Zero-Change Zero-Cost Electronics in Silicon Photonics for Real-Time Control of Programmable Interferometric Meshes

Introduction

Photonic integrated circuits (PICs) have emerged as a promising technology for various applications, including optical computing, telecommunications, and sensing. However, the development of large-scale programmable photonic circuits, such as meshes of interferometers, has been hindered by the required number of electrical connections to external control units. To ensure stability of operations and programmability, these circuits require a large number of light sensors and phase shifters, which need to be controlled in a feedback configuration. Traditionally, this has been achieved using external control electronics, but the practical number of electrical connections is limited to a few hundred, making it challenging to control complex photonic circuits.

In this tutorial, we will explore a novel approach proposed by researchers at Politecnico di Milano to integrate electronic circuits directly into a standard silicon photonic (SiP) technology platform, without modifying the existing technological stack. This zero-change, zero-cost approach enables the integration of multiplexers and memory devices, significantly reducing the number of external connections while still ensuring closed-loop control of the photonic functionalities.

Zero-Change Zero-Cost Electronics in SiP

The key to this approach is the leveraging of the n- and p-type dopings available in the active photonic process to build transistor gates on the sides of the waveguide. The thickness of the gate oxide is defined by the minimum etching distance between two contiguous silicon strips, which is approximately 200 nm in this case. Source and drain contacts are obtained by doping the waveguide layer along its length, and connection to metal layers is achieved using the metal vias traditionally employed to address photodiodes and heaters. Figure 1 shows a 3D view of the structure of an nMOS transistor next to a Mach-Zehnder interferometer (MZI) on the photonic process stack.

Fig. 1. 3D view of the structure of an nMOS transistor next to a MachZehnder interferometer (not to scale) on the photonic process stack.

Figure 2 displays the characteristic curves of the designed nMOSFET and pMOSFET, as well as a microscope photograph of the obtained device.

Fig. 2. Measured characteristic curves of the realized nMOS and pMOS transistors. The inlet shows a micrograph of the device.

Analog Multiplexers for Photonic Circuits

To reduce the number of electrical connections from the photonic chip to the external control electronics, the researchers combined nMOSFETs and pMOSFETs to design analog multiplexers (MUX). A MUX, composed of 596 nMOSFETs and 84 pMOSFETs, has been optimized to connect the 16 photodiodes of a mesh prototype to a single output pad, enabling a sequential readout of their signals using only one transimpedance amplifier (TIA) (Figure 3a).

Fig. 3: (a) Schematic of a multiplexer linked to multiple photodiodes. (b) Graph showing sensor current variation with changes in heater thermal power. (c) Microscope image of the photonic chip.

The update time of the MUX has been measured to be around 2 μs. Figure 3b shows the current in 4 PDs of the mesh, arranged as in Figure 3a, and read with the MUX as a function of the heater thermal power when injecting light to one input. PD1 and PD2 correctly provide complementary information about the MZI transfer function, while PD3 and PD4 only generate their dark current.

A similar DEMUX can also be used to drive 16 heaters available as phase shifters in photonic meshes, as shown in Figure 4a.

Fig. 4: (a) Diagram of a MZI mesh featuring multiplexed heaters. (b) Heater current response to driving voltage and multiplexer commands. (c) Bit error rate at 10 Gbps plotted against switching frequency, with an eye diagram for two cascaded MZIs depicted in the inset.

Multiplexed Control of Photonic Meshes

The presence of MUXes raises the question of whether it is still possible to configure and stabilize programmable photonic circuits at runtime. The researchers demonstrate that enforcement action can be effectively performed with electrical signals multiplexed over time directly on the photonic chip. The electronic controller has been designed to allow time-multiplexed readout of integrated photodetectors and activation of thermal phase shifters by automatically cycling the MUX/DEMUX bit commands to address each device.

To obtain sequential control of integrated volatile actuators, such as heaters, an on-chip memory device has been integrated on the photonic chip together with the actuator driver and the DEMUX, as shown in Figure 4a. The power transistor, made of 700 nMOSFETs in parallel, is connected in source follower configuration to supply the heater with a linear relation between gate voltage and current. The memory functionality has been obtained by integrating an on-chip Sample & Hold with a memory capacitance of a few pF, providing a sample time of about 20 ns and a hold time of 0.1 seconds (Figure 4b).

The precision of time-multiplexed control, achieved on a time scale of 2 ms for each MZI, is demonstrated by routing a 10 Gbit/s modulated signal through a cascade of two sequentially-controlled MZIs. Figure 4c reports the bit-error-rate and eye diagram at the output when changing the Sample & Hold refresh rate, showing that no penalties are observed when the switching frequency is increased above a few tens of Hz. This approach can thus be applied directly to large-scale photonic chips.

Conclusion

The integration of electronic circuits into a standard photonic platform without modifications to the technological stack, achieved through this zero-change, zero-cost approach, represents a significant advancement in the development of large-scale programmable photonic circuits. By drastically reducing the number of interconnections required for feedback control, this technique paves the way for the realization of complex photonic meshes and routers, enabling a wide range of applications in optical computing and telecommunications.

The successful demonstration of electronic integrated multiplexers drastically reducing the number of interconnections required to control large-scale programmable photonic circuits is a major milestone. This approach overcomes a significant bottleneck in the development of complex photonic integrated circuits, opening up new possibilities for the design and implementation of advanced optical systems.

Furthermore, the zero-change, zero-cost nature of this integration technique means that it can be readily adopted by existing silicon photonics foundries without the need for costly modifications to their technological processes. This seamless integration of electronics and photonics on the same chip represents a significant advantage, facilitating the widespread adoption of this technology and accelerating the development of innovative photonic solutions.

Overall, this work represents a pioneering effort in the field of integrated photonics, offering a practical and cost-effective solution for the realization of complex programmable photonic circuits. As the demand for high-performance, energy-efficient optical systems continues to grow, this zero-change, zero-cost approach to electronics-photonics integration holds great promise for enabling a new generation of advanced photonic technologies.

Reference

[2] F. Zanetto, F. Toso, M. Crico, F. Morichetti, A. Melloni, G. Ferrari, and M. Sampietro, "Zero-Change Zero-Cost Electronics in Silicon Photonics for Real-Time Control of Programmable Interferometric Meshes," in Department of Electronics Information and Bioengineering, Politecnico di Milano, Politecnico di Milano, Department of Physics, 20133 Milano, Italy, 2024.