An optical neuron structure based on an on-chip microring resonator

Abstract

This invention discloses an optical neuron structure based on an on-chip microring resonator, comprising, from front to back, a weighting unit, a summing unit, and a nonlinear unit. The weighting unit employs a wavelength division multiplexing (WDM) series microring resonator structure to apply weights to multiple input optical signals. The summing unit includes a photodetector and a preamplifier, used to sum the multiple weighted optical signals and drive the microring resonator of the nonlinear unit. The nonlinear unit employs a microring resonator structure to perform nonlinear output processing on the summed optical signals, primarily utilizing the nonlinear transmission characteristics of the microring resonator. This invention features high integration, and the proposed optical neuron exhibits strong scalability. Furthermore, the on-chip integration process is compatible with CMOS, enabling large-scale fabrication and facilitating the miniaturization, high integration, and high scalability of the optical neuron.

Description

Technical Field

[0001] This invention relates to the fields of silicon-based optoelectronic integrated devices and neuromorphic photonics, and particularly to optical neurons constructed based on on-chip microring resonators. Background Technology

[0002] With the rapid development of information technology, the amount of data has multiplied, making efficient data transmission and processing crucial. Due to the limitations of the von Neumann architecture, data computation and storage are not on the same unit, requiring additional data transmission. Therefore, computers cannot achieve ultra-high computational efficiency and ultra-low energy consumption. Our brains process vast amounts of information daily—what we see, hear, and smell—and make rapid decisions; the brain is a prime example of efficient data processing. The brain can process 10... (The sentence is incomplete and requires more context to translate accurately.) 20 With a data throughput of MAC / s and a computational efficiency approximately nine orders of magnitude higher than today's supercomputers, the brain, as a highly efficient data processor, has been extensively studied and simulated by researchers. Against this backdrop, neuromorphic computing has become an important research topic. Due to the advantages of optics, such as high speed, low power consumption, low crosstalk, scalability, and high interconnect bandwidth, combining neuromorphic computing with optics has become another important area of ​​research. Neuromorphic photonics primarily utilizes the isomorphism between optical devices and biological neurons. It uses optical devices to construct artificial neurons, and then builds neural networks based on these artificial neurons, thereby achieving neuromorphic computing. Therefore, how to construct artificial neurons using optical devices has become a crucial step in realizing neuromorphic computing.

[0003] Optical neurons primarily utilize optical devices to realize the signal processing processes of biological neurons. This signal processing mainly includes three processes: weighting, summation, and nonlinearity. The weighting process can be implemented using devices such as microring resonators, Mach-Zehnder interferometers, attenuators, and memristors. The summation process can be implemented using photodetectors and waveguide arrays. The nonlinearity can be implemented using Mach-Zehnder modulators, semiconductor optical amplifiers, nonlinear fibers, and lasers. Currently, in implemented optical neurons, the nonlinearity component is not implemented using on-chip microring resonators. Summary of the Invention

[0004] The innovation of this invention lies in proposing an optical neuron based on an on-chip microring resonator. Both the weighting and nonlinear components of the optical neuron are implemented using an on-chip microring resonator. The weighting component primarily utilizes the thermo-optical effect of the microring resonator. By applying a voltage to a thermally adjustable resistor to generate heat, the resonant wavelength of the microring resonator shifts, causing a change in the transmission state at the resonant peak, corresponding to a shift in the weight between low and high values. The nonlinear component mainly utilizes the nonlinear characteristics of the transmission spectrum at the direct-pass terminal of the microring resonator, achieving nonlinear processing under the excitation of a summing signal.

[0005] To achieve the above objectives, this invention proposes an optical neuron based on an on-chip microring resonator, comprising, from front to back, a weighting unit, a summing unit, and a nonlinear unit. The weighting unit includes a microring resonator and a thermally adjustable resistor. The summing unit includes a photodetector, a preamplifier, and a switching resistor. The nonlinear unit includes a microring resonator and a thermally adjustable resistor.

[0006] Wherein: the micro-ring resonator of the weighting unit adopts a wavelength division multiplexing architecture, multiplexing one download end optical waveguide; the preamplifier of the summing unit, combined with the conversion resistor, is used to drive the micro-ring resonator of the nonlinear unit; the resistance value of the conversion resistor of the summing unit is equivalent to the resistance value of the thermally adjustable resistor.

[0007] Preferably, the on-chip microring resonator adopts an on-off type structure;

[0008] Preferably, the waveguide height of the on-chip microring resonator is 220-800nm, the width is 0.4-2.8μm, and the material is independently selected from Si and Si3N4;

[0009] Preferably, the thermally adjustable resistor of the on-chip microring resonator has a width of 1-3 μm, a thickness of 100-200 nm, and is made of TiN independently.

[0010] Preferably, the on-chip microring resonator of the weighting unit has a radius of 10-30 μm, and adjacent microring resonators have the same radius difference of 0.008-0.05 μm.

[0011] Preferably, the photodetector of the summing unit has a response wavelength of 1500-1600nm and a responsivity of 0.7-1A / W.

[0012] Preferably, the summing unit uses a preamplifier with a gain of 1-5 times;

[0013] Preferably, the switching resistor of the preamplifier of the summing unit has a resistance value comparable to that of the thermally adjustable resistor, which is 100-500 ohms.

[0014] The basic working principle of this invention is to utilize optical devices to realize the weighting, summing, and nonlinear processing of signals by biological neurons. The weighting unit is implemented using a microring resonator. By applying voltage to a thermally adjustable resistor to generate heat, the resonant wavelength of the microring resonator shifts, changing the transmission state at the resonant peak, thereby achieving signal weighting. The summing unit is implemented using a photodetector, a preamplifier, and a conversion resistor. At the multiplexing port of the microring resonator in the wavelength division multiplexing architecture, the photodetector detects and acquires the summing current. This current is then converted into a voltage signal by the preamplifier and conversion resistor to drive the subsequent nonlinear unit. The nonlinear unit, also implemented using a microring resonator, primarily utilizes the nonlinear characteristics of the transmission spectrum at the direct-pass end of the microring resonator to achieve the nonlinear processing.

[0015] The advantages of this invention are that it uses on-chip microring resonators as weighting and nonlinear units, resulting in higher integration density; furthermore, due to the excellent cascading performance of microring resonators, the optical neuron proposed in this invention has strong scalability, and the on-chip integration process is compatible with CMOS, enabling large-scale fabrication. This invention facilitates the miniaturization, high integration density, and high scalability of optical neurons. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the architecture of an optical neuron based on an on-chip microring resonator in an embodiment of the present invention;

[0017] Figure 2 This is a photomicrograph of the microring resonator used in the optical neuron based on the on-chip microring resonator in this embodiment of the invention;

[0018] Figure 3 This is a schematic diagram and transmission spectrum of the weighting unit of the optical neuron based on the on-chip microring resonator in an embodiment of the present invention;

[0019] Figure 4 This is a schematic diagram and transmission spectrum of the nonlinear unit of the optical neuron based on the on-chip microring resonator in an embodiment of the present invention;

[0020] Explanation of reference numerals in the attached figures:

[0021] λ - Light source; OA - Optical attenuator; PD - Photodetector; EA - Preamplifier; OSA - Spectrometer. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0023] This invention proposes an optical neuron based on an on-chip microring resonator, where both the weighting and nonlinear components are implemented using on-chip microring resonators, resulting in higher integration density. Because on-chip microring resonators have excellent cascading performance, the optical neuron constructed from them also exhibits high scalability. Furthermore, the on-chip integration process is compatible with CMOS, enabling large-scale fabrication of optical neurons.

[0024] Reference Figure 1 This invention proposes an optical neuron based on an on-chip microring resonator, comprising, from front to back, a weighting unit, a summing unit, and a nonlinear unit. The weighting unit includes a light source, an optical attenuator, a microring resonator, and a thermally adjustable resistor. The summing unit includes a photodetector, a preamplifier, and a switching resistor. The nonlinear unit includes a light source, a microring resonator, a thermally adjustable resistor, and a spectrometer.

[0025] The microring resonator is made of silicon, with a waveguide width of 500 nm and a height of 220 nm. Its micrograph is shown below. Figure 2 The weighting unit uses a C-band light source, and the attenuation coefficient of the optical attenuator is adjustable within a 10dB range. The radius of the micro-ring resonator in the weighting unit is 10μm, and the thermally adjustable resistor is 260 ohms. The photodetector responsivity of the summing unit is 0.8A / W, the preamplifier gain is 1, and the conversion resistor is 260 ohms.

[0026] like Figure 1 As shown, combined with Figure 3 and Figure 4 The signal processing procedure of the optical neuron based on the on-chip microring resonator proposed in this invention is as follows:

[0027] (1) Weighted processing

[0028] Weighting process reference Figure 3 (a). First, multiple wavelengths λ n The optical signal is input to the input port of the micro-ring resonator of the weighting unit, and the input signal is denoted as x. n After weighting by the micro-ring resonator, the weight value is denoted as W. i Then, the optical signal is output from the multiplexed download port as ∑ n W i x i It is important to emphasize that both parameters n and i can be expanded according to the actual situation. The weighting principle can be found in [reference needed]. Figure 3(b) Understanding the transmission spectrum at the input end of the microring resonator: If the microring resonator is initially in a resonant state, i.e., the state at point A, then the weight is high. After thermal tuning, the resonance peak shifts to the right, causing the microring resonator to change from a resonant state to a detuned state, i.e., the state at point B, then the weight is low. Different high and low weights can be applied to the input optical signal by setting different initial microring states and thermal tuning conditions.

[0029] (2) Summation process

[0030] The summation process is as follows: Figure 1 For the summing unit, the input signal is the optical signal ∑ output by the weighting unit. n W i x i The signal is directly input to a photodetector for detection, where it undergoes photoelectric conversion, transforming into a summed photocurrent signal. This summed photocurrent signal is then converted into a summed photovoltage signal [∑] under the combined action of a preamplifier and a conversion resistor. n W i x i ] O / E Microring resonators are used to drive nonlinear units.

[0031] (3) Nonlinear processing

[0032] Nonlinear process reference Figure 4 (a) For the nonlinear unit, the input signal is λ n+1 The excitation signal is the summed photovoltage signal [∑] output by the summing unit. n W i x i ] O / E By utilizing the nonlinear characteristics of the transmission spectrum at the through end of a microring resonator, a nonlinear output signal F([∑ n W i x i ] O / E (), where F represents the nonlinear characteristics of the transmission spectrum at the through end of the microring resonator, i.e., the nonlinear function. The nonlinear principle can be found in [reference needed]. Figure 4 (b) Understanding. The microring resonator is initially set to a detuned state, i.e., the transmission spectrum at the through end is point C. As the excitation signal increases, the microring resonator gradually becomes resonant, and the transmission spectrum at the through end changes from point C to point D. The curve through which the transmission spectrum slides is a nonlinear function.

[0033] The optical neuron based on an on-chip microring resonator proposed in this invention uses the on-chip microring resonator as both the weighting and nonlinear units, resulting in higher integration density. Due to the excellent cascading performance of the microring resonator, the optical neuron proposed in this invention exhibits strong scalability, and the on-chip integration process is compatible with CMOS, enabling large-scale fabrication. This invention is beneficial for the miniaturization, high integration density, and high scalability of optical neurons.

[0034] The specific embodiments described above provide a detailed explanation of the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit, idea, and principle of the present invention should be included within the protection scope of the present invention.

Claims

1. An optical neuron structure based on an on-chip microring resonator, characterized in that: include: From front to back, the components are a weighting unit, a summing unit, and a nonlinear unit. The weighting unit uses a wavelength division multiplexing series microring resonator structure to apply weights to multiple input optical signals. The summing unit includes a photodetector and a preamplifier, used to sum the multiple weighted optical signals and drive the microring resonator of the nonlinear unit. The nonlinear unit uses a microring resonator structure to perform nonlinear output processing on the summed optical signal, utilizing the nonlinear transmission characteristics of the microring resonator. The weighting unit is used to apply weights to multiple input optical signals; The summation unit is used to sum multiple weighted optical signals; the nonlinear unit is used to perform nonlinear output processing on the summed optical signals.

2. The optical neuron structure based on an on-chip microring resonator as described in claim 1, characterized in that, The weighting unit adopts an on-chip micro-ring resonator, which uses an on-chip / off-chip structure.

3. The optical neuron structure based on an on-chip microring resonator as described in claim 1, characterized in that, The weighting unit uses an on-chip microring resonator with a waveguide height of 220-800 nm and a width of 0.4-2.8 μm, and the material is independently selected from Si and Si3N4.

4. The optical neuron structure based on an on-chip microring resonator as described in claim 1, characterized in that, The weighting unit uses an on-chip microring resonator, which adopts a thermally adjustable resistor structure. The thermally adjustable resistor has a width of 1-3 μm, a thickness of 100-200 nm, and is made of TiN.

5. The optical neuron structure based on an on-chip microring resonator as described in claim 1, characterized in that, The weighting unit uses on-chip microring resonators arranged in series in the form of wavelength division multiplexing. The radius of the initial on-chip microring resonator is 10-30 μm, and adjacent on-chip microring resonators have the same radius difference of 0.008-0.05 μm.

6. The optical neuron structure based on an on-chip microring resonator as described in claim 1, characterized in that, The summing unit employs a structure consisting of a photodetector and a preamplifier.

7. The optical neuron structure based on an on-chip microring resonator as described in claim 1, characterized in that, The summing unit uses a photodetector with a response wavelength of 1500-1600 nm and a responsivity of 0.7-1 A / W.

8. The optical neuron structure based on an on-chip microring resonator as described in claim 1, characterized in that, The summing unit uses a preamplifier with a gain of 1-5.

9. The optical neuron structure based on an on-chip microring resonator as described in claim 1, characterized in that, The summing unit uses a preamplifier with a switching resistor value of 100-500 ohms.

10. The optical neuron structure based on an on-chip microring resonator as described in claim 1, characterized in that, The nonlinear unit adopts an on-chip micro-ring resonator, and the on-chip micro-ring resonator adopts an up-down type structure; The weighting unit uses an on-chip micro-ring resonator with a radius of 10-30 μm; The on-chip microring resonator used in the weighting unit uses the same thermally adjustable resistor and material as the on-chip microring resonator in the weighting unit.

Images