Replacement Technologies Needed for Transistor-based Silicon Chips, but when?

Click here to find suppliers

Alternatives must pass the litmus test of feasibility for high-volume production. In the meantime, technologies that solve overheating or allow greater design flexibility are attracting attention.

By Mark Shortt
Editorial Director, Design-2-Part Magazine

According to Moore's Law, based on an observation in 1965 by Intel co-founder Gordon Moore, the number of transistors on an integrated circuit is expected to double approximately every year and a half. Although the technology revolution that followed Moore's prediction is still going strong, it's taken a new turn as conventional methods of fabricating increasingly tiny silicon chips approach their limits. In the race to achieve faster, more powerful chips at increasingly smaller scales, the control of electron flow becomes more problematic, resulting in higher levels of current leakage, power consumption, and heat generation. Industry watchers generally agree that new technologies will be needed to fabricate circuits that could measure just a few atoms in size within the next 10 to15 years.

Influential players within the semiconductor industry have seen the writing on the wall for some time, and have been actively exploring alternative technologies for fabricating circuits with feature sizes under 65 nanometers (nm), 45nm, and below. Materials ranging from nanowires to carbon nanotubes, and methods from self-assembly to nanoimprint lithography are being studied intensively as possible future replacements for the standard lithography techniques that are currently used to make transistor-based silicon chips.

The huge amount of R&D currently being done is a reminder that the innovative spirit of the semiconductor industry, a recognized leader in bringing R&D breakthroughs into the commercial sphere, is alive and well. Today, that innovation is reflected in the design and manufacturing of semiconductor devices for an ever-widening range of markets. For years, corporate computers were the main applications for semiconductors. Although computers remain a significant market, the industry today is getting increasing demand from consumer electronics, such as mobile phones, MP3 players, iPods, gaming devices, and from automotive applications—sensors, actuators, and accelerometers used for everything from airbag deployment to tire pressure sensing. Other industries that are fueling demand include telecommunications, WiFi, medical, aerospace/defense, and industrial process monitoring.

New product development is focused on devices that are not only smaller and faster, but also more powerful and more energy-efficient. The future is wide open for alternative device technologies that one day may or may not include silicon semiconductors. But until the market accepts practical alternatives that prove viable for high-volume manufacturing, silicon will be here to stay. At last fall's Intel Developer Forum, held in San Francisco in September, Intel Corporation president and CEO Paul Otellini told thousands of developers and engineers that advances in silicon technology will deliver new performance breakthroughs in an era of energy-efficient computing.

"The industry is going through the most profound shift in decades, moving to an era where performance and energy efficiency are critical in all market segments and all aspects of computing," Otellini said. "The solution begins with the transistor and extends to the chip and platform levels.

"More than ever processing power matters, even as the need to reduce heat, extend battery life, and reduce electricity costs in data centers becomes more critical," Otellini continued. "Silicon technology is at the heart of the solution. It is how we get there."

Two months later, Intel Corp. ushered in the computer industry's multi-core PC era by delivering the industry's first quad-core processors for PCs and high-volume servers. The two new processor families, consisting of the Quad-Core Intel^® Xeon^® processor 5300 series and the Intel^® Core^TM2 Extreme quad-core processor, feature four computer "brains," inside a single microprocessor. Both families are said to provide outstanding speed and responsiveness for general-purpose servers and workstations, as well as for digital media creation, high-end gaming, and other applications that require high performance.

The Intel^® Core^TM2 Extreme quad-core processor, targeted at gamers and content creators, is said to provide a dramatic 70 percent performance improvement over the Intel Core2 Extreme processor. It incorporates a metal integrated heat spreader (IHS) that spreads heat from the silicon chips and protects them. The IHS serves as a contact for the heat sink and provides additional surface area to permit improved cooling.

A major advance in fundamental transistor design was announced near the end of last month (January 27), when Intel reported that it is using two "dramatically new materials" to build the insulating walls and switching gates of its 45-nanometer (nm) transistors. The company said that hundreds of millions of these microscopic transistors, or switches, will be inside the next generation Intel^® Core^TM 2 Duo, Intel Core 2 Quad, and Xeon^® families of multi-core processors.

Intel is implementing an innovative combination of new materials that is said to drastically reduce transistor leakage and increase performance in its 45nm process technology. The company will use a new high-k material that replaces silicon dioxide for the transistor gate dielectric, and a new combination of metal materials for the transistor gate electrode.

"The implementation of high-k and metal materials marks the biggest change in transistor technology since the introduction of polysilicon gate MOS transistors in the late 1960s," said Intel co-founder, Gordon Moore.

Transistors are tiny switches that process the ones and zeroes of the digital world. The gate turns the transistor on and off; the gate dielectric is an insulator underneath it that separates it from the channel where current flows. The combination of the metal gates and the high-k gate dielectric is said to lead to transistors with very low current leakage and record high performance.

Silicon dioxide has been used to make the transistor gate dielectric for more than 40 years, largely because of its manufacturability and ability to deliver continued transistor performance improvements as it's been made increasingly thinner. According to Intel, it has successfully shrunk the silicon dioxide gate dielectric to as little as 1.2nm thick—equal to five atomic layers—on its previous 65nm process technology, but the continued shrinking has led to increased current leakage through the gate dielectric, resulting in wasted electric current and unnecessary heat.

Transistor gate leakage associated with the ever-thinning silicon dioxide gate dielectric is recognized by the industry as one of the most formidable technical challenges facing Moore's Law. To solve this critical issue, Intel replaced the silicon dioxide with a thicker, hafnium-based high-k material in the gate dielectric, reducing leakage by more than 10 times over the silicon dioxide used for more than 40 years.

"As more and more transistors are packed onto a single piece of silicon, the industry continues to research current leakage reduction solutions," said Mark Bohr, Intel senior fellow. "Our implementation of novel high-k and metal gate transistors for our 45nm process technology will help Intel deliver even faster, more energy-efficient multi-core products that build upon our successful Intel Core 2 and Xeon family of processors, and extend Moore's Law well into the next decade."

Hybrid Silicon Laser May Herald Faster Computers

By offering faster speed, a chip that can emit and guide light could help push silicon photonics into widespread use in the computers and data centers of tomorrow. A giant step toward that possibility was taken last fall, when researchers from Intel Corp. and the University of California at Santa Barbara (UCSB) announced that they had built, using standard silicon manufacturing processes, what they say is the world's first electrically-powered hybrid silicon laser. Their achievement, named one of the top 100 science stories of 2006 by Discover magazine, addresses one of the last major barriers to producing low-cost, high-bandwidth silicon photonics devices for use inside and around future computers and data centers.

The researchers succeeded in combining the light-emitting properties of indium phosphide with the light-routing capabilities of silicon into a single hybrid chip. Thirty-six lasers are built into the silicon chip, a feat that could help future computers move data at the speed of light. As voltage is applied, light generated in the indium phosphide enters the silicon waveguide to create a continuous laser beam that can be used to drive other silicon photonic devices, the researchers say. A laser based on silicon could lead to wider use of photonics in computers, they say, because the cost can be greatly reduced through high-volume silicon manufacturing techniques.

Mario Paniccia, director of Intel's Photonics Technology Lab, said that the achievement could serve to bring low-cost, terabit-level optical 'data pipes' inside future computers and help make possible a new era of high-performance computing applications. "While still far from becoming a commercial product, we believe dozens, maybe even hundreds of hybrid silicon lasers could be integrated with other silicon photonic components onto a single silicon chip," said Paniccia.

"By combining UCSB's expertise with indium phosphide and Intel's silicon photonics expertise, we have demonstrated a novel laser structure based on a bonding method that can be used at the wafer-, partial-wafer or die-level, and could be a solution for large-scale optical integration onto a silicon platform," said John Bowers, a professor of electrical and computer engineering at UC Santa Barbara. "This marks the beginning of highly integrated silicon photonic chips that can be mass produced at low cost."

The novel design of the hybrid silicon laser employs indium phosphide-based material for light generation and amplification; it uses the silicon waveguide to contain and control the laser. According to the researchers, the key to manufacturing the device is the use of a low-temperature, oxygen plasma—an electrically charged oxygen gas—to create a thin oxide layer, measuring approximately 25 atoms thick, on the surfaces of both materials.

When the oxide layer is heated and pressed together, it functions as a "glass-glue" that fuses the two materials into a single chip. When voltage is applied, light generated in the indium phosphide-based material passes through the oxide "glass-glue" layer and into the silicon chip's waveguide, where it is contained and controlled to create a hybrid silicon laser. The design of the waveguide is critical to determining the performance and specific wavelength of the hybrid silicon laser, the researchers say.

While much of today's research and development is aimed at new materials and processes, the ability to leverage existing semiconductor manufacturing methods is a key factor in pushing the industry forward. It's enabling emerging companies like Akustica and Nantero to introduce what many say are breakthrough technologies. Both companies are making use of complementary metal-oxide semiconductor (CMOS) manufacturing processes that are the standard in today's semiconductor foundries.

Single-chip Silicon Microphones Enhance Design Flexibility

Akustica, Inc., a privately held company based in Pittsburgh, has developed a proprietary technology through which PC manufacturers have the freedom to design single-chip digital microphones into notebook PCs in locations where the best audio performance can be achieved. The company's patented Sensory Silicon^TM technology makes use of standard CMOS processes to manufacture, in high volumes, micro-electromechanical system (MEMS) devices that combine the functionality of microphones with microelectronics and software onto a single chip. By enabling the single-chip integration of mechanical transducers with analog and digital circuitry, Sensory Silicon devices comprise a new class of monolithic products that can hear, speak, and sense the world around them.

Beginning with its introduction of the AKU2000 microphone in February 2006, Akustica has launched a family of CMOS MEMS devices—including the AKU 2001 and AKU2004—that are the industry's only single-chip digital microphones. Reported to provide acoustic performance equal to or superior to that of conventional analog microphones, the digital microphones integrate an acoustic transducer, an analog output amplifier, and a sigma-delta modulator on a single chip. Their design flexibility is enhanced by a single-bit digital output stream that connects to downstream electronics without shielded cabling or complex signal routing. Digital output and monolithic design (single-chip construction) also contribute to greater design flexibility by eliminating radio frequency interference (RFI) and electromagnetic interference (EMI).

Because they are immune to RFI/EMI, the chips can be placed in locations that are optimal for acoustic performance, regardless of its proximity to WiFi antennae or other sources of interference. Their small form factor also enables them to be placed in thin-profile areas and tiny products. The digital microphone chips are surface mountable and compatible with automated pick and place systems.

Akustica's first product, the AKU2000 single-chip silicon digital microphone, was recently selected by Fujitsu Computer Systems Corporation for use as an embedded microphone array in its LifeBook^® T4215 convertible Tablet PC. Designed to replace the electret condenser microphones (ECMs) currently used in portable electronic devices, the AKU2000 is said to dramatically improve the quality of voice input in digital and still-video cameras, mobile phones, headsets, and other digital media devices, in addition to notebook PCs. It also earned Akustica a Technology Innovation Showcase award last summer at SEMICON West in San Francisco.

"High-quality voice recording is becoming essential to mobile users as they rely more on their Tablet PC to conduct their day-to-day tasks," said Paul Moore, senior director of mobile product marketing, Fujitsu Computer Systems. "Demands for voice annotating files and interest in applications such as Voice over IP are increasing in vertical markets such as healthcare. By deploying the Akustica AKU2000 microphones in the LifeBook T4215 Tablet PC, Fujitsu can provide a superior voice input solution in a convertible tablet without compromising the form factor."

Factors that led the OEM to choose the AKU2000 reportedly included RFI/EMI immunity, small size, and compatibility with standard CMOS manufacturing processes. Single-chip construction (monolithic design) and digital output, resulting from the integrated sigma-delta modulator, are responsible for the RFI/EMI immunity. No additional circuitry or shielded cabling is required.

The AKU2000 is tiny enough to be embedded in the bezel of the Fujitsu tablet alongside the liquid crystal display (LCD), the optimum position for recording sound on a tablet PC. Also, as a CMOS MEMS chip, the AKU2000 can be manufactured in high quantities with guaranteed uniformity, a factor of critical importance to manufacturers.

Uniform performance between microphones is also important in applications such as the LifeBook T4215, in which noise suppression software is used in conjunction with the microphone array to further enhance the voice input quality.

"With the proliferation of VoIP applications coupled with rapid increase in VoIP users in many markets, manufacturers are recognizing the importance of sound quality," said Rob Enderle, principal analyst, Enderle Group. "As a result, digital microphones such as Akustica's are becoming increasingly critical in laptop designs to better address this growing VoIP user base. These microphones dramatically improve the quality of voice input, therefore improving the performance of VoIP and other voice-enabled applications."

Fujitsu has also embedded the AKU2000 digital microphones in its LifeBook Q2010 and P1610 computing platforms.

According to Akustica, its silicon microphones are smaller and more reliable than the current crop of ECMs, and can be customized with advanced sound capture features and noise reduction capabilities.

"In 2007, we would expect more design-ins in the PC notebook space, complementing the launch of Microsoft^® Windows Vista^TM and the next-generation Intel^® Centrino^TM Pro platform, codenamed Santa Rosa-both of which are optimally designed for VoIP and other voice-enabled applications," said James H. Rock, president and CEO of Akustica, Inc. "Additionally, we are continuing to expand our Sensory Silicon portfolio by using our CMOS MEMS platform to develop new types of products ranging from new and innovative microphone solutions to RF and inertial products."

Industry analyst Marlene Bourne of Bourne Research commented: "With microphones now already designed into notebook PCs, Akustica is well positioned to capture new customers both in and beyond the PC space. By broadening its CMOS MEMS platform for other markets dominated by traditional MEMS sensors, Akustica could gain significant traction within other applications, including automotive, digital cameras, and cell phones."

Next-Generation Devices Could Hinge on Carbon Nanotubes

The ability to effectively apply novel processes and materials in high-volume production environments is expected to go a long way toward determining the future of semiconductor manufacturing. Of all the materials currently being investigated, carbon nanotubes are generating the most buzz. The tiny, cylindrical sheets of graphite are intriguing to researchers because of their unique dimensions, amazing strength (100 times stronger than steel at one-sixth the weight), and excellent thermal and electrical conductivity. They are known to conduct electricity better than copper and heat better than diamond.

Although carbon nanotubes are widely acknowledged to have great potential, their use as a viable material for high-volume semiconductor manufacturing has always seemed a distant reach. The lack of a reliable method for positioning them across silicon wafers, coupled with a susceptibility to contamination that would render them incompatible with clean room requirements of semiconductor fabs, have held them back. But an announcement in November by Nantero, Inc., a nanotech company in Woburn, Mass., may have heralded the start of a new era in which carbon nanotubes begin to play a major role in electronics. That's when Nantero made it known that it had become the first company in the world to introduce and use carbon nanotubes in mass production semiconductor fabs.

The announcement followed several months of working jointly with ON Semiconductor at ON's Gresham, Ore., manufacturing facility to develop process technology that could effectively integrate carbon nanotubes in complementary metal-oxide semiconductor (CMOS) fabrication. With production capabilities down to the 130 nanometer (nm) node, the Gresham facility is well suited to development work in this area. Nantero's partnership with ON Semiconductor is a continuation of joint work that Nantero had undertaken with LSI Logic Corporation, from whom ON purchased the facility last year. When ON acquired the facility, it hired virtually all of the facility's personnel, including most of the in-house engineering team.

"We see tremendous potential in this joint project," said Greg Schmergel, co-founder and CEO of Nantero, in a statement announcing the company's joint development project with ON Semiconductor. "Nantero has successfully worked with the engineering team in Gresham for years, and we look forward to working with ON Semiconductor to make the world's first production semiconductor devices using carbon nanotubes."

Nantero appears well on its way to achieving that goal, having developed a process for mass production of carbon nanotube devices that includes very few steps and only one additional mask layer. As a result, the process is much simpler than processes for manufacturing most other forms of emerging memory, the company says. The key to the process is a method that the firm developed for positioning carbon nanotubes reliably on a large scale. By this method, the nanotubes are treated as a fabric that can be deposited using processes such as spin-coating, and then patterned using lithography and etching. All are processes that can be used in CMOS production fabs.

The U.S. Patent and Trademark Office has issued patents to Nantero on all the steps of the process and on the article of the carbon nanotube fabric itself (U.S. Patent No. 6,706,402, "Nanotube Films and Articles"). The patent relates to the article of a carbon nanotube film, comprising a conductive fabric of carbon nanotubes deposited on a surface.

Nantero has also developed a method for purifying carbon nanotubes to the standards required for use in a production semiconductor fab. Nanotubes purified by the method are said to contain, on a consistent basis, less than 25 parts per billion of any metal contamination. The two breakthroughs—the development of workable methods to reliably position carbon nanotubes and to purify them—have effectively resolved the major issues that had prevented carbon nanotubes from being used in mass production in semiconductor fabs.

As the company works to develop next-generation semiconductor devices, Nantero is focusing most of its efforts on the development of NRAM^TM, a high-density, nonvolatile random access memory chip that the company believes could replace dynamic RAM (DRAM), static RAM (SRAM), flash memory, and, ultimately, hard disk storage. It's a bold vision that Nantero is pursuing: to replace all existing forms of memory with a universal memory chip for semiconductor devices from MP3 players to networking applications. The company envisions a device that stores data permanently, even when not powered on, and enables instantaneous booting and rebooting of computers.

Dr. Thomas Rueckes, Nantero's chief scientific officer, is the inventor of the proprietary NRAM^TM design. By integrating the unique properties of carbon nanotubes with traditional semiconductor technologies, Dr. Rueckes has made it possible to use existing processes to manufacture semiconductor devices that incorporate the awesome benefits of carbon nanotubes.

"The goal is for NRAM to be a universal memory, combining the nonvolatility of flash, the speed of SRAM, and the density of DRAM, all in one chip," says Schmergel. "As such, the potential impact is enormous, as it would allow for substantially greater flexibility in designs and also for new features like true instant-on for computers."

The combination of nonvolatility and high speed is a unique feature of NRAM, which, by enabling instant-on computers, would eliminate the initialization process that occurs during startup. According to Nantero, the chip could be used as a replacement for memory in applications such as cell phones, MP3 players, digital cameras, PDAs, and networking. Revenue potential for a universal memory chip exceeds $100 billion per year, the company says.

"NRAM can be produced as either a stand-alone or embedded memory, since it is compatible with logic processes," says Schmergel. "The stand-alone memory market is far larger, but the embedded memory segment offers a lot of high-value applications. One example of this is replacing embedded SRAM with a fast, nonvolatile memory with 10-times smaller cell size."

Carbon nanotubes are the only material that would work in the NRAM design, Schmergel says, because they have unique properties that include extremely high tensile strength, high electrical conductivity, and high thermal conductivity. Their properties give them potential for use in numerous products, ranging from memory to logic, interconnects, sensors, and many more. "We expect to see quite a wide range of applications for them," said Schmergel.

Nantero is planning to license technology transfer packages to manufacturers that will enable them to produce, market, and sell nanotube-based semiconductor products. The company is seeking to partner with semiconductor manufacturers who are interested in producing either stand-alone or embedded NRAM, Schmergel says. It is also working with licensees on the development of additional applications of its core nanotube-based technology.

"Nantero has developed the full suite of solutions required to use carbon nanotubes in a mass production environment today, and this is being proven every day not in theory, but in practice," said Schmergel. "Our CMOS-grade carbon nanotube formulation and processes for handling them reliably across 200mm silicon wafers is being practiced today and enables the development of our NRAM as well as carbon nanotube-based products from many other companies as well."

MRAM Technology Reaches Commercialization
4 Mbit MRAM Device is now in Volume Production

Another non-volatile computer memory technology is Magnetoresistive Random Access Memory (MRAM), which uses magnetic storage elements, rather than electric charge or current flows, to store data. The first commercial MRAM device went into volume production last July, when Freescale Semiconductor, Inc. (Austin, Tex.) announced the commercialization of its four-megabit (4 Mbit) MRAM product, the MR2A16A. A fast, non-volatile memory device with "unlimited endurance," Freescale's MR2A16A is built on a foundation of technology protected by more than 100 Freescale patents, including toggle-bit switching.

"With the commercialization of MRAM, Freescale is the first-to-market with a technology of tremendous possibilities and profound implications," said Bob Merritt, Semico Research. "Competition to become the first company to market MRAM technology was fierce. This is a significant achievement that certainly confirms the dedication of Freescale's engineering team."

MRAM uses magnetic materials combined with conventional silicon circuitry to deliver the speed of SRAM with the non-volatility of flash in a single, high-endurance device. According to Freescale, its successful commercialization of the technology could hasten new classes of electronic products that offer dramatic advances in size, cost, power consumption, and system performance.

"The commercial launch of the industry's first MRAM product is a major milestone made possible by the pioneering research of Freescale technologists," said Sumit Sadana, senior vice president, strategy and business development, and chief technology officer, Freescale. "It underscores our commitment to deliver breakthrough technology to our customers to address real-world challenges. The unique capabilities of MRAM technology have numerous exciting applications in our target markets."

The MR2A16A is appropriate for a variety of commercial applications, such as networking, security, data storage, gaming, and printers. The part is engineered to be a reliable, economical, single-component replacement for battery-backed SRAM units. The device also could be used in cache buffers, configuration storage memories, and other applications that require the speed, endurance, and non-volatility of MRAM.

Manufactured at Freescale's 200 millimeter Chandler Fab in Arizona, the MR2A16A is a commercial temperature range, 3.3-volt device featuring 35-nanosecond read and write cycle times. It is an asynchronous memory organized as 256K words by 16 bits. An industry standard SRAM pinout arrangement allows for system design flexibility without bus contention. The device is housed in a 400 mil TSOP type-II RoHS package./p>

Industry's First 50-Nanometer NAND Flash Memory Devices under Development

Last July, Micron Technology, Inc., and Intel Corp. announced they were sampling the industry's first NAND flash memory built on 50-nm process technology. The two companies manufactured the samples through IM Flash Technologies (IMFT), a joint development and manufacturing venture formed by Micron and Intel in January 2006. Both companies sampled 4-gigabit (Gb) devices and confirmed their plans to mass produce a range of densities on the 50nm node in 2007.

According to the companies, the move to 50nm process technology will enable them to meet the increasing demand for higher-density NAND flash in a range of computing and consumer electronics applications, from digital music players and removable storage to handheld communications devices. Industry research forecasts peg the NAND flash market segment to reach $13 billion to $16 billion in 2006, before growing to approximately $25 billion to $30 billion by 2010.

"Micron entered the NAND business in 2004 using a 90nm process," said Brian Shirley, Micron's vice president of memory. "In a few short years and through our collaboration with Intel, we are now poised to introduce a leadership product based on a cutting-edge process technology. Micron will continue its commitment to NAND with a rapid transition to the 50nm process and through continued work on advanced nodes for the introduction of even higher-density products."

Micron Technology, Inc., provides advanced semiconductor components for use in computing, consumer, networking, and mobile products. In addition to NAND flash memory, the company manufactures and markets DRAM, CMOS image sensors, and memory modules. One of its sensors, the MT9V112 CMOS image sensor, measures just 4mm diagonally and is a complete camera system-on-a-chip (SOC) that requires only a power supply, lens, and clock source for basic operation. Its small form factor is suited for cell phones, PDAs, and other mobile wireless products.

This report includes news from Business Wire and Market Wire.

Intel Core, Xeon, and Centrino are trademarks or registered trademarks of Intel Corporation or its subsidiaries.

Akustica is a trademark of Akustica, Inc.

Micron is a trademark of Micron Technology, Inc.

Click here to find vendors for this process