The July 2022 issue of IEEE Spectrum is here!

Close bar

Bright Silicon LED Brings Light And Chips Closer

Vertical stacking of diode components opens door to on-chip infrared—and, possibly one day, a phone that senses distance

3 min read
color silicon chip
A color brightfield image of a 4um diameter vertical junction LEDs made in silicon. They were fabricated in the global foundries 55nm 55BCDL process.
Courtesy of MIT

For decades, researchers have tried to fashion an effective silicon LED, one that could be fabricated together with its chip. At the device level, this quest matters for all the applications we don’t have on our mobile devices that would rely on cheap and easily fabricated sources of infrared light. 

Silicon LEDs specialize in infrared light, making them useful for autofocusing cameras or measuring distances—abilities that most phones now have. But virtually no electronics use silicon LEDs, instead opting for more expensive materials that have to be manufactured separately.

However, prospects for the elusive, light-emitting, silicon-based diode may be looking up. MIT researchers, led by PhD student Jin Xue, have designed a functional CMOS chip with a silicon LED, manufactured by GlobalFoundries in Singapore. They presented their work at the recent IEEE International Electron Devices Meeting (IEDM).

The chief problem to date has been, to be blunt, silicon isn’t a very good LED material.

An LED consists of an n-type region, rich in excited free electrons, junctioned with a p-type region, containing positively-charged “holes” for those electrons to fill. As electrons plop into those holes, they drop energy levels, releasing that difference in energy. Standard LED materials like gallium nitride or gallium arsenide are direct bandgap materials, whose electrons are powerful emitters of light.

Silicon, on the other hand, is an indirect bandgap material. Its electrons tend to turn that energy into heat, rather than light. That makes silicon LEDs slower and less efficient than their counterparts. Silicon LED makers must find a way around that indirect bandgap.

One way might be to alloy silicon with germanium. Earlier this year, in fact, a group at Eindhoven University of Technology in the Netherlands fashioned a silicon-based laser out of a nanowire-grown silicon-germanium alloy. Their tiny laser, they reported, might one day send data cheaply and efficiently from one chip to another.

That’s one perhaps elaborate approach to the problem. Another has been considered for more than 50 years—operating silicon LEDs in what’s called reverse-biased mode. Here, the voltage is applied backwards to the direction that would normally allow current to flow. This changeup prevents electrons from filling their holes until the electrical field reaches a critical intensity. Then, the electrons accelerate with enough zeal to knock other electrons loose, multiplying the current into an electrical avalanche. LEDs can harness that avalanche to create bright light, but they need voltages several times higher than the norm for microelectronics.

Since the turn of the millennium, other researchers have tinkeredwith forward-biased LEDs, in which electrons flow easily and uninterrupted. These LEDs can operate at 1 volt, much closer to a transistor in a typical CMOS chip, but they’ve never been bright enough for consumer use.

The MIT-GlobalFoundries team followed the forward-biased path. The key to their advance is a new type of junction between the n-type and p-type regions. Previous silicon LEDs placed the two side-by-side, but the MIT-GlobalFoundries design stacks the two vertically. That shoves both the electrons and their holes away from the surfaces and edges. Doing that discourages the electrons from releasing energy as heat, channelling more of it into emitting light.

“We’re basically suppressing all the competing processes to make it feasible,” says Rajeev Ram, one of the MIT researchers. Ram says their design is ten times brighter than previous forward-biased silicon LEDs. That’s still not bright enough to be rolled out into smartphones quite yet, but Ram believes there’s more advances to come.

Sonia Buckley, a researcher at the U.S. National Institute of Standards and Technology (NIST) who isn’t part of the MIT-GlobalFoundries research group, says these LEDs prioritize power over efficiency. “If you have some application that can tolerate low efficiencies and high power driving your light source,” she says, “then this is a lot easier and, likely, a lot cheaper to make” than present LEDs, which aren’t integrated with their chips.

That application, Ram thinks, is proximity sensing. Ram says the team is close to creating an all-silicon system that could tell a phone how far away its surroundings are. “I think that might be a relatively near-term application,” he says, “and it’s certainly driving the collaboration we have with GlobalFoundries.”

The Conversation (0)

3 Ways 3D Chip Tech Is Upending Computing

AMD, Graphcore, and Intel show why the industry’s leading edge is going vertical

8 min read
A stack of 3 images.  One of a chip, another is a group of chips and a single grey chip.
Intel; Graphcore; AMD

A crop of high-performance processors is showing that the new direction for continuing Moore’s Law is all about up. Each generation of processor needs to perform better than the last, and, at its most basic, that means integrating more logic onto the silicon. But there are two problems: One is that our ability to shrink transistors and the logic and memory blocks they make up is slowing down. The other is that chips have reached their size limits. Photolithography tools can pattern only an area of about 850 square millimeters, which is about the size of a top-of-the-line Nvidia GPU.

For a few years now, developers of systems-on-chips have begun to break up their ever-larger designs into smaller chiplets and link them together inside the same package to effectively increase the silicon area, among other advantages. In CPUs, these links have mostly been so-called 2.5D, where the chiplets are set beside each other and connected using short, dense interconnects. Momentum for this type of integration will likely only grow now that most of the major manufacturers have agreed on a 2.5D chiplet-to-chiplet communications standard.

Keep Reading ↓Show less