Abstract
Silicon modulators are fundamental components in silicon photonics, enabling the integration of high-speed electro-optic (EO) functions on compact chips for data transmission and photonic computing. Despite significant progress, a practical bandwidth ceiling of approximately 67 GHz has persisted for pure silicon modulators, hindering the path toward Terabit-per-second (Tbps) data throughput. This paper presents a design strategy that utilizes the slow light effect within a coupled-resonator optical waveguide (CROW) architecture to shatter this limit. By balancing group index, photon lifetime, and electrical bandwidth, we demonstrate a silicon modulator with a record-breaking EO bandwidth of 110 GHz and a compact footprint of 124 μm. The device achieves 112 Gbps on-off keying (OOK) modulation without digital signal processing (DSP), proving the feasibility of silicon-based ultra-high-bandwidth applications.
Introduction
As global data traffic continues to surge, the demand for high-speed, energy-efficient, and compact optical communication components has never been higher. Silicon photonics has emerged as a leading platform due to its compatibility with complementary metal-oxide-semiconductor (CMOS) fabrication processes. However, the performance of "pure" silicon modulators—those relying on the plasma dispersion effect—has historically been limited. While heterogeneous integration with materials like lithium niobate or organic polymers can extend performance, these approaches introduce significant manufacturing complexity and cost. The challenge remains to push the limits of pure silicon to meet the requirements of next-generation Tbps-level networks.
The 67 GHz Bandwidth Ceiling
The primary mechanism for modulation in silicon is the plasma dispersion effect, where the concentration of free carriers in a PN junction is varied to change the refractive index. While effective, this process faces a practical "bandwidth ceiling" at roughly 67 GHz. This limitation arises from the inherent trade-offs between modulation efficiency, optical loss, and the electrical RC (resistance-capacitance) constants of the device. To achieve higher speeds, researchers typically have to increase doping levels, which unfortunately increases parasitic capacitance and optical absorption, ultimately throttling the device's speed and efficiency.
Design Strategy: Leveraging the Slow Light Effect
To overcome these limitations, this research employs a Coupled-Resonator Optical Waveguide (CROW) architecture. This structure consists of a series of cascaded Bragg gratings, often described as a "fish-bone" waveguide, integrated with periodic phase shifters.
The Balancing Act
The core innovation lies in the strategic use of the slow light effect. By designing the gratings to create a topological "mid-gap" mode, the group velocity of light is reduced. This increases the interaction time between the optical carrier and the electrical signal, which significantly boosts modulation efficiency.
Traditionally, slow light is used to minimize device size. However, slowing light also increases the "photon lifetime" within the structure, which can act as a bottleneck for high-speed operation. The researchers developed a theoretical framework to identify a "sweet spot"—targeting a modest group index ($n_g \approx 6.1$). This allows for:
- Reduced Footprint: Achieving high efficiency in a length of only 124 μm.
- Lower Doping Levels: Because the slow light effect provides an efficiency boost, the researchers could use lower doping concentrations. This keeps the electrical bandwidth high (approximately 200 GHz) by reducing parasitic capacitance.
- Optimized Bandwidth: Ensuring the photon lifetime does not limit the EO bandwidth below the desired 110 GHz threshold.
Experimental Results
The device was fabricated using a standard 90 nm Silicon-on-Insulator (SOI) process at a commercial foundry, ensuring that the design is compatible with mass-production environments.
Record-Breaking Performance
Testing using a 110 GHz vector network analyzer confirmed a 3-dB EO bandwidth of 110 GHz, the highest ever reported for a pure silicon modulator. This represents a significant leap over the previous 67 GHz benchmark. Furthermore, the device maintained stable performance across an 8 nm optical window, making it highly suitable for Wavelength Division Multiplexing (WDM) systems.
High-Speed Data Transmission
In real-world testing, the modulator demonstrated 112 Gbps OOK transmission. Notably, this was achieved with a Bit Error Rate (BER) below the soft-decision forward-error-correction (SD-FEC) threshold and, crucially, without the need for power-hungry Digital Signal Processing (DSP) for equalization. This "DSP-free" operation is a major advantage for reducing latency and power consumption in short-reach data center interconnects.
Conclusion and Future Outlook
By successfully integrating the slow light effect into a CROW architecture, this work proves that pure silicon modulators can exceed the 110 GHz bandwidth threshold while maintaining an ultra-compact footprint. While the current device shows an insertion loss of 6.8 dB, future optimizations in grating couplers and phase shifter geometry are expected to improve this metric.
Furthermore, while this study focused on simple OOK modulation, the architecture is inherently scalable. Future iterations could utilize multi-level modulation formats such as PAM-4 or 16-QAM, potentially pushing aggregate data rates beyond 1 Tbps per carrier. This breakthrough sheds light on the immense potential of silicon photonics for ultra-high-bandwidth applications in data communication, optical interconnection, and photonic machine learning.
Summary updated