Researchers use smart engineering to speed data transmission

AI data centers need to move lots of information quickly, and the demand for faster data transmission is only growing.

Over very short distances, data centers can use traditional copper wires to connect chips right next to one another. But over distances longer than 1 m, the connection requires fiber optic cables. This is because electrical signals transmitted at fast speeds over copper cables degrade quickly with distance and generate large amounts of heat. In contrast, light signals that travel over fiber optic cables do not degrade and travel quickly without producing heat.

One of the hurdles limiting the speed of data center transmission is the Mach-Zehnder modulator (MZM), a silicon microchip that converts electrical data into light signals. The technology has reached some of its physical limits, and increased transmission speeds will require a modulator that can convert signals more efficiently, ideally with a smaller physical and environmental footprint, reducing energy consumption and heat production.

To address this issue, a group of researchers from YOKOHAMA National University, the Institute of Science Tokyo, the National Institute of Advanced Industrial Science and Technology (AIST) and the Photonics Electronics Technology Research Association (PETRA) developed an engineering technique to generate slowly propagating light, called ‘slow light,’ to create a smaller and more efficient silicon modulator.

The team published their paper, “Reduced-DSP 200G/λ low-voltage slow-light silicon modulator enabled by built-in EO equalizer,” on October 22 in the journal Nature Communications.

“In this paper, we created a silicon photonic crystal waveguide optical modulator that integrates a built-in electro-optic equalizer to break the bandwidth limitation while preserving the benefits of the slow-light effect,” said Toshihiko Baba, professor in the Faculty of Engineering at YOKOHAMA National University and senior author of the research study.

The silicon photonic crystal waveguide consists of a nanostructure made up of tiny holes. This nanostructure generates slow light that enhances the interaction between the light and device and significantly improves modulation efficiency when it is applied to modulators. With conventional slow-light modulators, large electrical resistance limits the device’s high-frequency bandwidth more severely than other traditional modulators. The team overcame this limitation by introducing the electro-optic (EO) equalizer into the slow light modulator to boost the frequency response, increasing the frequency bandwidth beyond 80 GHz, which is comparable to traditional silicon modulators.

“Importantly, slow light can be fabricated using CMOS-compatible processes, simultaneously satisfying high speed, low power consumption and small footprint requirements, meeting the next-generation optical transmitter needs of AI data centers,” said Baba.

Complementary metal-oxide-semiconductor (CMOS) is the standard method used today to make computer chips, meaning the silicon slow-light modulator developed by the research team could be manufactured at scale without building expensive new facilities. In contrast, electro-optic modulators using next-generation materials, such as thin-film lithium niobate, often require specialized manufacturing steps for bonding materials other than silicon.

Ultimately, the team produced a modulator with a 320 µm length capable of 140 Gb/s data transmission speeds in the simple On-Off data format. The modulation only consumed 26 fJ (femtojoule, 1 x 10^-15 joule, or one-quadrillionth of a joule) of energy per bit of data. Using a multi-level modulation format called PAM-4 (pulse amplitude modulation four level), the data speed finally reached 200 Gb/s at maximum. In contrast, traditional MZM technologies that are typically between 2 and 4 mm in length are capable of the same modulation speed while expending several thousand fJ of energy per bit of data. This means that the slow light modulator achieved approximately one order smaller footprint and two orders higher modulation efficiency than conventional modulators.

While the team has achieved a smaller footprint, superior efficiency and sufficiently high data speeds with their new silicon modulator design, the component does suffer from a larger optical insertion loss, which decreases the optical signal when inserted into the transmission pathway. The researchers are confident that this loss can be minimized through design optimization, however.

“We believe that this architecture offers promise as a building block for future co-packaged optics in high-density data centers,” said Baba.

Keisuke Kawahara from the Department of Electrical and Electronic Engineering at the Institute of Science Tokyo in Tokyo, Japan and the Department of Electrical and Computer Engineering at YOKOHAMA National University in Yokohama, Japan; Tai Tsuchizawa, Noritsugu Yamamoto, Yuriko Maegami and Koji Yamada from the National Institute of Advanced Industrial Science and Technology (AIST) in Tsukuba, Japan; and Nobuhiko Nishiyama from Department of Electrical and Electronic Engineering at the Institute of Science Tokyo and the Photonics Electronics Technology Research Association (PETRA) in Tokyo, Japan also contributed to this research.

This work was supported by JSPS KAKENHI (Grant Numbers JP25K23449 and JP23KJ0988), by NEDO (Grant Number JPNP16007) and by Lumentum Japan Inc.

###

YOKOHAMA National University (YNU) is a leading research university dedicated to academic excellence and global collaboration. Its faculties and research institutes lead efforts in pioneering new academic fields, advancing research in artificial intelligence, robotics, quantum information, semiconductor innovation, energy, biotechnology, ecosystems, and smart city development. Through interdisciplinary research and international partnerships, YNU drives innovation and contributes to global societal advancement.