Semiconductor laser reservoir computing device based on silicon-based microcavity optical feedback
By introducing silicon-based microresonant cavity optical feedback into a semiconductor laser RC system, the virtual node states are enriched, solving the problems of computational error and insufficient memory capacity in traditional systems for complex tasks, and achieving higher prediction accuracy and stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SICHUAN NORMAL UNIV
- Filing Date
- 2026-02-05
- Publication Date
- 2026-06-09
AI Technical Summary
Traditional optical feedback semiconductor laser RC systems suffer from large computational errors, low memory capacity, and narrow feedback force operating range when handling complex tasks, making it difficult to effectively process complex chaotic time series.
By introducing silicon-based micro-resonant cavity optical feedback, the virtual node state is enriched through the nonlinear interaction between the silicon-based micro-ring resonant cavity and the semiconductor laser, forming a nonlinear optical feedback loop, and the training weights are optimized by combining the ridge regression algorithm.
It improves computational accuracy and memory capacity, enhances system stability and robustness, and enables better handling of complex chaotic time series.
Smart Images
Figure CN122174904A_ABST
Abstract
Description
Technical Field
[0001] This invention provides a semiconductor laser reservoir computing device based on silicon-based micro-resonant cavity optical feedback, belonging to the field of photonic neuromorphic computing technology, and is suitable for high-speed optical computing scenarios such as complex chaotic time series prediction and nonlinear signal processing. Background Technology
[0002] Existing photonic delay RC systems for semiconductor lasers typically employ a fiber optic delay feedback structure. This involves re-injecting the output of the EESL / VCSEL into the laser through a fiber optic loop of a certain length, forming an optical feedback loop. The output of the tunable laser (TSL) is modulated by a modulator and then injected into the response laser (SL, i.e., EESL / VCSEL). The output of the SL is converted into an electrical signal by a broadband photodetector, serving as the output of a virtual node, which is further used to calculate the training weights for the RC system.
[0003] Traditional optical feedback semiconductor laser RC systems suffer from drawbacks such as large calculation errors, low memory capacity, and narrow feedback force working area when dealing with complex tasks (such as NARMA10). Summary of the Invention
[0004] To address the aforementioned technical problems, this invention provides a semiconductor laser reservoir computing device based on silicon-based micro-resonator optical feedback. It integrates a silicon-based micro-resonator (such as a micro-ring resonator, MRR) into the EESL-RC / VCSEL-RC system to further enrich the virtual node states of the system, thereby achieving the goal of reducing computational errors and increasing memory capacity.
[0005] The specific technical solution is as follows:
[0006] A semiconductor laser reservoir computing device based on silicon-based microcavity optical feedback includes a driving laser, a phase modulator, a responding semiconductor laser, a silicon-based microresonator, a feedback cavity coupler, a circulator, a first variable optical attenuator, a second variable optical attenuator, a polarization controller, and a broadband photodetector.
[0007] The laser output from the driving laser is modulated by a phase modulator and then injected into the response semiconductor laser through a feedback cavity coupler and a circulator. The output of the response semiconductor laser is connected to the optical circulator in sequence through a second variable optical attenuator, a polarization controller, a silicon-based microresonator, and a first variable optical attenuator, forming a silicon-based microresonator optical feedback loop. The output of the feedback cavity coupler is connected to a broadband photodetector for acquiring the output signal of the reservoir. The phase modulator is loaded with a modulation signal containing the input signal and a random mask to achieve input layer signal encoding.
[0008] Furthermore, the nonlinear optical devices within the feedback cavity include both a responsive semiconductor laser and a silicon-based microresonator.
[0009] Preferably, the modulated signal S(t) satisfies S(t) = Mask(t)u(n), where γ is the scale factor, Mask(t) is the random mask, u(n) is the input signal with a sampling period of T and a modulation period of T=Mθ, where M is the total number of virtual nodes and θ is the time interval between adjacent virtual nodes.
[0010] The first variable optical attenuator is used to adjust the feedback intensity k of the response semiconductor laser. SL The second variable optical attenuator is used to adjust the feedback constant η of the silicon-based microresonator. MRR .
[0011] The types of responsive semiconductor lasers include edge-emitting semiconductor lasers or vertical-cavity surface-emitting lasers.
[0012] Furthermore, after obtaining the output light signal through a broadband photodetector, the training weights are obtained through a ridge regression algorithm, thereby completing the time series prediction or classification task.
[0013] This invention improves overall computational performance by integrating the nonlinear interactions of a driving semiconductor laser (SL2) and a silicon-based microring resonator (MRR). Specifically, in terms of prediction accuracy, compared to optical reservoirs using traditional optical feedback, this invention achieves superior prediction results on typical time series prediction tasks such as Santa Fe and NARMA10. Simultaneously, the system's memory capacity is effectively enhanced, enabling it to more efficiently handle complex chaotic time series with high memory requirements. Furthermore, the introduction of the microring resonator reduces the system's sensitivity to feedback parameters, thereby enhancing its stability and robustness, and further ensuring reliable operation in complex scenarios. Attached Figure Description
[0014] Figure 1 An RC device diagram is provided for this invention;
[0015] Wherein, PD: broadband photodetector; PM: phase modulator; Coupler: coupler; CIR: optical circulator; SL1: driving semiconductor laser; SL2: responding semiconductor laser (EESL or VCSEL); MRR: silicon-based microring resonator; VOA1: first attenuator; VOA2: second attenuator; Coupler: coupler; PC: polarization controller.
[0016] Figure 2This is one of the comparisons of the prediction performance of the MRR feedback (MRR-OF) technology proposed in this invention with that of traditional optical feedback (COF).
[0017] Figure 3 shows the second comparison of the prediction performance of the MRR feedback (MRR-OF) technology proposed in this invention with that of traditional optical feedback (COF).
[0018] Figure 4 This invention compares the memory capabilities of MRR-OF and COF. Detailed Implementation
[0019] The specific technical solutions of the present invention will be described with reference to the embodiments.
[0020] To address the shortcomings of existing technologies, this embodiment proposes a nonlinear feedback SL-RC system based on silicon-based microresonators (such as MRR), aiming to improve the prediction accuracy and memory capacity of chaotic time series, especially for the prediction of complex chaotic time series (such as NARMA10).
[0021] like Figure 1 As shown, the SL-RC system proposed in this invention uses a silicon-based micro-resonant cavity (such as MRR) as an optical feedback element. By adjusting the feedback constant of MRR and the feedback intensity of SL2, the feedback state can be enriched.
[0022] The core components of the semiconductor laser reservoir computing device based on silicon-based microcavity optical feedback include a driver laser (SL1), a phase modulator (PM), a response semiconductor laser (SL2, EESL / VCSEL), a silicon-based microresonator (MRR), a 2×1 feedback cavity coupler, a circulator (CIR), a first variable optical attenuator (VOA1), a second variable optical attenuator (VOA2), a polarization controller (PC), and a broadband photodetector (PD).
[0023] The laser output from the driving laser (SL1) is modulated by a phase modulator (PM) and injected into the response semiconductor laser (SL2) through a feedback cavity coupler and a circulator (CIR). The output of the response semiconductor laser (SL2) is connected to the optical circulator (CIR) in sequence through a second variable optical attenuator (VOA2), a polarization controller (PC), a silicon microresonator (MRR), and a first variable optical attenuator (VOA1), forming a silicon microresonator optical feedback loop. The output of the cavity coupler is connected to a broadband photodetector (PD) for acquiring the output signal of the reservoir. The phase modulator (PM) loads a modulated signal containing the input signal and a random mask to achieve input layer signal encoding.
[0024] The specific connection relationships and working principles of each component are as follows:
[0025] 1. Signal Input and Modulation: The driving semiconductor laser (SL1) outputs continuous laser light, which is then modulated by a phase modulator (PM) with a modulation signal S(t) = Mask(t)u(n) is used to perform optical modulation of the input signal u(n), where Mask(t) is a random mask and γ is a scale factor.
[0026] 2. Laser Injection and Feedback: The modulated laser is injected into the Response Semiconductor Laser (SL2, which can be implemented using an EESL or VCSEL) through a feedback cavity coupler and circulator (CIR); the output laser of SL2 is adjusted by a variable optical attenuator (VOA2) to regulate the feedback constant (η). MRR Afterwards, the polarization state is calibrated by a polarization controller (PC) and then injected into a silicon-based microring resonator (MRR) to form a nonlinear optical feedback. The output of the MRR is then adjusted by a variable optical attenuator (VOA1) to regulate the SL2 feedback intensity (k). SL Finally, the feedback is sent to SL2 through a circulator (CIR) to form a nonlinear closed-loop feedback system.
[0027] 3. Signal detection and training weight calculation: The optical signal output by the feedback cavity coupler is collected by a broadband photodetector (PD), converted into an electrical signal, and used as the output of the virtual node in the reservoir. The weights of the output layer are then trained by the ridge regression algorithm to complete the time series prediction or classification task.
[0028] The nonlinear effects of the silicon-based microring resonator (MRR) include two-photon absorption (TPA), free carrier absorption (FCA), free carrier dispersion (FCD), Kerr effect, etc. The system improves the system's memory capacity and nonlinear mapping capability by driving the semiconductor laser (SL2) and the MRR in a synergistic nonlinear effect.
[0029] In this embodiment, after careful optimization, optimization parameters were selected for the Santa Fe and NARMA10 prediction tasks, respectively, as shown in Figure 2 and... Figure 3 This paper presents a comparison of the prediction performance of the MRR optical feedback (MRR-OF) proposed in this invention and the conventional optical feedback (COF), showing the normalized mean square error (NMSE) as a function of feedback strength k. SL The changes. According to Figure 2 and Figure 3It can be seen that the minimum NMSE for the Santa Fe task is 0.017 (COF), while the proposed MRR-OF further reduces it to 0.008. On the other hand, when performing the NARMA10 task, even with optimized feedback parameters, the minimum NMSE (0.505) for COF is still too high, resulting in a large test error between the predicted and target values. In contrast, the minimum NMSE of the proposed MRR-OF is reduced to 0.125, see [link to relevant documentation]. Figure 3 The red curve in the image.
[0030] In this embodiment, Figure 4 The memory capacity (MC) as a function of the semiconductor laser feedback intensity (k) was plotted. SL The figure shows the curves of change, comparing the memory capacity of the proposed MRR optical feedback (MRR-OF) with that of the conventional optical feedback (COF). It can be observed that, under optimized parameters, the maximum MC using COF is 17.8; while the maximum MC of the proposed MRR-OF RC system can be increased to 23.5.
[0031] The above is merely one embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope disclosed in the present invention, based on the technical solution and concept of the present invention, shall fall within the scope of protection of the present invention.
Claims
1. A semiconductor laser reservoir computing device based on silicon-based microcavity optical feedback, characterized in that: It includes a driving laser, a phase modulator, a responsive semiconductor laser, a silicon-based microresonator, a feedback intracavity coupler, a circulator, a first variable optical attenuator, a second variable optical attenuator, a polarization controller, and a broadband photodetector; The laser output from the driving laser is modulated by a phase modulator and then injected into the response semiconductor laser through a feedback intracavity coupler and a circulator. The output of the response semiconductor laser is connected to an optical circulator via a second variable optical attenuator, a polarization controller, a silicon-based microresonator, and a first variable optical attenuator, forming a silicon-based microresonator optical feedback loop; the output of the feedback cavity coupler is connected to a broadband photodetector for acquiring the output signal of the reservoir; the phase modulator is loaded with a modulation signal containing the input signal and a random mask to achieve input layer signal encoding.
2. The semiconductor laser reservoir computing device based on silicon-based microcavity optical feedback according to claim 1, characterized in that: The nonlinear optical devices within the feedback cavity include both a responsive semiconductor laser and a silicon-based microresonator.
3. The semiconductor laser reservoir computing device based on silicon-based microcavity optical feedback according to claim 1, characterized in that: The modulated signal S(t) satisfies S(t) = Mask(t)u(n), where γ is the scale factor, Mask(t) is the random mask, u(n) is the input signal with a sampling period of T and a modulation period of T=Mθ, where M is the total number of virtual nodes and θ is the time interval between adjacent virtual nodes.
4. The semiconductor laser reservoir computing device based on silicon-based microcavity optical feedback according to claim 1, characterized in that: The first variable optical attenuator is used to adjust the feedback intensity k of the response semiconductor laser. SL The second variable optical attenuator is used to adjust the feedback constant η of the silicon-based microresonator. MRR .
5. The semiconductor laser reservoir computing device based on silicon-based microcavity optical feedback according to claim 1, characterized in that: The types of responsive semiconductor lasers include edge-emitting semiconductor lasers or vertical-cavity surface-emitting lasers.
6. The semiconductor laser reservoir computing device based on silicon-based microcavity optical feedback according to claim 1, characterized in that: After obtaining the output light signal through a broadband photodetector, the ridge regression algorithm is used to obtain training weights, thereby completing the time series prediction or classification task.