Last week, at the semiannual Intel Developer Forum in San Francisco, chip-maker Intel announced a transistor made from a material called indium antimonide (InSb) that had some impressive stats: it was clocked at 1.5 times the speed of silicon-based transistors and used one-tenth the power.
According to Intel’s director of technology strategy, Paolo Gargini, who presented the results, a shift from silicon might be crucial for the chip-making industry, so it can build smaller devices over the next couple of decades. As transistors made of silicon keep shrinking, the material’s limitations are becoming more apparent. “Silicon is not the best semiconductor,” Gargini says.
But of course silicon is both highly prevalent and relatively inexpensive, and its manufacturing process has been honed for 30 years. What makes so-called “compound semiconductors” -– those made out of more than one element, such as indium antimonide -– so attractive is their special electrical and optical properties.
Electrons can pass through an indium antimonide crystal 50 times faster than through a silicon crystal, Gargini says. As a result, not only are electronic operations significantly faster, but less power is needed to push the electrons.
Compound semiconductors also have optical properties that could help speed up communication between transistors on a chip and multiple chips within a device. These materials easily emit and detect light -– a characteristic that has been studied and improved for decades, says David Hodges, electrical engineer at the University of California, Berkeley. Therefore, he says, light emitters and detectors made of compound materials could potentially replace copper wires, which are a major “impediment of speed.”
Compound materials also have their disadvantages, though. Currently, hundreds of billions of transistors are manufactured at a time on top of silicon wafers that can be as large as 12 inches in diameter. The crystals of compound materials, such as indium antimonide (InSb), gallium arsenide (GaAs), indium arsenide (InAs), and indium gallium arsenide (InGaAs), however, tend to break apart easily, and so can’t be made into such large wafers, says Gargini. This means that compound materials could never completely replace silicon as the wafer base for electrical devices, he says.
Instead, “islands” of InSb transistors must be deposited on the large-diameter silicon substrate. But depositing indium antimonide transistors onto silicon creates an additional challenge. The atoms in a silicon crystal are spaced 0.543 nanometers apart, while the atoms in indium antimonide are 0.648 nanometers apart. Because of this mismatch, when the two materials are placed next to each other, not all of the atoms at the interface bond together, resulting in ineffective devices.
The way to overcome this, Gargini explains, is to add thin layers of “buffer” materials on the silicon that have an atom spacing similar to it, then gradually adjust the chemical compositions of the buffer layers, until they have atom spacing similar to that of indium antimonide. Finding the ideal chemical ratios to provide the best buffer layers will be one of the major challenges to integrating indium antimonide on Intel’s existing silicon platform, he says.
In addition to finding the best buffer for the InSb “islands” on the silicon wafer, according to Jesus del Alamo, an electrical engineer at MIT who specializes in microelectronics, engineers must also consider the insulating layer, the “gate dielectric,” on top of the transistor, which is crucial to the electrical operations of the device. Currently, silicon transistors use a layer of silicon dioxide as the gate dielectric. For compound semiconductors, though, silicon dioxide does not work as an insulating material, says del Alamo. The interface quality between compound semiconductors and silicon dioxide is not good enough and the dielectric constant of silicon dioxide is too small. Therefore, a whole new class of high-quality gate dielectrics will need to be developed. “That will be a huge challenge,” he says.
Del Alamo still believes, however, that such hurdles will be overcome as the field matures. “I’m very optimistic that we’ll come up with these breakthroughs,” he says.
Intel’s Gargini expects that the technology, which Intel began researching about three years ago, will move toward manufacturing in about another decade. He also emphasizes that compound semiconductors are only one of a number of possibilities for future microprocessors. In fact, Intel has many ideas in the works, from extreme ultraviolet lithography, for making silicon transistors smaller, to developing silicon lasers, modulators, and detectors, in which beams of light instead of copper wires could be used to transmit data within a chip (see “Intel’s Breakthrough”). “Don’t expect [compound semiconductors] in a product tomorrow,” Gargini says. “But it’s in the pipeline.”
Home page image courtesy of Jesus del Alamo, electrical engineer, department of electrical engineering and computer science, MIT. Caption: Chip with transistors and test structures made of the compound semiconductor indium gallium arsenide (InGaAs). Chip is used to diagnose device operations.
AI is here. Will you lead or follow?
Join us at EmTech Digital 2019.