Plastic LEDs deliver more light, applications about to boom
03/01/2001
Plastic LEDs deliver more light, applications about to boom
Results from work at the University of Arizona (UA) show that electrically conducting plastic (that is, the material in organic or plastic light-emitting devices, i.e., OLEDs) can convert 45% or more of electricity into light. Until now, the maximum efficiency was thought to be about 25%. The work is being led by UA physics professor Sumit Mazumdar and was recently reported in the journal Nature. Mazumdar's collaborators include Valy Vardeny of the University of Utah (UU), UU postdoctoral physicist Markus Wohlgenannt, and S. Ramasesha and Kunj Tandon at the Indian Institute of Science in Bangalore, India.
While most of today's microelectronics is based on conventional inorganic semiconductors, over the past 15 years, several laboratories and research institutions have focused on organic molecular and polymeric semiconductors for optoelectronics. Mazumdar told Solid State Technology, "This field has become an extremely multidisciplinary field with scientists from physics, chemistry, optics, and materials science working together. In fact, the 2000 Nobel Prize in chemistry was awarded to researchers in this area who demonstrated that the polymer polyacetylene could be doped so that its conductivity increases a trillion times, reaching the conductivity of copper."
 |
Schematic of a "plastic" light-emitting device showing attached electrodes (upper left). The resulting light emission (green arrows) emerges through the transparent electrode. The enlarged view of the device interface shows positively (brown) and negatively (green) charged molecules. At the interface there are molecules in excited states — excitons — with both charges. When excitons come back to ground state, they give off light. Insert shows an actual plastic light-emitting device. (Source: Mazumdar, University of Arizona, Polymer Optics Group)
Now, emphasis in such research has shifted from doped conducting polymers to undoped semiconducting plastics that absorb visible light. These plastics are cheap, processable, lightweight, and flexible. "Equally important," said Mazumdar, "the optical gaps of these plastics and the color of light they absorb or emit can be tuned by simple chemical substitutions."
Masumdar has been working to find niche applications for semiconducting plastics. One is for LEDs that use organic electroluminescent molecules or polymers. In the simplest configuration, electrodes are attached on both sides of a thin organic film and pass current through the material (see figure). Electrons and holes are thereby injected into the system. These jump from molecule to molecule until an electron and a hole both find themselves on the same molecule and form an exciton — the excited state of the molecule. The exciton then radioactively decays and emits light.
The efficiency of the device depends on how much light output is obtained per unit of charge injected, or the number of excitons formed. However, not all excitons emit light, which reduces the efficiency. Previous research has shown that only one fourth of the injected electrons will emit light, setting a maximum efficiency of 25% for the polymer LEDs.
Mazumdar and his colleagues have shown through both theory and experiment that the previous research is not correct and that the measurement of efficiency could be 45% or higher. "If larger amounts of current can be injected into the system, organic solid-state lasers will become feasible," says Mazumdar. "In fact, a group from Bell Laboratories has already made such injection-driven lasers and this is one of the possibilities driving this field."
On the commercial side. Increasingly, OLEDs are being put into commercial applications. For example, in Japan, Pioneer and Toyota have a joint venture producing automobile stereos that use molecular blue-green OLED displays. Also, engineers at eMagin Corp., Hopewell Junction, NY, have developed a full-color microdisplay, and Philips has built a 3000-employee plant in The Netherlands to develop and commercialize OLED and PLED-based displays in a licensing agreement with Cambridge Display Technology (CDT). CDT itself has invested in a UK manufacturing plant to commercialize light-emitting polymers (LEPs) for integration into OLEDs.
David Fyfe, CEO at CDT said, "We have spent the last five years developing the technology base to enable the commercialization of LEP displays. Several of our licensees have announced that they will launch commercial products in 2001."
In another example of extensive development work behind this technology and the rapid emergence of production capability, Universal Display Corp. (UDC), Ewing, NJ, is building on its stacked and transparent (SOLEDs and TOLEDs) technologies (see "A description of TOLED and SOLED technology"). UDC has been developing OLED technology with Princeton University and the University of Southern California since 1994.
Sidney Rosenblatt, executive VP and CFO of UDC, said, "This year, we increased our technical staff, expanded our facilities, and acquired the license for more than 70 OLED patents from Motorola Inc., increasing our OLED patent portfolio to more than 103 patents, with an additional 50 patents pending. In addition, we entered into strategic relationships with PPG Industries and AIXTRON A.G."
According to research by DisplaySearch, OLED display revenue in 2000 was $24 million, up by 100% compared to 1999; it is expected to grow to $3.3 billion by 2005. Mazumdar said, "This will be an exciting area to watch. Plastic LEDs are much more promising than previously thought, lasers are a distinct possibility, and, while it will take a few more years, it will be a multibillion dollar industry."
EUV takes NGL lead, but huge problems remain to be solved
A panel of experts told executives at January's ISS 2001 in Pebble Beach that EUV (extreme ultraviolet) has taken the lead as the next-generation lithography (NGL) method most likely to succeed, perhaps at the 50nm node. But all of them peppered their talks with dozens of unsolved problems for NGL technologies and for 193nm and 157nm optical lithography.
"There are no pellicle solutions for 157nm or EUV," commented Gerhard Gross, director of lithography at International Sematech, who is on loan from Infineon. He also cited severe challenges with mask repair and with mask and edge overlay.
"It appears too risky to develop a whole technology solution for only one node," Gross said, suggesting 157nm might be used for devices at the 70nm node and for some features at 50nm, while EUV could start at 50nm and be useful at 30nm and beyond. He cited the very high costs that might be associated with EUV, especially the optics and scan stage, and for operating in vacuum, particularly mask handling under vacuum with no pellicle. Perhaps at 50nm, Gross commented, we may go to maskless lithography to eliminate the need for a maskmaking infrastructure.
Adding to the urgency of finding solutions, Gross pointed out that the traditional three-year gap between technology cycles has turned into a two-year cycle, which he expects to continue to the 70nm and maybe 50nm nodes.
Responding to the call for a maskless technology, Philip Ware of Canon's Semiconductor Equipment Division revealed that his company is working on a multi-e-beam, direct-write (MEBDW) system for below 50nm, with aberration correction for space charge effects. He also said that Canon already has a production-ready x-ray system. It has patents on multilayer thin films for optics that may be applicable to EUV, and is now in its fifth generation of aspheric polishing technology. Canon has developed a vacuum-operated stage, Ware said.
Nikon is also exploring alternate NGL technologies, according to John Weisner, senior VP, engineering and technology. He expects it will have an early learning electron projection lithography (EPL) tool by 2002, and a production tool by 2004. "EUV is further down the line, probably after 70nm," he commented.
Both 157nm optical lithography and EPL are being developed in parallel. Only illumination power limits throughput for a 157nm tool using an F2 laser, Weisner said. Suitable resists aren't expected until 2002; there is no pellicle; and optics offer an engineering challenge. It would probably be introduced at the 70nm node, already using sub-wavelength reticle enhancement techniques (RETs), and a fast ramp would be needed. It might only be a one-generation technology, Weisner pointed out, raising ROI questions.
Initial tools would be designed to process 60 300mm wafers/hr. The industry may need to go to 5x rather than 4x reduction to cope with mask error enhancement factors, he said, but this decision will have to be made very soon. He expects NA to be >0.8 for 22mm x 26mm fields.
The Scalpel-based EPL system would have unlimited resolution, down to angstroms, with a huge DOF and process margins. Masks would be an extension of membrane technology. This system would likely be used mix-and-match with ArF or F2 tools, with EPL used for vias and contacts. This technology faces mainly engineering problems, he said, and it will be hard to raise production throughput, which looks like about 30 wafers/hr. Cooling will be needed because lots of electrons heat a wafer, and stitching errors must be eliminated. He expects this technology will be used for shorter run ICs.
Paul Van Attekum, VP of ASML, said that ArF, 193nm tools would be ready for the 100nm node, and F2, 157nm tools for 70nm. Many RETs will be needed to lower k1 values, but specific solutions will be needed for each application since there is no generic solution. He pointed out that 248nm tools hit a real brick wall at 75nm. He said there was a narrowing window for NGL techniques, and global agreement will be needed to reduce risk for developers.
Although EUV offers good throughput, is extendible to smaller features, and has a growing consensus, Van Attekum suggests there needs to be better alignment between the US and Japan. Many challenges still remain. With no pellicle, mask blank contamination is a problem, and source power needs to be high enough for production throughput. A few companies feel they can develop sources giving 30-40W output, but 50-150W in-band, collected x-rays will be needed for 80 wafers/hr, he said. He also pointed out that it has taken four years to get suitable resists for 193nm litho.
Resists will be tougher for 157nm, according to Robert Allen of IBM Almaden Research Center's resist group, than for EUV or EPL, which can use chemical amplification. "It is hard to break away from the industry's phenol legacy," Allen said, adding, however, that three new non-phenol polymer families have already been developed for 193nm. For 157nm, he suggested that third-generation materials with rapid development cycles would be needed. It is a challenge to come up with new formulations meeting transparency, resolution, and etch resistance requirements, he explained.
We are learning about thin-film imaging, Allen said, with films 100-150nm thick rather than the traditional 500-600nm. For 157nm, we may need bi-layers, he suggested. For EUV, the thinner films could do the job, but we need to find out which factors influence resolution. They will need to be lightning fast while keeping all other parameters in order.
Bare die play key role in world's smallest robot
Researchers at the US Department of Energy's Sandia National Laboratories, Albuquerque, NM, are developing what may be the world's smallest autonomous, untethered robot. It is only 0.25 in3 and weighs <1oz; it's powered by three watch batteries and rides on tank-track-like wheels (see figure on p. 40). It consists of an 8K ROM processor, temperature sensor, and two drive motors. Enhancements being considered include a miniature camera, a microphone, communication devices, and chemical microsensors.
Sandia's Ed Heller, one of the project's researchers, said, "This could be the robot of the future. It may eventually be capable of performing difficult tasks that are done with much larger robots today, such as locating and disabling land mines or detecting chemical and biological weapons." Heller envisions the tiny robot scrambling through pipes or prowling around buildings looking for chemical plumes or human movement.
"In packs, these robots may be capable of relaying information to a manned station and communicating with each other. They will be able to work together in swarms, like insects. The miniature robots will be able to go into locations too small for their larger relatives," said Heller. One of these minirobots has already maneuvered through a field of dimes and nickels (it can park easily on a nickel), traveling at ~20 in/min.
This development work is rooted in Laboratories Directed Research and Development (LDRD) work started in Sandia's Intelligent Systems Sensors and Controls Department. In 1996, the department unveiled a mini autonomous robot vehicle (MARV), a 1-in3 robot that contained all the necessary power, sensors, computers, and controls on board. It was made primarily from commercial parts using conventional machining techniques. Over the next several years, the department improved MARV, making robot bodies out of PC boards and adding an obstacle detector sensor, radio, temperature sensor, and batteries, but he was still too "big." About three years ago, the Intelligent Systems and Robotics Center teamed with Sandia's Sensor Technologies Department with its expertise in building miniature sensors and devices, to further miniaturize robots.
"By trying new techniques at packaging electronics, wheel design, and body material, our team of researchers shrunk robots to the current size," said Heller, who developed the necessary microelectronics. "One significant innovation behind the current design has been the use of commercially available unpackaged microelectronics. By eliminating packaging, using components in die form, we reduced the size of the robot's electronics considerably," said Heller. The unpackaged parts are assembled onto a simple multichip module on a glass substrate.
Doug Adkins, who developed the mechanical design for the new minirobot, said, "We further reduced its size by using a new rapid prototyping technique to form the device's body. Called stereolithography, the material-building method lays down a very thin polymer deposit that is cured by a laser. The material, which 'grows' as each layer is added, is lightweight, strong, and can be formed in complex shapes." Using this technique, robot bodies have cavities for batteries, electronics-embedded glass substrates, axles, tiny motors, switches, and other parts.
 |
Sandia National Laboratories researcher Doug Adkins takes a close-up view of the minirobots he and Ed Heller are developing. At 0.25 in3 and weighing ~ 1 oz, these are possibly the smallest autonomous robots ever created. (Photo by Randy Montoya)
Adkins also redesigned the wheel structure of the device. Earlier models had standard wheels; mobility, however, was limited. "I thought of how tanks with their track wheels can maneuver over large objects and realized the minirobots could benefit from the same type of wheels," said Adkins.
The ultimate size of the miniature robots is primarily limited by the size of the power source — watch batteries. The body must be large enough to hold batteries to support power requirements of the robot. For further development, Heller said, "The batteries need to run longer and be smaller."
Over the next few years, with additional help from other Sandia groups, Heller and Adkins expect to add to their minirobot design either infrared or radio wireless two-way communications capability, as well as a miniature video camera, a microphone, and chemical microsensors. — P.B.
Low-k dielectric advances from TSMC and Applied Materials
TSMC has selected Applied Materials' Black Diamond CVD low-k dielectric for its 0.13µm copper dual damascene production needs. The announcement came with a demonstration of an eight-level copper process. The production ramp is expected in the second half of 2001, in both 200mm and 300mm fabs.
TSMC's decision came after two years of joint development with Applied, according to Farhad Moghadam, VP and GM of Applied's dielectric systems and modules product group. Moghadam emphasized the significance of this work being done at a foundry as opposed to an IDM (integrated device manufacturer), citing the need for foundries to meet a wide range of needs while having little if any control over the design. "They have a much tougher job than the IDMs, accommodating 50 different customers. That's not an easy task," said Moghadam. For example, a low modulus or weak dielectric layers can require dummy vias to act as pillars to strengthen the structure. An IDM can design this into the product, but a foundry would not be able to require its customers to do that. Foundries have fewer options for fixing marginal processes.
One of the biggest recent challenges in the industry has been the integration of low-k dielectrics into production-worthy process flows. Applied and TSMC worked together to optimize lithography and etch chemistry to allow patterning of the films with no etch stop between the trench and via. This innovation lowers the effective k by 0.3, according to Moghadam. The thermal-mechanical properties of Black Diamond also performed well in various evaluations. A critical one was temperature and humidity testing in plastic packages. This is where organic and inorganic materials might behave differently, with organics at risk for taking in moisture in the nonhermetic package. Moisture uptake would increase the dielectric constant of the material, and speed testing of devices before and after T&H testing showed that this did not happen with Black Diamond. Mechanical integrity during CMP and wire bonding were also evaluated.
The evaluations were done with Black Diamond as the dielectric for all of the layers in the interconnect structure. Other approaches require a low-k film between metal lines in the metallization layers, but then have a stronger and higher thermal conductivity dielectric between the metal layers. The inorganic Black Diamond material is robust enough and transfers heat well enough to serve as both the interlayer and intralayer dielectric. Effective heat transfer is critical for keeping the temperature low enough to prevent electromigration failures.
Current work by Applied and TSMC is targeting 0.10µm technology with Black Diamond. People have been skeptical about a CVD material reaching and going below a k of 2.5, but research with IMEC has resulted in processes that take Black Diamond to that level and below. It is widely agreed that some kind of porosity is needed to get k<2.5. Ideally, the voids would be as small as possible, all the same size, not connected, and uniformly distributed in the dielectric. According to Moghadam, Applied's work with IMEC has accomplished exactly this by finding a way to remove individual carbon atoms from the material, leaving behind atomic scale voids in a very repeatable fashion. More details will be revealed at the International Interconnect Technology Conference in June. Also for 0.10µm technology, Applied will be using a barrier film — "BLOk" — with a lower k (~4.5) to replace the nitride layer (k = 7) used now. — J.D.
 |
New stacked chip-scale package
ST Assembly Test Services (STATS), Singapore and Milpitas, CA, has developed and introduced a "near" chip-scale package (CSP) dubbed stacked-die ball-grid array (SDBGA). Typical SDBGA package height is 1.4mm, with popular 8mm x 8mm to 14mm x 14mm sizes and pin counts between 80 and 140. This package targets significantly reducing manufacturing costs, test time, and circuit board real estate for wired and wireless communications products.
 |
a) Inside the SDBGA package, showing two die (stair-like, dark gray surfaces, blank silicon for tests) and connections to the package (lower right). b) Schematic of the SDBGA package. Die thickness = 0.20mm; die adhesive thickness = 0.0254mm; and substrate thickness = 0.26mm. (Source: STATS)
B.J. Han, chief technology officer at STATS, said, "Both the mounting area and weight for a SDBGA are as much as 70% less, compared to conventional packages. The SDBGA's multidie application easily meets current market demand for communications devices where increasingly more capability is packed into a smaller size, [which] cost less, and are easier to produce and offer faster time to market than a single-chip solution."
It is likely that this package will be used in hand phones for wireless communications, which contain both flash and SRAM ICs. By stacking these on top of one another in an SDBGA package (see figure below), chip real estate is reduced while system capability is enhanced. In another example, the SDBGA offers an ideal configuration for emerging memory and logic combinations in one package; memory density and electrical performance are improved while reducing the package and testing costs that can occur prior to installation into a product. — P.B.
400mm wafers ahead?
Usage of 400mm wafers is now seen arriving around 2014, according to Japan's Super Silicon Crystal Research Institute (SSI), which provided an update on 400mm wafer technology at the recent 3rd International Symposium on Advanced Science and Technology of Silicon Materials in Kona, Hawaii.
More than ten 400mm wafer-related papers discussed 400mm wafer technology during the symposium; several 400mm wafers (with both sides polished) were on display as well.
SSI, the 400mm silicon crystal and wafer R&D consortium formed in 1996 (with the help of funding from Japan's government), will end operations next year as planned. According to SSI, 400mm activities have been delayed. In 1996, SSI officials expected 400mm wafer usage by 2008, but the group now expects to see the 400mm substrates come to market in 2014.
SSI's Yamagishi reported that the group has developed large silica crucibles and graphite hot-zone parts for 400mm silicon crystal growth. The group was able to "grow 400mm silicon single crystal weighing 438kg and dislocation free along a full body length of 110cm," Yamagishi said. "The outlook for fundamental technology in 400mm is quite optimistic except for the cost issue. We had to develop infrastructure for these silica and graphite materials and had to use greatly expensive hot-zone parts. We have to solve cost issues in the future."
SSI's Hagino reported on the present status of 400mm crystal slicing, grinding, polishing, and cleaning. "In wiresaw slicing, global wafer flatness and warp level were 19µm and 21µm, respectively," Hagino said. For surface grinding, Hagino reported global wafer flatness of 51µm; for double disk grinding, global wafer flatness was <1µm.
In final cleaning, particles over 80nm in diameter and metallic concentration levels were 60 counts/400mm wafer, and below 5 }mp 108 atoms/cm2, respectively.
The group also reported success in growing an epitaxial layer on a 400mm wafer by using newly developed equipment that uses a low-temperature SiH4 gas process.
___________________________________
A description of TOLED and SOLED technology
The Universal Display Corp. (UDC) TOLED (transparent organic light emitting device) is a monolithic, solid-state device consisting of a series of small molecular weight organic thin materials (typically smaller than 1000 molecular weight; see figure a) as thin films sandwiched between two transparent, conductive layers (see figure b). These are used for bright, self-emitting displays that can emit from either or both surfaces. This is possible because, in addition to having transparent contacts, the organic materials are also transparent over their own emission spectrum and throughout most of the visible spectrum.
 |
The UDC TOLED (transparent organic light-emitting device) is a monolithic, solid-state device consisting of a series of a) small molecular weight organic thin materials (typically smaller than 1000 molecular weight) as b) thin films sandwiched between two transparent, conductive layers. (Source: Universal Display Corp.)
Typically, a transparent conductive material (e.g., indium tin oxide or ITO) for hole-injection is deposited directly onto a glass substrate. Then, a series of organic materials are deposited by vacuum sublimation on the ITO layer. The first organic layer serves as a hole-transporting layer (HTL) and the second layer serves as both a light-emitting (EL) and electron-transporting layer (ETL). Finally, a UDC proprietary transparent top contact is deposited for electron injection by vacuum evaporation or sputtering on top of the organic films. When a voltage is applied across the device, it emits light based on the exciton luminescence phenomenon.
A SOLED (stacked organic light-emitting device) display consists of an array of vertically stacked TOLED sub-pixels. To separately tune color and brightness, each of the red, green, and blue sub-pixel elements is individually controlled. By adjusting the ratio of currents in the three elements, color is tuned. By varying the total current through the stack, brightness is varied. By modulating the pulse width, grayscale is achieved. With this SOLED architecture, each pixel can, in principle, provide full color. According to UDC, this is the first demonstration of a vertically integrated structure where intensity, color, and gray scale can be independently tuned to achieve high-resolution full-color.