Thermal Crosstalk Modelling and Compensation Methods for Programmable Photonic Integrated Circuits

TL;DR

This study tackles deterministic thermal crosstalk in programmable photonic integrated circuits by developing and experimentally evaluating three off-line crosstalk-predictive models that link interferer phase shifts to microring resonance shifts. The TPM provides a physics-grounded baseline, the ThDM adds distance-based diffusion effects for better generalization, and the LR offers data-driven per-PUC weighting that achieves the highest accuracy while remaining layout-agnostic. Experimental results show sub-picometer RMSEs (typical <0.5 pm) and demonstrate crosstalk compensation on MRRs, including cross-MRR generalization with RMSEs below ~1.2 pm. The work highlights a practical pathway for off-line PIC modelling and compensation, enabling more reliable, scalable, and compact photonic neural network hardware.

Abstract

Photonic integrated circuits play an important role in the field of optical computing, promising faster and more energy-efficient operations compared to their digital counterparts. This advantage stems from the inherent suitability of optical signals to carry out matrix multiplication. However, even deterministic phenomena such as thermal crosstalk make precise programming of photonic chips a challenging task. Here, we train and experimentally evaluate three models incorporating varying degrees of physics intuition to predict the effect of thermal crosstalk in different locations of an integrated programmable photonic mesh. We quantify the effect of thermal crosstalk by the resonance wavelength shift in the power spectrum of a microring resonator implemented in the chip, achieving modelling errors <0.5 pm. We experimentally validate the models through compensation of the crosstalk-induced wavelength shift. Finally, we evaluate the generalization capabilities of one of the models by employing it to predict and compensate for the effect of thermal crosstalk for parts of the chip it was not trained on, revealing root-mean-square-errors of <2.0 pm.

Figures (10)

• Figure 1: a) Schematic of the 72-PUC waveguide mesh with the MRR filter. Blue-colored PUCs denote PUCs in bar-state. Yellow-colored PUCs denote PUCs used as tunable couplers. The green arrows indicate the simulation of the coupling waveguides. Red arrows indicate input (in) and output (out) ports. b) Spectral response of an MRR filter. Resonance wavelengths are marked with red stars. FSR denotes the period of the MRR's spectral response.
• Figure 2: Hexagonal mesh with 3 MRR filters used for thermal crosstalk modelling. MRR 1 (green), MRR 2 (orange), and MRR 3 (blue) are indicated using different colors. Coupling ratios of the PUCs coupling to the input and output ports (34, 36, and 53 for MRRs 1-3, respectively) were set to 0.9. Maximum ERs were achieved by setting coupling ratios of the PUCs 35, 37 and 54 (in MRR 1, MRR 2, and MRR 3, respectively) to 0.77. Achieved ER values were 25, 20, and 30 dB for MRRs 1-3, respectively. The rest of the PUCs forming the MRRs were set to the bar state. Programming MRR 3 required fixing PUCs accross the blue dashed line to guide the light to the input and the output port. Therefore, PUCs 31 and 41 were set to the bar-state, and PUCs 28, 29, 30, 38, 39, 40 were set to the cross-state.
• Figure 3: Histogram of the distribution of the total driven phase on the PUCs
• Figure 4: Experimental setup for thermal crosstalk observation. ASE: amplified spontaneous emission, TF: tunable filter, EDFA: Erbium-doped fiber amplifier, PC: polarization controller, OSA: optical spectrum analizer.
• Figure 5: Total Phase Model fitted for for MRR 1, MRR 2, and MRR 3