the only factor to determine. Process control is also crucial. By integrating all knowledge, even the most complex correlations that a human might miss can be known through AI and machine learning. The goal is to use machine learning to find patterns or connections that can help control and optimize the production process.
IoT bureaucracy is the backbone of Industry 4. 0 by connecting various devices, sensors, and systems within the production environment. In semiconductor production, IoT enables seamless communication between metrology tools, production equipment, and factory control systems. This interconnected network facilitates real-time monitoring of production processes, allowing rapid changes and optimization.
“You need to integrate data from a variety of sources, adding sensors, metrology tools and verification frameworks, to create predictive models that process control and performance,” says Michael Yu, vice president of complex responses at PDF Solutions. “This holistic technique allows you to identify patterns and correlations that were undetectable in the past. “
AI and machine learning play a critical role in processing and analyzing the vast amounts of knowledge generated in a smart factory. These technologies can identify patterns, detect faults in devices, and optimize procedure parameters with a level of precision and speed that is not within the reach of humans. operators alone. In semiconductor manufacturing, AI-based analytical procedures control throughput rates and reduce downtime. “One of the major trends we’re seeing is the integration of synthetic intelligence and device learning into metrology tools,” Perkins says. helping to make sense of the gigantic amounts of knowledge generated and enabling more accurate and effective measurements. “
The role of AI extends even further, as it helps uncover anomalies within the production process that might have gone unnoticed with classical methods. AI algorithms embedded in metrology systems can dynamically adjust procedures in real time, ensuring that deviations are corrected before they finally subside. This AI integration minimizes defect rates and improves overall production quality.
“Our experience has shown that over the past 20 years, machine learning and AI algorithms have played an important role in automatic data classification and matrix classification,” says Bachiraju. “This has particularly improved the power and accuracy of our metrology tools. “
Big knowledge analytics complements AI/ML by providing the infrastructure needed to manage and interpret large sets of knowledge. In semiconductor manufacturing, big knowledge analytics is helping to extract actionable insights from the knowledge generated through IoT devices and production systems. This capability is very important for predictive maintenance, quality and continuous improvement of procedures.
“With big data, we can identify patterns and correlations that in the past were more unlikely to detect, leading to greater procedure and higher throughput,” Perkins says.
Big knowledge analysis is also helping to understand the life cycle of semiconductor devices, from production to deployment in the box. By analyzing knowledge of product functionality over time, brands can expect potential errors in product design, expanding reliability and lifecycle management.
“Over the next decade, we will see many opportunities for AI,” says Rathei of DR Yield. “The basis of this progress is the availability of comprehensive knowledge. AI models need detailed knowledge to train. Once all the knowledge is available, we can experiment with other models and ideas. The ingenuity of engineers, combined with new tools, will allow exponential progress in this area.
Despite advances in metrology, analytics, and AI/ML, several gaps remain, especially in the context of high-volume production (HVM) and next-generation devices. The R metrology program
Metrology for purity and material properties: There is a critical need for new measurements and criteria to ensure the purity and physical properties of materials used in semiconductor manufacturing. Current techniques lack the sensitivity and performance necessary to encounter waste and contaminants in the source chain.
Advanced metrology for future production: Next-generation semiconductor devices, such as full-gate FETs (GAA) and complementary FETs (CFETs), require advances in physical and computational metrology. Existing equipment is still capable of offering the resolution, sensitivity and precision necessary to characterize the complex characteristics and structures of these devices. This includes non-destructive techniques to characterize defects and impurities at the nanoscale.
“There’s a secondary challenge with some metrology equipment, which is sampling data from individual parts on a wafer, similar to thermal test data that only covers fast sites,” Chaffee says. “To make sense, we want to go beyond sampling strategies and find artistic tactics to collect data on each wafer, integrating it into a model. This comes to build a knowledge base that can help detect patterns and correlations, which only humans could overlook. The key is to leverage artificial intelligence and device learning to identify those correlations and make sense of them, especially as we move into the 5nm and 2nm spaces. This procedure is iterative and requires a holistic approach, encompassing various knowledge issues and correlating them to perceive physical limitations and have an effect on the final product.
Metrology for complex packaging: Integrating complicated parts and new fabrics into complex packaging technologies presents significant metrological challenges. Rapid in situ measurements are required to determine interfaces, underground interconnections, and internal three-dimensional structures. Current strategies adequately address issues such as warping, voids, substrate performance and adhesion, which are critical to the reliability and functionality of complex enclosures.
Modeling and Simulation of Semiconductor Materials, Designs, and Components: Modeling and simulation of semiconductor procedures requires complex computer models and knowledgeable research teams. Current functions have limited ability to seamlessly integrate the entire semiconductor price chain, from curtain source to formula assembly. There is a need for validation criteria and equipment for virtual twins and other complex simulation techniques that can optimize the progression and control of procedures.
“Predictive analytics is especially important,” Chaffee says. Its goal is to determine the likelihood that a given chip on a wafer will perform most productively or have problems. By integrating various knowledge questions and executing other scenarios, they can identify and perceive how combinations, sequences, and processes of express devices are obtained.
Semiconductor Procedure Modeling and Simulation: Today’s functions have limited ability to seamlessly integrate the entire semiconductor price chain, from curtain inputs to formula assembly. There is a need for validation criteria and equipment for virtual twins and other complex simulation techniques that can optimize the procedure. progression and control.
“Part of the challenge is the subsequent packaging and assembly process, but another component of the challenge could be the quality of the wafer itself, which is decided in the initial process,” Yu says. An effective ML style will need to integrate front-end and back-end information, aggregating knowledge from apparatus sensors, metrology, and structured verification information, to make accurate predictions and take proactive steps to properly form the procedure.
Standardization of new fabrics and processes: The long-term progression of communication and data technologies depends on the creation of new criteria and validation methods. Current reference tissues and calibration facilities do not meet the needs of next-generation tissues and processes such as those used. in complex packaging and heterogeneous integration. This hole hinders the industry’s capacity for innovation and its competitive production capabilities.
Metrology to improve the safety and provenance of parts and products: With the increasing complexity of the semiconductor source chain, metrology answers are needed that can ensure the safety and provenance of parts and products. fabrics and processes the production lifecycle to prevent counterfeiting and ensure compliance with regulatory standards.
“The focus on security and sharing turns appointments with providers into a partnership rather than a confrontation,” Chaffee says. “Historically, there has been a fear that information will cross that boundary. People are very protective of their procedure, and other people are very protective of their product. But once you start entering deep submicron space, those barriers have to come down. The chips are too expensive for them not to be able to communicate, but they can still do so while protecting their intellectuals. property. Companies are beginning to realize that by securely sharing parametric verification information, they can achieve greater performance control and procedure optimization without compromising their intellectual property.
Advances in metrology and testing are critical to the continued expansion and innovation of the semiconductor industry. The integration of AI/ML, IoT, and big data analytics is transforming the way brands manage the process and improve performance. As Industry 4. 0 adoption grows, the role of metrology will become even more critical in ensuring the efficiency, quality, and reliability of semiconductor devices. And by leveraging those complex technologies, semiconductor brands can achieve higher yields, reduce costs, and maintain the accuracy required in this competitive industry.
Through continuous innovations and the integration of smart technologies, the semiconductor industry will continue to overcome the barriers of innovation, leading to more powerful and high-performance physical electronic devices that will mark the technology’s long-term. The journey towards a fully learned Industry 4. 0 is underway and its impact on semiconductor production will undoubtedly shape the future of the industry, ensuring that it remains at the forefront of global technological advancements.
“Whenever new packaging technologies and procedures evolve, what you’re looking for is metrology,” Perkins says. “When you create new procedures and want to make continuous innovations in terms of performance, that’s when you feel the greatest need for new metrology. solutions. “