High-Performance Flexible Silicon
A new way to make bendable high-speed silicon devices could result in advanced circuits on virtually any surface.
The same high-quality form of silicon that is used inside many new computers could soon be rolled up on a sheet of plastic. Researchers from the University of Wisconsin, Madison, have shown that the type of high-speed silicon used for the past few years in Intel’s microprocessors, called “strained” silicon, can be made thin enough to be transferred to a flexible substrate.
The ability to put sheets of strained-silicon transistors on malleable materials could lead to high-quality flexible displays and solar cells – or eventually even to improved prosthetics, or computerized clothes, according to the researchers.
For the most part, flexible electronics are made of organic polymers, which, although bendable, produce relatively poor performance. So researchers have been giving silicon – the standard material in electronics – a second look as a way to make pliable circuits (see “Stretchable Silicon”).
Although silicon is usually brittle, it can bend when thin enough. In particular, the Wisconsin researchers set out to make flexible forms of strained silicon, a type of high-performance silicon recently commercialized by Intel. Electrons move through strained silicon 80 percent faster than in conventional silicon, and transistors switch on and off up to about 30 percent faster.
Until now, however, strained silicon has been far too bulky – micrometers thick – to flex. The Wisconsin researchers, led by Max Lagally, professor of materials science, found a way to thin out the material to a couple of hundred nanometers, as well as to effectively remove it from a silicon wafer, allowing its use in flexible and high-speed electronics. (Their work is described in a recent issue of Journal of Applied Physics.)
Strained silicon is typically fabricated using multiple layers of a material called silicon germanium, which has larger spaces between its atoms than silicon. Each layer of silicon germanium is chemically altered to gradually introduce more space between the atoms. Finally, a thin layer of silicon is deposited on top. When the silicon atoms (naturally spaced closer together than the silicon germanium atoms) contact the top layer of silicon germanium, they “strain” to bond to it. “If you strain the silicon lattice, then you can improve the electron mobility and performance in your device,” says John Rogers, professor of material science at the University of Illinois, Urbana.
But by using multiple layers of silicon germanium, the device becomes too thick to bend. To make the strained silicon thin enough to flex, Lagally and his team first start with a silicon wafer with two additional layers on top: a silicon oxide layer and a thin layer of silicon. On top of the thin layer of silicon they apply only one layer of silicon germanium, just 150 nanometers thick. Since the silicon layer below the silicon germanium is fixed, the silicon germanium atoms, while spaced wider than silicon, squeeze together, compressing to conform to the silicon layer below it. Then, the researchers add a thin layer of silicon on top of the silicon germanium, forming a sandwich 250 nanometers thick.
At this point, Lagally explains, there’s no strain in the silicon; there’s only compression in the silicon germanium. In order to add strain, the sandwich is removed from the silicon wafer by bathing it in hydrofluoric acid, which eats away at the silicon oxide – the layer connecting the sandwich to the wafer. Once the device is free, the spacing of the atoms in all the layers adjusts slightly: the silicon germanium atoms, formerly compressed, loosen up, and the silicon atoms, which had normal spacing before, developed strain.
Removing the device from the wafer not only adds strain to the silicon, but also allows it to be transferred to another material, says Lagally. From here, the device is pressed into a flexible material, which it sticks to with the aid of special glue.
Sigurd Wagner, professor of electrical engineering at Princeton University, says the work is “a well-executed example of transferring high-quality devices to a low-quality substrate.” Importantly, he says, it proves that strained silicon retains its properties after the transfer process, something that hadn’t been shown before. In addition, says Rogers, the process could be applicable to most inorganic materials, from strained silicon used in microprocessors to gallium arsenide transistors in light-emitting diodes.
Lagally expects that this type of flexible, high-speed silicon will find its way into commercial products within a few years, most likely initially in flexible imaging systems and high-quality displays.
Become an MIT Technology Review Insider for in-depth analysis and unparalleled perspective.Subscribe today