Silicon remains a contender in flexible electronics
09/01/2007
As discussed in “High interest in low-end printable electronics” (SST, July 07, p. 34), a number of flexible electronics researchers are focusing on organic semiconductors, and in particular on ink-jet and other printing technologies. Many of them assume that silicon will never match the low cost and mechanical robustness promised by organic semiconductors on flexible substrates.
Yet writing silicon off entirely may be a mistake. As the world’s most successful circuit technology, silicon is backed by enormous manufacturing and research infrastructure. Over the last several decades, it has captured and held many market segments that skeptics doubted it could serve at all. Technologies aimed at silicon’s perceived limitations have been unable to keep up with silicon’s continuous improvement.
One of the reasons for silicon’s success has been its ability to improve performance while cutting customer cost. Moore’s Law turns today’s niche applications into tomorrow’s consumer commodities. Moore’s Law doesn’t apply to many of the devices targeted by flexible electronics, however. When the size of the circuit is defined by the application, as in large area displays, smaller transistors do little to reduce costs. At the other extreme, in very small circuits like RFID tags, the silicon itself is almost free, while packaging and assembly dominate the total cost.
Although cost reduction for these applications is challenging, silicon still sets a performance benchmark that organic semiconductors cannot meet. The best organic semiconductors deliver device mobilities comparable to amorphous silicon, but only for p-channel transistors. Good n-type organic semiconductors remain elusive, and even the best p-channel devices fall far short of what polysilicon and single crystal silicon can achieve. Superior transistor performance means more capable circuits that can potentially serve more demanding applications. Given its long track record and advantage in resources, silicon must be considered the incumbent technology until proven otherwise.
Low temperature deposition
Still, it is not yet clear how to make flexible electronics with silicon technology. The first step, depositing silicon on a flexible substrate, is among the most challenging. In general, the quality of a silicon film increases with deposition temperature. High quality layers of polysilicon or epitaxial silicon demand deposition temperatures near 600°C. Such high temperatures are problematic even for the glass substrates used in conventional flat panel displays: most displays depend on amorphous silicon, deposited at temperatures of around 275°C. Most plastics have glass transition temperatures below 200°C. According to Michael Chabinyc and coworkers at the Palo Alto Research Center, thin film transistors have been made from silicon deposited at temperatures as low as 150°C [1]. These devices had mobility of ~0.5 cm2/V-sec, but were unstable under gate-bias stress.
The simplest way to deposit high quality silicon on flexible substrates may be to choose the correct substrate. Stainless steel and other metal foils are mechanically strong and can tolerate whatever temperatures the process requires. Metal foils are already used as substrates for flexible solar cells in space applications. In solar cells, the metal’s conductivity allows it to serve as an electrode for the panel. Circuits, in contrast, require an insulating barrier layer to both prevent short circuits and stop diffusion of metal ions into the semiconductor. Though this layer does increase the cost of the substrate, it is not a significant obstacle; polymer substrates require even more complex barrier layers to protect the circuit from oxygen and water vapor.
Conventional LCD displays cannot use metal foil substrates. These displays change the transmission of light through the liquid crystal and require a light source or reflector behind the backplane. Flexible LCDs are problematic for other reasons, too, since it is difficult to maintain uniform spacing between the backplane and the viewing surface as the display flexes. Organic light emitting devices, on the other hand, provide their own light. The electrophoretic displays used in electronic paper work by switching beads of pigment between two states, causing them to appear black or white. Neither of these designs necessarily requires a transparent substrate.
A more serious limitation of metal foil substrates may be weight. A 100µm thick steel foil weighs about 800g/m2, compared to 220g/m2 for glass of the same thickness, and only 120 g/m2 for plastic. Weight is especially important for the large area displays expected to be a major application for flexible electronics. Still, metal foils combine durability and flexibility, while remaining compatible with conventional CMOS processing. They are likely to be an especially important alternative for applications that demand both flexibility and performance.
Flexible silicon may also be able to leverage existing display technology because very thin glass is flexible. Unlike metal foil, it requires no barrier layer and is transparent. Most process technologies used for rigid glass displays would still apply to flexible glass. The main challenge is that thin glass is still glass. The safe bending radius is large, about 40cm for a 100µm thick sheet, and is likely to preclude true reel-to-reel processing. Though plastic coatings can reduce the risk of breakage, glass remains the most fragile candidate substrate. Thin glass substrates are likely to be the most appropriate for applications that need transparency, yet require only moderate flexibility in use. For instance, a curved display could use flexible glass to achieve the desired shape, while remaining static after installation.
More exotic approaches separate silicon synthesis from device assembly, transferring single crystal silicon from a wafer to another, more flexible, substrate. For example, researchers at the University of Illinois at Urbana-Champaign transferred silicon strips from an SOI wafer to a thin polyimide substrate, using self-assembled monolayers (SAMs) as molecular adhesion layers (Figs. 1 and 2). These devices achieved mobilities in excess of 500cm2/V-sec, while remaining stable under a 3mm bending radius [2]. Another group, at Harvard, assembled Si nanowires from solution onto pre-patterned gate electrodes, achieving hole mobility of about 200cm2/V-sec [3]. Layer transfer is especially suitable for bottom contact transistor designs, where the silicon strips are placed onto a pre-fabricated array of metal contacts and dielectric layers.
Patterning without optics
A second major challenge for flexible silicon, as for any flexible electronics technology, is patterning. For silicon, conventional projection lithography appears to be one option. After all, silicon is compatible with conventional resists, developers, and etch chemistries, while organic semiconductors are not. However, projection lithography is one of the most expensive steps in conventional CMOS manufacturing. Most flexible electronics applications are unlikely to need the very high resolution it offers, and most will strive to avoid the costs it imposes.
 Figure 2. Strips of single crystal silicon “ink” from SOI wafers. (Courtesy U. of Illinois at Urbana-Champaign) |
Projection lithography also faces practical constraints. For example, the field size for projection lithography is limited by the size of the lens. Larger areas must be stitched together from many exposure fields, introducing alignment errors at the field edges. Large area circuits may require several different reticles to cover the whole area, increasing cost and adding opportunities for error.
Flexible substrates also tend to wrinkle and stretch during manufacturing. The amount of distortion can vary from one field to the next, due to pressure variations across the width of a roll, for instance. Overlay error can vary depending on position, and may not even be consistent within a single exposure field. Conventional projection lithography assumes a flat substrate and cannot easily adapt to these distortions.
For all of these reasons, flexible electronics researchers are considering alternatives to projection lithography. Microcontact printing has been used to prepare substrates for selective growth of high quality organic semiconductors and, as discussed above, to transfer silicon strips to flexible substrates. Integration of microcontact printing with conventional CMOS processes is likely to be challenging, however. For example, the SAMs used as microcontact “inks” are not adequate etch resists by themselves, and may not be compatible with existing silicon resist systems. Selective deposition of silicon from the vapor phase is more challenging than selective deposition of a liquid phase organic semiconductor onto a prepared surface, too. Silicon researchers are also wary of the defect issues inherent in contact printing methods. Silicon’s performance advantage encourages researchers to focus on smaller, more defect-sensitive transistors.
One alternative approach to patterning for flexible silicon electronics uses a laser to write the desired pattern directly into a conventional resist. Though already used in maskmaking, laser direct write patterning may prove to be too slow for volume production of large area circuits. Also, conventional liquid resists pose problems in roll-to-roll processing. It can be difficult to maintain a uniform film thickness as the substrate web passes through a roll system. Dry resists are used in printed circuit board manufacturing, but require several minutes for exposure.
Digital lithography, in contrast, dispenses a resist directly onto the substrate using an ink-jet print head. The resolution of digital lithography is limited by blurring of feature edges as the resist droplets spread out before drying. One possible solution, phase change resist, takes advantage of a temperature difference between the warm print head and the cooler substrate. The liquid resist, usually wax-based, cools and solidifies on contact with the substrate, allowing only minimal droplet spread. One of the important advantages of digital lithography is its ability to adapt to variations in the substrate pattern. The pattern is dispensed one droplet at a time, rather than conforming to a pre-defined mask. Thus, the system can measure the exact placement of the previous layer, and digitally modify the resist pattern immediately before printing [4].
Printed silicon? Maybe…
Printing technologies offer a potential cost advantage, in part because they are additive. CMOS manufacturing is subtractive. It alternates deposition and etch or CMP steps, ultimately removing much of the material deposited on the wafer. Printing processes can combine deposition and patterning in a single step. The greatly simplified process sequence should lead to dramatic cost reductions. Proponents of organic semiconductors argue that their liquid phase materials are amenable to printing processes, while silicon and other materials deposited from the vapor phase are not. The need for vapor phase deposition may set a lower limit on silicon process cost. However, some steps toward liquid phase deposition of silicon devices can be found in the photovoltaic industry.
In photovoltaic manufacturing, weight is an important contributor to the total module cost. Heavier modules require more complex support structures and are more expensive to transport. Flexible cells are lighter and easier to handle, both in manufacturing and in the field. As in the IC industry, eliminating vacuum processing would reduce solar cell costs and simplify the manufacturing process. Accordingly, several companies have been exploring the use of photovoltaic “inks.” The most mature of these, based on nanoparticles of cadmium indium gallium diselenide (CIGS), are now nearing commercialization. Ink-based processing of CIGS is especially appealing because the indium/gallium ratio is critical to performance, and is difficult to control in PVD deposition. Nanoparticles offer better composition control in addition to the advantages of printing.
Innovalight, a Santa Clara, CA, startup, claims to have developed nanocrystalline silicon inks, allowing deposition of silicon from solution. Though their products are aimed at the photovoltaic market, flexible electronics are another potential application.
References
- Michael L. Chabinyc, et al., “Printing Methods and Materials for Large-area Electronic Devices,” Proc. IEEE, Vol. 93(8), pp. 1491-1499, 2005.
- Jong-Hyun Ahn, et al., “High-Speed Mechanically Flexible Single-Crystal Silicon Thin-film Transistors on Plastic Substrates,” IEEE Elec. Dev. Lett., Vol. 27(6), pp. 460-462, 2006.
- Michael C. McAlpine, Robin S. Friedman, Charles M. Lieber, “High-performance Nanowire Electronics and Photonics and Nanoscale Patterning on Flexible Plastic Substrates,” Proc. IEEE, Vol. 93(7), pp. 1357-1363, 2005.
- William S. Wong, et al., “Hydrogenated Amorphous Silicon Thin-film Transistor Arrays Fabricated by Digital Lithography,” IEEE Elec. Dev. Lett., Vol. 24(9), pp. 577-579, 2003.
 |
Katherine Derbyshire is a contributing editor at Solid State Technology. She received her engineering degrees from the Massachusetts Institute of Technology and the U. of California, Santa Barbara. She is the founder of consulting firm Thin Film Manufacturing, [email protected], http://www.thinfilmmfg.com.