Photonics Research, Volume. 11, Issue 1, 65(2023)
Experimental demonstration of coherent photonic neural computing based on a Fabry–Perot laser with a saturable absorber
Fig. 1. Schematic diagram of the coherent photonic spiking neural network.
Fig. 2. (a) Experimental setup for the coherent photonic spiking neural network; (b) integrated FP-SA chip and (c) two input signals
Fig. 3. (a) PI curve and (b) optical spectrum of the FP-SA measured at 25°C. The total length of the FP-SA chip is 1500 μm, the width of the laser chip is 300 μm, the ridge waveguide width is 2.5 μm, and the length of the SA region is 75 μm. The SA section side facet and gain section side facet are coated with reflections of 95% and 30%, respectively.
Fig. 4. Modulation characteristics of the DQPSK single-drive MZM. The phase shifter is 0. The bias voltage of MZM1 (MZM2) is
Fig. 5. Experimental demonstration of temporal integration property and spike threshold property. (a1) and (a2) indicate temporal integration property. (a1) represents the input signals for the weighted summation and (a2) denotes the response of the FP-SA neuron. (b1) and (b2) indicate spike threshold property. (b1) represents the input signals for the weighted summation and (b2) denotes the response of the FP-SA neuron. Temporal maps show the laser neuron’s responses to 100 successive external stimuli, where (c1) corresponds to the stimulus in (a1) and (c2) to the stimulus in (b1). The bias voltage of the FP-SA is
Fig. 6. Experimental demonstration of the refractory period property. (a1) and (a2) indicate that FP-SA neurons are in the refractory period regime. (a1) represents the input signals for the weighted summation and (a2) denotes the response of the FP-SA neuron. (b1) and (b2) indicate that the FP-SA neuron has left the refractory period regime. (b1) represents the input signals for the weighted summation and (b2) denotes the response of the FP-SA neuron. Temporal maps show the laser neuron’s responses to 100 successive external stimuli, where (c1) corresponds to the stimulus in (a1) and (c2) to the stimulus in (b1). The bias voltage of the FP-SA is
Fig. 7. Experimental demonstration of the effect of frequency detuning on spike threshold property. (a) represents the input stimulus signal. (b1)–(b3) denote the response of the FP-SA neuron, corresponding to the three optical spectra for the FP-SA that operates as a photonic spiking neuron, as shown in (c1)–(c3). The bias voltage of the FP-SA is
Fig. 8. Spike response of the FP-SA neuron when injecting the triangular stimulus signal. (a) represents the input signal where the light intensity varies as a triangular wave. (b1) and (b2) denote the response of the FP-SA neuron. (c) represents the ISI interval distribution corresponding to the first type of spike and the second type of spike. The bias voltage of the FP-SA is
|
Get Citation
Copy Citation Text
Dianzhuang Zheng, Shuiying Xiang, Xingxing Guo, Yahui Zhang, Biling Gu, Hongji Wang, Zhenzhen Xu, Xiaojun Zhu, Yuechun Shi, Yue Hao, "Experimental demonstration of coherent photonic neural computing based on a Fabry–Perot laser with a saturable absorber," Photonics Res. 11, 65 (2023)
Category: Lasers and Laser Optics
Received: Aug. 1, 2022
Accepted: Nov. 8, 2022
Published Online: Dec. 21, 2022
The Author Email: Shuiying Xiang (syxiang@xidian.edu.cn), Yuechun Shi (yuechun-shi@ylab.ac.cn)