Beyond CMOS: the Future of Semiconductors
What is "Beyond CMOS?"
Beyond CMOS refers to potential future digital logic technologies that expand beyond the present CMOS scaling limits. These limits are designed to keep device intensity and speeds in check in an effort to combat heating effects.
Digital logic is a fundamental component in the creation of electronic and logic devices. It's what allows us to create circuits, check computer chips, and perform other functions that are necessary to a system's success. The acronym CMOS, which is short for complementary metal oxide semiconductor, is typically used to describe the fundamental switches that are normally on and off in modern semiconductor products.
This means that "Beyond CMOS" refers to future technologies concerning digital logic, or future technologies, which is what we rely on to represent the sequences and signals within a digital circuit. These technologies are anticipated to stretch beyond the present CMOS scaling limits, which have already spanned over an order of magnitude in feature size and two orders of magnitude in speed (Are we reaching the end of CMOS scaling limits?).
What distinguishes Beyond CMOS from current and past semiconductor technology?
Beyond CMOS research and development places a focus on extending integrated circuit technology to fundamentally new approaches in order to determine the best means for technological advancements following the end of CMOS dimensional scaling. For this reason, much of the technology being researched looks different than the semiconductor technology of yesterday and today.
"An overarching goal of the Beyond CMOS chapter is to survey, assess, and catalog viable emerging devices and novel architectures for their long-range potential and technological maturity and to identify the scientific and technological challenges gating their acceptance by the semiconductor industry as having acceptable risk for further development.” (Beyond CMOS).
Researchers hope to create both volatile and non-volatile memory technologies to replace the commonly used SRAM and FLASH options in appropriate applications. Present scaling limits have created a need to obtain electrically accessible memories that are:
- High-speed
- High-density
- Low power
- Embeddable
- Possible non-volatile
Today's memory systems range in size considerably. Some may be so small that they're gigabyte-based--like mobile systems--and others require exabytes' worth of storage. No matter their size, however, the challenge is the same: most computer systems do not consistently run at peak load. Data servers today require exceptional amounts of power even when utilization rates are low.
Issues like these highlight how valuable a persistent memory could be. If this capability did not require constant refresh, it could mean a notably lightened load for logic devices to grapple with during operation. As much as one-third of a large computer system's power is utilized for refresh power; with that power freed up, device users could be operating much faster and more powerful computers.
What does the IRDS™ have to say about Beyond CMOS?
The IRDS™ seeks to identify Beyond CMOS devices that will help enable computing paradigms beyond the capabilities of conventional CMOS architectures and technologies. The Beyond CMOS chapter of the IRDS™ was specifically created to identify research opportunities, assess current devices, and map emerging devices.
A number of components make up the foundation of the Beyond CMOS IRDS™ chapter's efforts. The chapter places heavy emphasis on emerging memory, storage, logic, and information process devices. There are also several emerging application areas of great interest to those involved with the IRDS™. Neural computing, hardware security, and big data analytics are an integral part of the Beyond CMOS chapter's work.
The Beyond CMOS team has noted that the circuit, architecture, and device communities are presently working towards varying goals. They see great importance in synergizing the development of emerging devices in order to enable significant gains in the realm of computing (IEEE Beyond CMOS Workshop, July 2017 (PDF, 12 MB)).
The IRDS™ also recognizes that while conventional CMOS scaling and Moore's Law are anticipated to continue in the coming years, it is clear that we are closing in on traditional scaling limits. Factors like heightening costs and fundamental physical effects will likely put a halt to scaling as we presently know it. The roadmap recognizes that critical dimensions and statistical distributions present roadblocks to efforts to extend CMOS scaling; factors such as gate length (Outside System Connectivity) serve as one example of this.
The IRDS™ places a heavy focus on high performance at low-power consumption. Recent advancements and innovations within the technological sphere have shifted this focus somewhat towards more novel computing paradigms, functionalities, and applications. The roadmap lays out a number of guiding principles for present and future research of Beyond CMOS technology:
- Computational state variable(s) other than solely electron charge
- Non-thermal equilibrium systems
- Novel energy transfer interactions
- Nanoscale thermal management
- Sub-lithographic manufacturing process
- Alternative Architectures
Evolution of CMOS and Beyond CMOS
How has CMOS technology evolved?
For more than thirty years, the performance of integrated circuits (More Moore) has consistently seen exponential improvement; but the history of CMOS reaches much further back than the past three decades. The technology was invented by Frank Wanlass in 1963 (US PATENT 3,356,858, filed June 1963, Issued Dec. 1967) but CMOS technology did not became mainstream until the introduction of the 256Kb CMOS DRAM in 1984; since then, has been essential in constructing integrated circuits, microprocessors, microcontrollers, sensors, RAM, and a host of other digital circuits.
Since the fateful discovery that an integrated circuit's transistors double every eighteen months, much has changed (Moore, Gordon E. (1965). "Cramming more components onto integrated circuits" (PDF, 638 KB). Electronics Magazine.). Conventional CMOS presents a number of challenges that dedicated researchers have had to expend considerable energy overcoming. Problems like high-gate leakage currents, high source-to-drain leakage, gate stack reliability, and channel mobility degradation have all been an essential part of CMOS' evolution.
Within the past fifteen years, a new CMOS technology node (Evolution of CMOS Technology at 32 nm and Beyond) has emerged approximately every twenty-four months. Each node has come with its own changes; generally, 2× density shrink and ~35% performance gain per node. Chip power capabilities have steadily increased and power limits have driven the process of CMOS scaling to a more density-focused process.
Why is Beyond CMOS technology possible now?
It is important to keep in mind that much of Beyond CMOS technology is not really possible now. Niche offerings are available on commercial levels, but the general public certainly isn't working with devices that incorporate Beyond CMOS capabilities. What is possible now is working towards these capabilities.
This work on Beyond CMOS is possible now due to the alignment of several factors: need, outside technological advancement and innovation, and an exceptional level of flexibility. The potential for scaling CMOS technology is rapidly approaching an end. With no other options in sight, researchers and academics worldwide are pushed into considering technologies like these out of sheer necessity.
Obviously, the immense work devoted to the evolution of computing over the past several decades has contributed to present ability as well. Professional research groups are deeply concerned with magnetic devices, resistive-switching electronics, and a number of other technologies and capabilities that have all been researched for some time prior to now.
Because the scope of potential Beyond CMOS technologies is fairly large, researchers also have a great deal of flexibility in terms of where they choose to place focus and channel efforts. This means that those who are experts regarding circuits and architectures, for example, can focus on those and offer thorough insight into development; meanwhile, other groups can channel their work into creating Nano/Micro-ElectroMechanical Systems (N/MEMS) or improving magnetic logic.
These factors, combined with the work of forward-thinking individuals and groups who identified the need for this technology in the first place, have granted us the capability to begin striving towards Beyond CMOS technology. Without the many avenues for research, the many niches to select from, and outside advancement pushing things along, it's entirely possible that research and development would stagnate.
What companies and researchers are driving Beyond CMOS technology?
Intel has been a key player in the game to drive Beyond CMOS forward. The company produced and presented their own magneto-electric spin-orbit (MESO) logic device, which has the capability to lower voltage by 5× and energy by as much as 10-30× (provided it's combined with ultralow sleep state power). The complementary metal oxide semiconductors of today are simply not capable of achieving such things.
As Intel has worked behind the scenes to pursue technologies for the Beyond CMOS era, numerous other companies and researchers have been putting their own efforts into the cause. Semiconductor research centers worldwide are rushing to discover a worthy CMOS successor; and the Nanoelectronics Research Initiative (DOI: 10.1109/ICSICT.2016.7998829) (NRI) is on a mission to benchmark these potential solutions against each other.
UCLA's California NanoSystems Institute (CNSI) (CNSI: Beyond CMOS) has also recognized the global demand for Beyond CMOS technologies and has been at the helm of numerous research- based efforts within the field. CNSI researchers are dedicated to creating flexible, scalable electronic devices that contribute to painting the picture of a future after CMOS.
What related areas of research are closely tied to the development and evolution of Beyond CMOS?
Perhaps the simplest way to pinpoint related areas of research is to revisit the IRDS™ points of focus. Beyond CMOS is a massive, integral part of the goal set but there are other key elements for research and development, too. A number of the roadmap IFTs have been created to assess the present status and potential for future evolution concerning:
- Yield enhancement
- Lithography
- Application benchmarking
- Systems and architectures
- More Moore
- Developing traditional integrated circuits (Colloquial term utilized to refer to historical improvements as described by Moore's Law)
- Outside system connectivity
- Emerging research materials
Applications of Beyond CMOS Technology
What are some current applications of Beyond CMOS technology?
There are a number of niche applications that already rely on Beyond CMOS process technologies. Josephson junctions, for example, are an integral part of quantum computing; they are utilized to help implement qubits and the control systems that manage them. Because qubits are temperature-sensitive, Josephson junctions' low-power dissipation (orders of magnitude less power per computation than CMOS) makes them an ideal solution.
Beyond CMOS technology has also been an integral part of the medical industry for some time. The sensitivity of superconducting quantum interference devices (SQUIDs) has opened doors to opportunities for non-invasive measurements of electrophysiological activity. Magnetoencephalography (or MEG) alone, which is a technology for magnetic source imaging, is responsible for more than half a billion USD in sales to date. The magnetocardiography capabilities medical professionals rely on to diagnose critical problems such as fetal heart rhythm abnormalities are also possible due to Beyond CMOS tech.
What innovations will be made possible by Beyond CMOS technology?
The push for Beyond CMOS technology has resulted in incredible strides in the realm of cryogenic electronics. Low-temperature electronics like these are needed by operations at temperatures lower than 4 degrees Kelvin, typically operating in the milli-Kelvin range and can involve numerous components and devices crafted from insulators, conductors, semiconductors, and superconductors. The presence of preexisting applications and the movement towards emerging ones have translated to novel developments in the realm of cryogenic electronics. This year, Cryogenic Electronics has been spun out from Beyond CMOS to form a separate chapter.
These cryogenic electronics are anticipated to have a number of potential initial application areas. Sensors may be capable of improved accuracy, resolution, and rate; signal and media processing (along with sensor array readouts) may also see improvements. Digital and quantum computing capabilities will likely see a sharp increase, resulting in improvements concerning operations per second, energy per operation, and coherence time.
Magnetic logic is expected to be another key opportunity born from Beyond CMOS tech. Researchers have proposed that mLogic (or logic based on novel magnetic devices) will make an ideal technology for electronic systems in energy-constrained environments. These systems will be able to be powered by notably low voltage and are anticipated to be 3D stackable, resulting in high-density logic and memory.
One key aspect of the innovative process should be considered during discussions regarding this topic. More than any specific technology or application, the innovations that experts predict they will witness (thanks to Beyond CMOS technology) have to do with a vast array of devices and architectural approaches. The goal of Beyond CMOS researchers centers around utilizing the wealth of approaches and devices available in order to invent a new information processing platform technology.
What future devices will incorporate Beyond CMOS technology?
Several emerging devices are poised to become critical in the realm of security applications. Hardware security is becoming an increasingly essential design consideration and Beyond CMOS technologies are anticipated to replace the passive role that CMOS technology currently plays in security. Some demonstrate unique features that could prove useful in simplifying circuit structures, which makes protection easier.
As an example, numerous structures based on magnetic tunnel junctions have been proposed; and some researchers have even exploited variations in write times to produce unique responses in phase change array (PCM) arrays. Because so many of these emerging devices come along with the element of inherent randomness, they lend themselves well to random number generation (RNG). New RNGs based on emerging devices successfully pass randomness tests put forth by the NIST
Digital and quantum computing processes are anticipated to rely on this and other technologies as we move into the future. There are already quantum-annealing processors available for commercial purchase. Based on superconducting qubits, these processors are presently a costly and rare solution to very unique problems. To date, none of them has demonstrated the versatility of current technologies. Fortunately, these technologies are not needed for another 10-15 years at least.
As researchers work to make progress in the realm of Beyond CMOS technology, these sorts of powerful tools are anticipated to become more readily available. Microprocessor units and memories, for example, are presently under development. While they are not ready for public or commercial consumption just yet, their prototypes serve as proof that Beyond CMOS technology may be incorporated into far more tools than we believe.
The future market for digital superconductor computing can hardly be predicted, but analysts and researchers are readying themselves for heightened interest and demand as time goes on. These emerging technologies take time, funding, and expertise to perfect—the focus today remains on developing small-scale systems and locating markets for future expansion.