© Gorodenkoff #165717610, source:stock.adobe.com 2020
The EU-funded SiLAS project has overturned the long-held notion that silicon, the abundant elementary building block of all commercial computer chips, is incapable of emitting light efficiently. By changing the atomic structure of a silicon germanium (SiGe) alloy from a typical cubic shape into a novel hexagonal form, the researchers have been able to develop an innovative material for fabricating silicon-compatible lasers to transmit data quickly and efficiently.
For decades, it has been the holy grail of the semiconductor industry to demonstrate light emission out of silicon, but nobody had succeeded until now, says SILAS project coordinator Jos E.M. Haverkort at Eindhoven University of Technology in the Netherlands.
The fundamental breakthrough in the SILAS project is that SiGe, which is mainstream in electronics today, has been shown to provide very efficient light emission when converted to a hexagonal crystal form.
Integrated into a computer chip, the hexagonal silicon germanium, or Hex-SiGe, technology would revolutionise the way processor cores are connected. It would use light from miniature nano-scale lasers to transmit data instead of energy inefficient metal wiring that slows data-transfer rates. This means your laptop or smartphone could operate much faster and for far longer on battery power alone, while also dissipating much less heat.
The SiLAS technology would also enable a scaling up of high-performance computing infrastructure, and help the semiconductor industry overcome the energy, heat and size obstacles that have undermined Moores Law over the past decade as the pace of chip performance improvements using conventional silicon technology has slowed.
Haverkort points out that silicon-based photonics circuitry could achieve energy dissipation below one femtojoule (one quintillionth of a joule) per bit of data transferred. That is at least 100 times less than conventional connections, which can dissipate as much as 100 watts of energy as heat over just a millimetre-long metal interconnecting wire, once data-transfer rates reach one petabit per second.
High efficiency, low cost
Because silicon chips are so well established and cheap to produce at scale, the integration of Hex-SiGe photonics would also open pathways to developing small, energy efficient and low-cost devices. These could include optical sensors, radar-like light-based LiDAR systems, gas, pollution and environmental monitoring devices and biomedical sensors, such as disposable lab-on-a-chip solutions for diagnosing disease.
Now that we have shown that Hex-SiGe has the proper physical properties for efficient light emission, the demonstration of a scalable pathway to integrating Hex-SiGe into conventional silicon electronics or silicon photonics circuitry is the next big challenge, the project coordinator says. The fundamental difference between now and the situation before the SILAS project started is that we know any successful integration method will pay off. It will result in a light emitter in silicon technology that can be used for intra-chip or chip-to-chip communication.
He says that once a successful integration method has been developed, the project consortium can foresee sizeable cost reductions in manufacturing in high volumes in existing silicon foundries.
Industrial partner IBM is addressing the integration challenge, working on methods to introduce Hex-SiGe into silicon chip fabrication processes. SILAS researchers are also planning to develop a prototype Hex-SiGe nano-laser before the end of the project, alongside making progress on light-emitting nano-LEDs and other experimental optoelectronic devices. Their results to date are reported in a scientific paper on the breakthrough technology which is available on the open access ArXiv website.
The SILAS project has removed the existing fundamental barriers for light emission out of silicon germanium. If industry and the scientific community jump on it, silicon-based photonics circuits with integrated Hex-SiGe lasers and optical amplifiers will be demonstrated and commercialised in the next 5 to 10 years, Haverkort predicts.
About the Author
