Building phononic integrated circuits with GaN
The ability to guide high-frequency sound around a semiconducting chip positions GaN as a promising platform for producing compact, high-performance acoustic wave devices.
BY Mahmut Bicer, Stefano Valle, Jacob Brown, Martin Kuball and Krishna C. Balram FROM THE UNIVERSITY OF BRISTOL.
A long-standing theme of engineering research is exploring and exploiting the similarities between different wave phenomena. Their similarities, such as reflection, superposition and the creation of standing waves, has been appreciated for decades, thanks in part to excellent demonstrations by the late John Shive, who worked at Bell Labs. Back in 1959 Shive gave a lucid, insightful lecture on this topic that can still be enjoyed via Youtube. In that talk, standing as one of the earliest examples of recognizing the unifying nature of the underlying physics and illustrating how to translate ideas from one field to another, Shive demonstrated his wave generator, a great contribution to teaching this topic.
One modern iteration of exploiting wave phenomena is an extension of silicon integrated photonics, with chips designed to control the propagation of sound waves with frequencies in the gigahertz domain. Such efforts build on progress in silicon photonics over the last decade that spawned a revolution in optical telecommunication, with CMOS foundries now used to build photonic chips. The underlying physics behind this development is the high refractive index contrast provided by a silicon-on-insulator platform – this enables tight confinement of propagating light in wavelength-scale waveguides. For instance, the standard waveguides used to route optical telecommunication signals with wavelengths of around 1550 nm have a cross-section of around just 220 nm by 550 nm. That’s substantial miniaturisation compared with an optical fibre: it has a typical core size of 8.2 mm and a cladding diameter of 125 mm, due to its far weaker refractive index contrast.
There are many benefits associated with the extreme confinement that enables compact devices. One great attribute is that a single chip can accommodate various passive and active functions, such as splitting signals, combining them, mode transformation modulation and detection. In addition, there is a low propagation loss that accompanies the strong confinement, resulting in high electric field strengths and associated non-linearities. These properties may be harnessed to implement a variety of functions for quantum information processing, such as single-photon generation using spontaneous four-wave mixing.
As the wavelength of sound waves in the gigahertz range is similar to that of the light used for telecommunication, it is natural to ask how far these ideas from integrated photonics can be extended to gigahertz frequency acoustic waves. One may also wonder what new device paradigms might be enabled.
It should be noted that the idea of guiding sound in waveguides is not new – this had been discussed from the very early days of waveguide research. However, this interest has evolved, and is now motivated by considerations related to the acoustic waves that underlie all the filtering in modern smartphones. Today’s state-of-the-art smartphones are packed with between 30 and 50 acoustic wave filters, with their number increasing with each successive generation. It is an ever-increasing challenge to accommodate this increasing number of discrete filters into a given area, along with the associated switches, amplifiers and other signal processing circuitry. Current devices are a packaging tour-de-force, but we cannot expect these methods to work well into the future.
The tremendous successes that have come from microelectronics, and more recently silicon photonics, indicate that monolithic integration could be the best solution to this growing problem. Success hinges on figuring out a way to trim the size of traditional acoustic wave devices, and how to implement tight integration between active and passive components on the same die. Much effort has already been directed at accomplishing this, with approaches relying on methods that employ a CMOS process to integrate existing acoustic-wave devices, such as FBAR and SAW filters, which rely on quasi plane-wave resonators. But at the University of Bristol we are taking a different path, investigating whether acoustic-wave devices can be redesigned around phononic integrated circuits (PnIC), featuring strong geometrical confinement of sound. We are keen to explore the benefits and challenges of this approach.