Programmable Integrated Photonics: A New Era of Multifunctional Optics

Introduction

Integrated photonics is a technology focused on generating, processing, controlling, and detecting guided lightwave signals carrying information. It supports a wide range of applications like telecommunications, computing, sensing, robotics, and aerospace.

While integrated photonics shares some similarities with microelectronics like leveraging semiconductor fabrication processes, there are key differences. Unlike silicon's dominance in electronics, no single material platform can monolithically integrate all required photonic components. This necessitates hybrid or heterogeneous integration approaches combining multiple chips or materials.

Additionally, photonics operates in the analog domain with linear optical signals that cannot be easily stored, contrasting with microelectronics' digital and nonlinear switching operations. However, this complementarity is precisely what opens opportunities for integrated photonics to address the limitations facing microelectronics from stagnating scaling laws.

The Programmable Photonics Approach

One promising solution is to adopt programmability in integrated photonics, analogous to programmable electronic devices like microprocessors, FPGAs, and DSPs. This leverages a single reconfigurable hardware platform that can implement and emulate multiple circuits and functions through software programming, enabling significant cost savings and reduced development cycles.

The principles behind programmable integrated photonics revolve around the ability to independently control the amplitude and phase of interfering optical signals. This is achieved using basic building blocks like tunable couplers, Mach-Zehnder interferometers, and phase shifters, which can be programmed via external electronic signals.

Figure 1. Basic building blocks in programmable photonics

By combining and interconnecting these basic blocks, three generations of programmable photonic hardware have emerged, as illustrated in Figure 2:

Figure 2. The three families of Programmable Photonics Hardware

1. Reconfigurable ASPICs: Retaining core features of fixed designs but with tunable operation points.

2. Multipoint Interferometers: 2D fixed topologies built from tunable interferometers, emulating arbitrary linear transformations.

3. Photonic Waveguide Meshes: Balanced interferometers in 2D patterns like hexagonal meshes, capable of emulating any feed-forward, feed-backward, and reconfigurable ASPIC transformation.

The Technology Stack

Operating a programmable integrated photonic processor involves a layered "technology stack" coordinating different tasks, as depicted in Figure 3:

Figure 3. The technology stack of a general-purpose integrated photonics processor

The photonic layer consists of the programmable waveguide mesh chip and surrounding components like input/output ports. The electronic layer handles monitoring, control, and driving the photonic elements via electronic signals. Finally, the software layer provides the user programming environment and lower-level resource optimization and coordination.

Applications of Programmable Integrated Photonics

As a transversal enabling technology, programmable integrated photonics finds applications across numerous domains, as illustrated in Figure 4:

Figure 4. Application fields for programmable integrated photonics

1. Telecommunications and Data Centers: Flexible transceivers, routers, switches, and multiplexers/demultiplexers for high-capacity, low-power circuit switching.

2. 5G/6G Wireless Communications: Programmable interfaces between fiber and radio frequency segments for multi-band, multi-channel, and multi-input/multi-output capabilities.

3. High-Performance Computing: Hardware accelerators for parallel computing, multi-core interconnects, and disaggregated clusters.

4. Sensors: Simultaneous communication and sensing capabilities for IoT, smart cities, LIDAR, and autonomous driving.

5. Artificial Intelligence and Novel Computing: Fast matrix-vector multiplications for deep learning, integration with novel photonic computing paradigms.

6. ASPIC Manufacturing: Rapid prototyping and emulation of ASPIC designs.

7. Quantum Information Systems: Emulation of linear optical transformers for quantum logic gates, boson sampling, and numerical operations.

This list is expected to grow as programmable integrated photonics matures further.

Challenges and Outlook

While promising, programmable integrated photonics faces challenges regarding scalability, losses, and power consumption. Efforts are underway to increase the component count, reduce propagation and coupling losses (targeting <4 dB overall), and achieve low or negligible power consumption through non-volatile tuning mechanisms.

If these challenges are addressed, programmable integrated photonics could become as instrumental as key programmable electronic devices, enabling a new era of highly flexible, multifunctional, and cost-effective optical systems across diverse application domains.

Reference

[1] J. Capmany and D. Perez-Lopez, "Programmable Integrated Photonics: A New Paradigm for Low Cost Multifunctional Optics," *Photoniques*, no. 125, pp. 34-38, May. 2024. doi: 10.1051/photon/202412534.