Global Semiconductor Technology Roadmap
Release Date:
2014-08-04
Source:
Author:
Precision amplifiers play a vital role in data acquisition systems, which are essential in a wide range of electronic applications. These amplifiers are specifically designed to amplify small signals with minimal distortion, ensuring that data collected from sensors or other sources is as accurate as possible. The significance of precision amplifiers lies in their ability to maintain high fidelity
I. The Semiconductor Industry Ecosystem
The semiconductor industry emerged in the 1970s, driven primarily by two factors: first, the need to provide the computing sector with more cost‑effective memory; and second, the demand for rapidly manufactured application‑specific integrated circuits to support companies in developing new products with specialized functions.
By the 1980s, system standards were firmly in the hands of system integrators. Storage devices updated their semiconductor technology every three years, and these advancements were promptly adopted by logic‑device manufacturers.
In the 1990s, logic‑device integrated circuit manufacturers accelerated the adoption of new technologies, updating their product generations every two years—keeping pace with memory‑chip makers. The unusually strong correlation between technological advances and improvements in product performance enabled a significant portion of control over system performance and profitability to shift to IC manufacturers. Leveraging this new balance of power, they drove the semiconductor industry’s revenue growth to an average annual rate of 17% during this period.
In the first decade of the 21st century, a completely new ecosystem had taken shape in the semiconductor industry:
First, semiconductor technology is updated every two years, enabling the efficient and low-cost production of integrated circuits and millions of transistors. This allows highly complex systems to be integrated at a lower cost on a single chip or within the same package. Moreover, advances in packaging technologies make it possible to place multiple chips within a single package. Such devices are classified as system-on-chip (SoC) and system-in-package (SiP).
Second, integrated‑circuit foundries can once again offer “next‑generation application‑specific integrated circuits” at highly attractive costs, giving rise to a highly profitable sector: integrated‑circuit design.
Third, advances in high-end integrated‑circuit manufacturing equipment have spurred progress in related technological fields, significantly reducing the costs of devices such as flat‑panel displays, microelectromechanical systems (MEMS) sensors, radio‑frequency equipment, and passive components. Under these conditions, system integrators have once again assumed control over system design and product integration.
Fourth, the rise of Internet applications and mobile smart devices has spurred the widespread deployment of fiber-optic cables and the development of diverse wireless technologies, enabling unprecedented global mobile connectivity. This ecosystem has given birth to the emerging “Internet of Things” market, with innovative product manufacturers, telecom operators, data and information distributors, and content providers vying for dominance in this space.
Semiconductors are the cornerstone of all the aforementioned applications, and no innovation can be realized without the support of the semiconductor industry.
II. Global Semiconductor Technology Roadmap
In the late 1960s, the invention of the self-aligned silicon‑gate process laid the foundation for modern semiconductor scaling. Moore’s “Moore’s Law,” which posited a doubling of transistor density every two years starting in 1965, along with Dennard’s “Dennard Scaling” introduced in 1975, fueled the growth of the semiconductor industry and defined the era of classical geometrically driven scaling through the early 21st century. The advent of the equivalent scaling era was underpinned by technologies such as strained silicon, high‑k metal gates, multi‑gate transistors, and compound semiconductors. These innovations have propelled the semiconductor industry over the past decade and will continue to drive its future evolution.
(1) Components
Information-processing technologies are driving the semiconductor industry into an increasingly broad range of applications, and device cost and performance will continue to be closely tied to the scaling and functional expansion of complementary metal-oxide-semiconductor (CMOS) transistors.
Strained silicon, high‑k metal gates, and multi‑gate transistors are now widely used in integrated circuit manufacturing. The next frontier for further performance enhancement lies in III–V compound materials and germanium. Compared with silicon devices, these materials offer significantly higher carrier mobilities. To leverage the advantages of the mature silicon platform, it is anticipated that new high‑mobility materials will be grown epitaxially on silicon substrates.
2D scaling will ultimately reach its fundamental limits by the 2013 International Technology Roadmap for Semiconductors (ITRS), and both logic and memory devices are now exploring ways to leverage the vertical dimension—namely, 3D. The convergence of 3D device architectures with low‑power technologies will usher in an era of “3D power scaling,” with further increases in transistor density per unit area ultimately achieved through multi‑layer transistor stacking.
Unfortunately, there have been no breakthroughs in interconnect technology, as no viable material yet exists with a resistivity lower than that of copper. However, advances in handling edgeless wrapped materials such as carbon nanotubes and graphene composites are laying the groundwork for the development of “ballistic conductors,” which could emerge within the next decade.
Three-dimensional packaging of multiple chips offers a viable approach to reducing interconnect resistance, primarily by increasing the cross-sectional area of conductive traces (in the vertical direction) and shortening the length of each interconnect path.
However, the lateral scaling of CMOS—and of equivalent devices currently under investigation—will ultimately reach its physical limits. Future opportunities in semiconductor technology lie in two areas: first, extending the functionality of the CMOS platform through heterogeneous integration with new technologies; and second, developing devices that support next-generation information-processing paradigms.
(II) System Integration
System integration has shifted from a data‑processing, PC‑centric paradigm to a highly diversified mobile‑communication model. Integrated circuit design is transitioning from a performance‑driven approach to one focused on low power consumption, enabling the heterogeneous integration of multiple technologies within constrained form factors—such as GPS devices, telephones, tablets, and smartphones—and thereby fundamentally transforming the semiconductor industry. In short, whereas performance was once the singular objective, today the goal of minimizing power consumption is steering integrated circuit design.
System-on-chip and system-in-package products have become the primary drivers of the semiconductor industry. Over the past few years, the production volume of smartphones and tablets has surpassed that of microprocessors. Heterogeneous integration rests on the integration of “More Moore” (MM) devices with “More than Moore” (MtM) components.
For example, today, microelectromechanical systems (MEMS) devices are integrated into a wide range of systems, including automobiles, video projectors, tablets, smartphones, and gaming platforms. In general, MEMS devices add valuable functionality to these systems, enhancing their core capabilities. For instance, the MEMS accelerometer in a smartphone can detect the device’s orientation and rotate the screen display accordingly. While such additional features improve the user interface, the phone can still function without them. By contrast, video projectors and inkjet printers that rely on digital light processing (DLP) technology would be unable to operate properly without MEMS components. Moreover, multimodal sensing technologies have become an integral part of mobile devices, serving as a key enabler of the Internet of Things.
The rapid advancement of digital data and connectivity technologies is transforming healthcare. Silicon, microelectromechanical systems, and optical sensing technologies are making this revolution possible.
Mobile phones can now provide a wealth of health information. Accelerometers track physical activity and sleep, while the built-in light sensor measures heart rate when the user touches the device. The phone’s camera can be used for various purposes, such as estimating the calorie content of food or detecting emotions based on facial expressions. A wide range of mobile apps has emerged that can analyze this data and present it to users in an intuitive, easy-to-understand format.
Looking ahead to the next 7–15 years (beyond 2020), advances in devices and systems will give rise to entirely new architectures based on groundbreaking principles. For example, spin‑wave devices (SWDs) are magnetic logic components that leverage collective rotational oscillations—spin waves—to transmit and process information. These devices convert input voltage signals into spin waves, perform computations on the spin waves, and then map the output spin waves back into voltage signals. Within a single‑core architecture, large‑scale parallel data processing across multiple frequencies can be achieved by assigning each frequency to a distinct information channel, all while consuming very low power. Moreover, emerging device technologies are driving the development of novel architectures. For instance, storage‑class memory (SCM) integrates the high performance and robustness of solid‑state storage with archival capabilities and the cost‑effectiveness of conventional hard‑disk magnetic storage. Such a device requires a nonvolatile memory (NVM) technology capable of delivering extremely low per‑bit storage costs.
(3) Manufacturing
Driven by dimensional scaling, the precision of integrated circuit manufacturing will reach the sub‑nanometer regime within the next 15 years. Measuring physical properties on wafers has become increasingly challenging, but this task can be largely accomplished by correlating process and equipment parameters. By ensuring equipment stability and process repeatability, precise control of process parameters such as feature size is now achievable.
Fab operations are increasingly data‑driven, with requirements for data volume, communication speed, data quality, and availability being clearly defined and quantified. The transition of wafers from 300 mm to 450 mm presents significant challenges. Efforts should focus on developing common technologies applicable to both 300 mm and 450 mm, as 450 mm fabs will benefit from enhancements derived from 300 mm‑compatible process improvements.
System-on-chip and system-in-package integration will continue to gain momentum. Rising levels of integration are driving the re‑integration of test solutions to balance test costs with product quality requirements. Optimized test solutions may necessitate access to and testing of embedded modules and cores. Moreover, Known Good Die (KGD) technology—providing high‑quality die for multi‑chip packages—is becoming increasingly critical, serving as a key factor in balancing test complexity and cost.
III. Major Challenges
(1) Short-term challenges (from now until 2020): Performance improvement
1. Logic Devices
The conventional scaling path of planar complementary metal–oxide–semiconductor (CMOS) technology will face severe challenges in both performance and power consumption.
Despite the introduction of high‑k/metal‑gate (HKMG) technology, reducing the equivalent gate oxide thickness (EOT) remains challenging in the short term. Integrating high‑k dielectrics while simultaneously mitigating the increase in gate leakage current caused by bandgap narrowing will also pose significant challenges. A comprehensive optimization of the gate stack material system is required to achieve optimal device characteristics—both power and performance—while minimizing costs.
New device architectures, such as multi-gate metal–oxide–semiconductor field-effect transistors (MOSFETs) and fully depleted silicon-on-insulator (FD-SOI), are emerging. A particularly challenging issue is the precise control of the gate oxide thickness in these ultra-thin MOSFETs. Addressing these challenges should proceed in parallel with advances in circuit design and system architecture.
Some high-mobility materials, such as germanium and III–V compounds, have been considered as upgrades or replacements for silicon channels in CMOS logic applications. Key challenges include developing high‑k metal gate dielectrics with low bulk traps and low leakage current, an unpinned Fermi level, and low ohmic contact resistance.
2. Storage Devices
The challenge in dynamic random-access memory (DRAM) lies in achieving an adequate storage capacitance under conditions of shrinking feature sizes, the use of high‑k dielectrics, low‑leakage access‑device designs, and low‑resistivity materials for bitlines and wordlines. To increase bit density and reduce manufacturing costs, 4F²‑type cell drivers require high aspect ratios and non‑planar transistor architectures.
Flash memory has become a key driver of front-end-of-line (FEOL) technologies, including critical dimension scaling, materials, and process‑level techniques such as lithography and etching. In the short term, the continued increase in flash‑memory density hinges on further thinning of the tunnel oxide layer and higher levels of dielectric integration.
To meet the requirements for charge retention and endurance, the adoption of high‑dielectric materials will be essential. Achieving both cost‑effectiveness and reliable performance for 3D NAND flash capacities exceeding 256 GB, while maintaining multi‑level cell (MLC) technology, remains a formidable challenge. Additional challenges include the ongoing evolution of new memory‑type manufacturing processes, as well as emerging memory concepts such as magnetic random-access memory (MRAM), phase-change memory (PCM), resistive random-access memory (ReRAM), and ferroelectric random-access memory (FeRAM).
3. High-performance, low-cost RF and analog/mixed-signal solutions
Advancing the adoption of CMOS technology—featuring high‑k dielectrics and strain engineering—in wireless transceiver integrated circuits and millimeter‑wave applications may require techniques to keep device mismatch and 1/f noise within acceptable limits. Additional challenges include integrating lower‑cost, high‑density passive components, incorporating MEMS that combine efficient on‑chip passives with off‑chip passive networks, and developing low‑cost, non‑silicon (e.g., gallium nitride) devices.
As chip complexity and operating frequencies increase while supply voltages decrease, signal isolation between digital and analog regions on the chip becomes increasingly critical. Noise reduction may require further innovation—such as adopting advanced design techniques to address power‑supply and ground‑connection challenges in substrates with resistivities on the order of thousands of ohms per centimeter.
Many material‑driven and structural changes—such as the adoption of multi‑gate architectures and silicon‑on‑insulator (SOI) technologies, or shifts in the behavior of RF and analog devices—introduce significant trade‑offs. These trade‑offs, coupled with the need to optimize RF, high‑frequency, and analog‑mixed‑signal (AMS) performance while steadily reducing supply voltages, pose substantial challenges for integrated circuit design.
4. 32, 22 nanometer half-pitch and below
Lithography is becoming an extremely costly and challenging technology. For 22-nanometer half-pitch lithography, spacer‑based lithography or multi‑pattern 193‑nanometer immersion lithography systems are being employed to overcome the limitations of single‑pattern approaches; however, these methods suffer from very high mask error enhancement factors (MEEF), significant line edge roughness (LER), stringent design rule constraints, and substantially increased costs. Deep ultraviolet lithography at a wavelength of 13.5 nanometers—known as extreme‑ultraviolet lithography (EUVL)—is widely regarded as the industry’s official path forward for sustaining Moore’s Law.
The challenges of deep ultraviolet lithography include the lack of high‑power light sources, the need for high‑speed photoresists, and the latency introduced by defect‑free, highly planar masks. Additional challenges encompass increasing the numerical aperture of deep‑UV systems beyond 0.35 and exploring the feasibility of incorporating more reflective mirrors into the imaging system.
Multiple-e-beam maskless lithography holds the potential to circumvent mask‑based challenges, eliminate design‑rule constraints, and offer greater manufacturing flexibility. Progress has already been made in rendering high‑resolution images and achieving tight CD control. However, key areas requiring further development include tool timing, cost, defect management, overlay accuracy, and photoresist performance.
Direct Self-Assembly (DSA) technology has made new advances, but defects and positioning accuracy still require urgent improvement.
Other challenges include controlling and suppressing the gate‑length CD of the light‑emitting resist (LER) in lithography and etching, as well as characterizing new gate materials, non‑planar transistor architectures, photoresist‑based LERs, and deep‑ultraviolet lithography.
5. Introduction of new materials
Since low‑dielectric materials—including porous materials and air gaps—must possess sufficient mechanical strength to withstand cutting, packaging, and assembly, minimizing dielectric damage becomes increasingly critical when accounting for etching and chemical‑mechanical polishing (CMP) processes. On the metallization front, ultrathin, conformal, low‑resistivity barrier metals must be integrated with copper to achieve both low resistivity and high reliability.
6. Power Management
In most application domains, power management is currently the top priority. As each new generation sees transistor counts double, the cost‑effective thermal performance of packaged chips remains virtually unchanged. To keep systems operational while reducing leakage power, advances in circuit technologies must extend to system design requirements, enhancements in computer‑aided design (CAD) tools, reductions in leakage power, and the performance demands of novel device architectures.
(II) Short-term Challenges (Now to 2020): Cost-effectiveness
1. Lithography
Although deep ultraviolet lithography at a wavelength of 13.5 nanometers is the industry’s official target, it requires very high source power to remain cost‑competitive at 10 nanometers and beyond. If multi‑electron‑beam maskless lithography can maintain per‑exposure throughput, process costs, and pattern fidelity comparable to mask‑based exposure tools, it could be the most economical option. With fewer masks introduced into the process flow, the DSA (Directed Self‑Assembly) architecture for 193‑nm immersion lithography has gained widespread adoption.
2. Front-end Process
We need to achieve low parasitic effects, continue shrinking the gate pitch, scale the area of next-generation substrates to 450‑mm wafers, and adopt breakthrough technologies to address the challenges of lithography.
3. Factory Integration
The key challenges include: first, addressing rapidly evolving and complex business requirements; second, managing the ever-increasing complexity of manufacturing facilities; third, achieving economic growth targets while facing declining marginal returns; fourth, meeting stringent demands for plant and equipment reliability, functionality, efficiency, and cost-effectiveness; fifth, leveraging cross‑boundary integration technologies—such as combining 300‑mm and 450‑mm processes—to realize economies of scale; and sixth, overcoming the unique challenges associated with transitioning to 450‑mm wafers.
4. Meet the market’s evolving cost requirements
Challenges in assembly and packaging include three-dimensional chip stacking (testing: access, cost, and known-good chips), three-dimensional packaging and integration, and testing access to individual wafers or dies.
5. Environment, Safety, and Health
The challenges facing the environmental safety and health sector include: management and efficiency of chemicals and raw materials; process and equipment management; facility technical requirements; product management; and the reuse, recycling, and remanufacturing of end-of-life products.
6. Measurement
Measurement integration at the plant and corporate levels: Measurement strategies should be carefully selected, and sampling must be statistically optimized to support process control based on total cost of ownership (CoO).
(3) Long-Term Challenges (2021–2028): Performance Enhancement
1. Realization of an atypical complementary metal–oxide–semiconductor channel material
To supply sufficient drive current for highly scaled metal–oxide–semiconductor field-effect transistors (MOSFETs), enhanced thermal velocity and quasi‑ballistic transport at the source are deemed essential. Consequently, high‑speed channel materials—such as III–V compounds, germanium‑based narrow channels on silicon substrates, or even semiconductor nanowires, carbon nanotubes, graphene, and other alternatives—remain to be developed. Nonconventional complementary metal–oxide–semiconductor (CMOS) devices will need to be physically or functionally integrated onto a CMOS platform. Such integration necessitates the epitaxial growth of foreign semiconductors on silicon substrates, a process that poses significant challenges. Ideally, material and device performance must remain stable despite exposure to high temperatures and aggressive chemical processing. Moreover, reliability concerns should be identified and addressed early in the technology development cycle.
2. Identification, selection, and implementation of a new storage structure
Nonvolatile memory (NVM) with dense, fast operation and low supply voltages will become highly desirable; further density scaling may necessitate three-dimensional architectures, such as vertically stacking memory cells on a single integrated circuit, provided that acceptable manufacturing yields and performance can be achieved. The challenges of scaling dynamic random-access memory (DRAM) are expected to intensify, particularly in terms of reducing the equivalent oxide thickness (EOT) and achieving extremely low leakage currents and power consumption. All existing forms of nonvolatile memory (NVM) face material‑related limitations, and their success will hinge on the ability to identify and develop alternative materials or to pioneer new, alternative technologies.
3. There is a shift from conventional specifications toward equivalent scaling and functional diversity through unconventional approaches.
The roughness of wire edges, trench depth and cross‑section, sidewall roughness during via formation, etch deviations, thinning caused by cleaning, chemical mechanical polishing (CMP), the intersection of porous low‑k dielectrics with sidewall voids, barrier layer roughness, and copper surface roughness all adversely affect electron scattering in copper interconnects, leading to increased resistivity. Coupled with the adoption of new materials in multi‑layer stacks, shrinking feature sizes, process‑dependent patterning, the use of alternative storage devices, and optical and RF interconnects, these challenges will persist. Moreover, significant hurdles remain in etching, cleaning, and filling high‑aspect‑ratio structures—particularly in low‑dielectric‑constant metal dual‑damascene architectures and in nanoscale‑dimensioned dynamic random‑access memory devices.
The integration of materials and processes used to fabricate novel structures gives rise to compounded complexity, while the increasing number of stacked layers amplifies deformation‑induced effects. Novel or high‑performance devices can be reconfigured into interconnect networks. Three-dimensional chip stacking offers greater functional diversity and helps circumvent the limitations of conventional interconnect architectures. Developing cost‑effective, manufacturable engineering solutions remains a critical challenge.
4. Deep Ultraviolet Lithography Technology
Since extreme ultraviolet lithography (EUVL) remains the optimal solution for 22‑nm and 16‑nm half-pitch nodes, extending it to higher resolutions will pose a significant long-term challenge. As currently understood, achieving an effective numerical aperture (NA) of 0.5 or greater will require either an eight‑mirror unobstructed design or a six‑mirror design with a central obscuration.
The eight-mirror design will incur greater reflection losses, as the additional mirrors require a higher‑power source to achieve the same wafer‑level flux. In the six‑mirror configuration, the mirror angles are smaller, necessitating a reduced field size and potentially longer track lengths. The increase in numerical aperture poses a significant challenge to both designs in terms of depth of focus. Furthermore, to mitigate shadowing and other three‑dimensional effects on the mask, the absorber material, absorber thickness, and multilayer stack must be carefully optimized.
Another potential solution is to reduce the wavelength of deep‑ultraviolet lithography to the 6‑nanometer node. In the short term, this approach would inherit all the existing challenges of deep‑UV lithography—ranging from energy availability to mask infrastructure and photoresist performance. Multi‑mode deep‑UV lithography could also be an option, though it would entail greater process complexity and higher operating costs for users.
(4) Long-term Challenges (2021–2028): Cost-effectiveness
It meets the scalability requirements of cutting-edge factories that are flexible, scalable, and cost‑effective. To maintain production profitability, it is essential to be able to ramp up output within controllable limits in response to fluctuating market demand, as well as to leverage opportunities for task sharing, such as manufacturing outsourcing. Enhancing customer perception of quality—particularly for outsourced manufacturing—remains a challenge.
Scalability refers to the ability to meet the demands of large 300‑mm fabs [40K–50K WSPM] while ensuring that the facility, products, ancillary equipment, production data, and control systems can be reused across technology generations. Cost‑control and task‑sharing strategies are highly anticipated in industry‑wide standardization efforts for infrastructure development, such as data‑standardization and visualization methodologies.
Keywords:
Hot News
Leave a Message Online
*Note: Please ensure that all information provided is accurate and keep your contact details up to date. We will get in touch with you as soon as possible.
Request for Samples / Product Manuals / Contact Information:
Mr. Jia:13022932841(WeChat)
Landline:029-89259995Transfer 833
Technical Support / Contact Information:
Mr. Zhao:13572473285
Landline:029-89259995Transfer to 803
Ms. Xu:18729396827
Company landline:029-89259995
Fax: 029-89259996
Unit 4, Building 6, Beihang Science and Technology Park, Hangtian Middle Road, Aerospace Base, Chang’an District, Xi’an, Shaanxi Province