top of page
Writer's pictureLatitude Design Systems

Metamaterials for analog optical computing

Introduction

The idea of an analog computer – a device that uses continuous variables rather than zeros and ones – may evoke obsolete machinery from the past. However, emerging technologies, including artificial intelligence, could greatly benefit from this computing approach. A promising direction involves analog computers that process information with light rather than electrical currents.

Metamaterials, a class of engineered materials with exotic optical properties, offer a powerful platform for building these analog optical computers. Recent work has demonstrated metamaterial-based schemes that can perform mathematical operations on light, potentially enabling computing tasks to be carried out much faster and more efficiently than with conventional digital processors.

Harnessing the Unique Properties of Metamaterials

At the heart of this approach are the unique properties of metamaterials. These synthetic materials are composed of many subwavelength-scale building blocks that can be tailored to exhibit optical behaviors not found in natural materials. For example, metamaterials can be engineered to have a near-zero or even negative refractive index, enabling exotic applications like sub-wavelength imaging and invisibility cloaking.


Sketch of the equation-solving scheme
Figure 1. Sketch of the equation-solving scheme used in the 2019 demonstration.

The design flexibility of metamaterials has inspired researchers to explore strategies for turning them into computing machines. In 2014, a team led by Nader Engheta of the University of Pennsylvania proposed that metamaterials could be used to realize a suite of mathematical operations, including differentiation, integration, and convolution. The key idea is to take an electromagnetic wave as an input and manipulate it through interaction with the metamaterial structure, such that the output wave corresponds to the desired mathematical transformation of the input.

Experimental Demonstrations in Microwaves and Optics

In 2019, Engheta's group demonstrated this concept experimentally, using a microwave-frequency metamaterial device. Their scheme involved a block of metamaterial with several input and output ports connected by waveguides in a feedback loop. For a given input, the device's output corresponded to the solution of a Fredholm integral equation – a mathematical problem with applications in diverse fields like fluid mechanics, antenna design, and quantum mechanics.

The metamaterial structure that enabled this computation had a nontrivial "Swiss-cheese" pattern, with an inhomogeneous distribution of small islands with different dielectric properties (air holes, polystyrene, and microwave-absorbing materials). This structure was designed using an "inverse design" approach – an iterative optimization process to find the metamaterial parameters that would realize the desired mathematical operation.

While the microwave demonstration was an important proof of concept, the bulky, impractical nature of microwave setups motivated researchers to extend these ideas to optical frequencies. Several groups have since reported a variety of optical computing schemes using thin metamaterial sheets, known as metasurfaces, to manipulate the propagation of light in free space.

Integrated Metamaterial Platforms for Optical Computing

However, metasurface approaches often require sophisticated, customized fabrication processes, limiting their potential for mass production. To address this challenge, Engheta and his colleagues have now developed an on-chip metamaterial platform for analog optical computing.


Sketch of the metamaterial-based silicon-photonics chip
Figure 2. Sketch of the metamaterial-based silicon-photonics chip.

Unlike the free-space light propagation in metasurface schemes, Engheta's team channels light through structured waveguides on a silicon chip. Using inverse design, they have fabricated a micron-scale chip with a structure reminiscent of their 2019 microwave design: a set of waveguides feeding light into and out of a flat cavity containing a Swiss-cheese-like metamaterial.

This integrated, chip-based approach offers several advantages. The device can be easily ordered from commercial foundries, simplifying the manufacturing process. And compared to the microwave version, the optical chip performs a simpler mathematical operation – matrix-vector multiplication, which is useful for AI tools like neural networks.

To solve more complex equations, the team plans to incorporate feedback waveguides linking the outputs back to the inputs, as was done in the microwave experiments. This reconfigurability feature is another key aspect of Engheta's work, enabling the analog computer to be reprogrammed to perform different mathematical tasks.

Reconfigurable Analog Computers

In a recent proof-of-principle demonstration, Engheta's group built a reconfigurable analog computer operating at radio frequencies (45 MHz). This device consisted of a 5x5 module of elements like amplifiers and phase shifters, whose parameters could be adjusted to change the mathematical operations performed by the system.

As examples, the researchers used this reconfigurable device to solve the roots of a system of polynomials and to perform the inverse design of a metastructure – both of which are nonstationary problems requiring a sequence of different mathematical steps. Engheta envisions that this reconfigurability could ultimately be integrated into the silicon-photonics chips, potentially by depositing a patterned layer of a "phase-change" material on top of the waveguides.

Conclusion

The programmable, metamaterial-based silicon-photonics chip could be a game-changer for analog optical computing. By processing information at the speed of light with a fraction of the energy required by conventional digital processors, these devices could excel at tasks like neural network inference. And since photons do not interact with each other, the chips could perform parallel operations simply by using different wavelengths of light.

Moreover, these analog optical computers offer potential privacy benefits, as they do not require intermediate steps that store information in a potentially hackable memory. Instead, the computation is performed "in-flight" as the light propagates through the waveguide labyrinth.

As the field of metamaterials continues to advance, the development of these integrated, reconfigurable analog optical computing platforms promises to open new frontiers in information processing. By harnessing the unique properties of engineered materials, researchers are paving the way for a new generation of high-speed, energy-efficient computing devices that could revolutionize a wide range of applications, from artificial intelligence to scientific simulations.


Reference

[1] M. Rini, "Metamaterials for Analog Optical Computing," Physics, vol. 17, no. 52, Mar. 29, 2024. Available: https://physics.aps.org/articles/v17/52.

Comments


bottom of page