An approach that entails printing compound-semiconductor ribbons on a silicon substrate offers the means to build nanoscale transistors that can be switched on and off much more effectively than their bulk analogues. See Letter page 286
For applications in electronics, silicon is often referred to as 'God's material'. Characteristics such as high natural abundance, and relative ease in crystal growth, purification and doping, combine with favourable electronic-transport properties to provide unmatched capabilities for commercial integrated circuits. As a result, silicon has held a dominant position in microelectronics since the early days of the industry, with compound semiconductors, used mainly in radio-frequency devices, consistently a distant second.
Sometime in the latter part of this decade, however, fundamental limitations on the switching speed and energy efficiency of silicon transistors may force a shift to a certain level of diversity in semiconductor materials1,2. One future approach might involve integrating non-silicon semiconductors onto silicon platforms, to yield heterogeneous systems that exploit different types of materials for different functions. On page 286 of this issue, Ko et al.3 report an intriguing route to this goal, which is based on organized arrays of ribbons of indium arsenide (InAs) delivered to silicon wafers in a type of printing process4. Transistors built with such ribbons at nanoscale thicknesses exhibit impressive characteristics, suggesting their potential for enhancing the performance of next-generation silicon electronics.
Compound semiconductors such as InAs are attractive because their extremely high electron mobilities and conductivities lead to transistors that can be faster (up to twice as fast) and more power efficient (up to ten times) than silicon transistors with comparable dimensions2. Although poor hole mobilities (where a hole is a 'missing electron') and lack of high-quality, interfacial insulators will probably prevent their exclusive use in large-scale complementary logic circuits1,2, these materials have potential as strategic additions to silicon-based technologies. The most widely explored means of exploiting compound semiconductors in this fashion involve specialized procedures for growing or bonding these materials on silicon wafers. Although certain research demonstrations are encouraging, such strategies have serious shortcomings, ranging from defects in the materials to challenges in manufacturability. Ko et al.3 present an advanced procedure that avoids these limitations, and they demonstrate their ideas with InAs.
In the first step of the procedure, Ko and colleagues exploit optimized techniques to grow pristine, ultrathin films of InAs on gallium antimonide (GaSb) wafers coated with layers of aluminium gallium antimonide (AlGaSb). Next, the authors pattern the InAs films into narrow, nanoscale-thick strips that they release from the underlying substrate by selectively removing the AlGaSb with a chemical etchant. In a final step, they use a silicone rubber stamp to lift arrays of the nanoribbons from the substrate, and then to deliver them to the silicon dioxide (SiO2) insulator surface of a silicon wafer, in a type of printing process4 in which the InAs serves as the 'ink' (Fig. 1, overleaf). Because the procedure can be used with different types of material, the authors refer to the resulting structure as 'X' on insulator, or XOI, where X represents a semiconductor, by analogy to the widely used acronym SOI for silicon on SiO2/Si substrates.
The printing process used by the authors3 represents a recent and increasingly sophisticated method for transferring nanoscale ribbons, wires and sheets of semiconductors (such as silicon, gallium arsenide, gallium nitride and indium phosphide) from substrates on which they are formed to other surfaces, including those of silicon, glass, plastic and even paper and rubber4,5. Demonstrated applications of the process include electronics integrated with biological systems6, hemispherical 'eyeball' and near-infrared imagers7,8, flexible display and lighting devices9 and photovoltaic modules8. In many of these examples, viscoelastic effects4 and/or specialized relief structures10 on the stamps enable printing of pristine, unaltered material onto bare substrate surfaces, even without any separate adhesive layers. Yields approaching 99.99% are now possible with highly developed tools that also offer micrometre-scale precision in the positions of the printed parts, and throughputs corresponding to millions of printed structures per hour, or more8.
These printing methods are presently in use for the pre-commercial manufacture of photovoltaic modules that incorporate sparse arrays of thin compound-semiconductor solar cells and micro-optics for focusing incident sunlight11. Although the same methods have been suggested for integrating compound semiconductors with silicon4,5,12, Ko and colleagues3 achieve by far the most impressive results in this context, accomplished by using semiconductor-material layers at exceptionally small thicknesses, down to just a few nanometres.
With remarkably clean, adhesiveless interfaces and high-quality, thermally grown oxides, these ultrathin semiconductor layers yield transistors that can be switched on and off much more effectively than their conventional, bulk counterparts. The authors3 describe systematic experimental studies that capture the essential physics of operation of one such type of device, in which an interesting and gradual transition from three- to two-dimensional electronic transport occurs as the thickness decreases from 50 nm to less than 10 nm. Device simulations not only quantitatively capture these trends, but also explain related improvements in switching properties. This match between theory and experiment provides further evidence of the defect-free, predictable nature of the printed material stacks from which the devices are made.
The transistor's performance parameters are highly promising, with electron mobilities that significantly exceed those of silicon transistors of similar design. The behaviour of the device at high switching speeds, however, must be evaluated to determine the potential for enhancing the performance of state-of-the-art silicon platforms. Exploring aspects of operation in this regime and demonstrating interconnected operation with silicon transistors represent directions for future work. Research of this type is appealing because it advances knowledge in both science and engineering, in the context of potential solutions to problems of practical importance. The increasingly ubiquitous nature of electronics in modern society suggests that successful outcomes will have widespread, positive implications.