Wavelength division multiplexing matrix-vector multiplication unit based on micro-ring resonators and photosensitive silicon and implementation method

Abstract

The invention belongs to the technical field of optical signal processing, and provides a matrix-vector multiplication unit based on micro-ring resonators and photosensitive silicon and an implementation method. Systems and method examples disclosed herein include, based on microring resonator structures, photosensitive silicon, and waveguides, completing optical signal screening by utilizing differences in resonant wavelengths of the microring resonators, and mapping weight information onto the transmissivity of the photosensitive silicon utilizing a characteristic that the transmissivity of the photosensitive silicon can vary by illumination intensity. The method has the function of realizing optical matrix-vector multiplication, is simple in structure and rapid in calculation, and meets the requirements of diversified application scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of optical matrix-vector multiplication, and in particular to a wavelength division multiplexing matrix-vector multiplication unit based on a microring resonator and photosensitive silicon and an implementation method thereof. Background Art

[0002] With the rapid development of information technology, traditional electronic integrated circuits are facing increasingly prominent bottlenecks in computing speed and energy efficiency, making it difficult to meet the growing demand for high-performance computing. In contrast, photonic computing, with its ultra-high bandwidth, low latency, and inherent parallel processing capabilities, demonstrates unique advantages in improving computing performance and reducing power consumption, becoming a key approach to breaking through the limitations of traditional computing architectures.

[0003] As a core device in photonic computing, silicon-based microring resonators have been successfully applied in optical interconnects and optical logic operations due to their miniaturization, low power consumption, and excellent anti-interference properties. The device's free spectral range (FSR) allows for precise wavelength selection. The photosensitive silicon material's high absorption coefficient, nanosecond-level response, and wide spectral response from the visible to near-infrared wavelength range provide an ideal platform for building novel optical computing architectures.

[0004] However, existing optical computing systems often face challenges in implementing matrix-vector multiplication, including complex structures, high costs, and asynchronous outputs. Therefore, there is an urgent need to develop a novel matrix-vector multiplication unit with a simple structure, convenient operation, and synchronous outputs to meet the needs of diverse application scenarios. Summary of the Invention

[0005] In response to the above problems, the purpose of the present invention is to provide a matrix-vector multiplication unit based on microring resonators and photosensitive silicon, which realizes the multiplication of m×n matrices and n×1 vectors. The unit is characterized in that m columns of microring resonator multiplication units are integrated, and each column of multiplication units contains n weighted multiplication calculation groups based on different transmittances of photosensitive silicon, where each group consists of a microring resonator, a photosensitive silicon unit, an on-chip delay line and a corresponding waveguide.

[0006] The matrix-vector multiplication unit is characterized in that each multiplication unit is composed of n multiplication calculation groups, each of which is composed of a microring resonator, a photosensitive silicon unit, an on-chip delay line, and a corresponding waveguide. N optical signals are modulated onto n different wavelengths of light. After aggregation, they are split into different specific ratios through a Y-shaped waveguide and equally divided into each column of multiplication units. Light of different wavelengths resonates with the microring resonators corresponding to the resonant frequencies, thus completing the screening of different modulated signals. The modulated signal carried by a specific wavelength enters a multiplication calculation group, which is output from the output of the microring resonator, passes through the photosensitive silicon, completes the multiplication operation with the weight, and finally passes through the delay line to complete the output of the multiplication calculation group. The n different wavelengths of light pass through the n multiplication calculation groups in sequence to complete the above steps, and a synchronous output is obtained at the output of a column of multiplication units. After each column of multiplication units completes the above steps, the output result of the matrix-vector multiplication unit can be obtained.

[0007] The matrix-vector multiplication unit is characterized by the fact that the transmittance of the photosensitive silicon can be changed by light intensity, resulting in an extremely fast response speed. Based on this characteristic, each photosensitive silicon has an independent and non-interfering light source to change its transmittance, thereby achieving the multiplication purpose by setting a weight between 0 and 1.

[0008] The matrix-vector multiplication unit is characterized in that the on-chip delay line can delay the optical signals that are not output synchronously after completing the multiplication operation due to different optical path lengths for a specific time and then output them synchronously so that they can be added at the output end to obtain the calculation result.

[0009] The matrix-vector multiplication unit is characterized in that the function of the matrix-vector multiplication unit is realized as follows: n optical signals a1, a2, ..., an are modulated to input n different wavelengths λ1, λ2, ..., λ n Light is input from the structure's input port and split by a Y-shaped waveguide, with each multiplication group generating 1 / m of light. Within each multiplication group, light with wavelength λ1 passes through the corresponding microring resonator and then through photosensitive silicon with a specific transmittance w11, resulting in a calculation result a1w11. After light signals of different wavelengths pass through n multiplication groups, their calculation results are delayed to produce synchronized outputs: a1w11 + a2w12 + ... a3w1n (using the first column of microring resonator multiplication units as an example). The outputs of the remaining columns are: a1w21 + a2w22 + ... anw2n; ...; a1wm1 + a2wm2 + ... anwmn. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a structural schematic diagram of the present invention.

[0011] Figure 2 It is a schematic diagram of the 2×2 matrix and 2×1 vector multiplication operation of the present invention.

[0012] Figure 3 Schematic diagram of the structure of the silicon-based microring resonator of the present invention. Specific implementation methods

[0013] In order to more clearly illustrate the purpose, technical solutions and advantages of the present invention, the specific implementation details of the present invention will be introduced in depth below with the help of schematic diagrams. Through the following description and the contents of the claims, the advantages and characteristics of the present invention will be fully demonstrated. It should be noted that all the drawings are in a highly simplified form and the proportions are not accurate. Their sole purpose is to assist in understanding the gist of the embodiments of the present invention in an intuitive and clear manner. It is hereby noted that the specific embodiments shown here are only intended to illustrate the present invention and are not intended to limit its scope of application.

[0014] The principle of matrix multiplication is to implement the multiplication operation between an input matrix α and an input vector β, and obtain the output vector γ. The specific calculation formula is as follows:

[0015] like Figure 1 As shown, the present invention can realize the above matrix-vector multiplication operation.

[0016] Figure 2 The figure is a schematic diagram of an embodiment of a 2×2 matrix and 2×1 vector multiplication operation of the matrix-vector multiplication operation unit. The present invention is described in detail below in conjunction with the embodiments and drawings.

[0017] The multiplication formula of a 2×2 matrix and a 2×1 vector is as follows:

[0018] like Figure 2As shown, the vector-matrix multiplication unit is integrated on a silicon substrate, comprising a first column of microring resonator multiplication units 1 and a second column of single-waveguide microring resonator multiplication units 2. Optical signals a1 and a2 are modulated onto wavelengths λ1 and λ2, respectively, and then synthesized and fed into the input port. After being split by a Y-shaped waveguide, the optical signals are equally divided and enter multiplication units 1 and 2, respectively. In multiplication unit 1, when the optical signal passes through the first multiplication unit, the wavelength that satisfies the resonance equation of the microring resonator in this multiplication unit is λ1, and its photosensitive silicon transmittance is w11. Therefore, the output of this multiplication unit is a1w11, and it passes through the delay line. When the optical signal that does not satisfy the resonance equation of the first multiplication unit passes through the second multiplication unit along the optical path, the wavelength that satisfies the resonance equation of the microring resonator in this multiplication unit is λ2, and its photosensitive silicon transmittance is w12. Therefore, the output of this multiplication unit is a2w12, and it passes through the delay line. The optical signals that do not satisfy the above resonance equation are output from the through end. The result of the synchronous output of the optical signals that satisfy the above multiplication unit is Y1=a1w11+a2w12. Similarly, the output of the multiplication unit 2 of the second column of microring resonators is a1w21+a2w22. In this way, the result is (a1w11+a2w12, a1w21+a2w22) T .

[0019] The above description is merely a preferred embodiment of the present invention and does not limit the present invention in any way. Any person skilled in the art who, without departing from the scope of the present invention, makes any equivalent substitution, modification, or other changes to the technical solution and technical content disclosed in the present invention shall be deemed to be within the scope of the present invention and still fall within the scope of protection of the present invention.

Claims

1. A matrix-vector multiplication unit based on a microring resonator and photosensitive silicon for multiplying an m×n matrix by an n×1 vector, characterized in that: M columns of microring resonator multiplication units are integrated, each column of multiplication units contains n weighted multiplication calculation groups based on the different transmittances of photosensitive silicon, and each group consists of a microring resonator, a photosensitive silicon unit, an on-chip delay line and corresponding waveguide.

2. The matrix-vector multiplication unit according to claim 1, characterized in that The multiplication unit is composed of n multiplication calculation groups, each of which is composed of a microring resonator, a photosensitive silicon unit, an on-chip delay line and a corresponding waveguide. The n optical signals are modulated onto n lights of different wavelengths. After aggregation, the light splitting operation of different specific proportions is completed through the Y-type waveguide, and the lights are equally divided into each column of multiplication units. The lights of different wavelengths resonate from the microring resonators with corresponding resonant frequencies, that is, the screening of different modulated signals is completed. The modulated signal carried by the specific wavelength enters a multiplication calculation group, and the modulated signal is output from the output end of the microring resonator, and then passes through the photosensitive silicon to complete the multiplication operation with the weight, and finally passes through the delay line to complete the output of a multiplication calculation group. The n lights of different wavelengths pass through the n multiplication calculation groups in sequence to complete the above steps, and obtain synchronous output at the output end of a column of multiplication units. After each column of multiplication units completes the above steps, the output result of the matrix-vector multiplication unit can be obtained.

3. The matrix-vector multiplication unit according to claim 1, wherein: The transmittance of photosensitive silicon can be changed by light intensity, and it has an extremely fast response speed. Based on this characteristic, each photosensitive silicon has an independent and non-interfering light source used to change its transmittance, thereby setting the weight between 0-1 to achieve the multiplication purpose.

4. The matrix-vector multiplication unit according to claim 1, wherein: The on-chip delay line can delay the optical signals that are not output synchronously after completing the multiplication operation due to different optical path lengths for a specific time and then output them synchronously, so that they can be added at the output end to obtain the calculation result.

5. The matrix-vector multiplication unit according to claim 1, wherein: The function of the matrix-vector multiplication unit is realized as follows: n optical signals a1, a2, ..., an are modulated to input n different wavelengths λ1, λ2, ..., λ n Light is input from the structure's input port and split by a Y-shaped waveguide, with each multiplication group generating 1 / m of light. Within each multiplication group, light with wavelength λ1 passes through the corresponding microring resonator and then through photosensitive silicon with a specific transmittance w11, resulting in a calculation result a1w11. After light signals of different wavelengths pass through n multiplication groups, their calculation results are delayed to produce synchronized outputs: a1w11 + a2w12 + ... a3w1n (using the first column of microring resonator multiplication units as an example). The outputs of the remaining columns are: a1w21 + a2w22 + ... anw2n; ...; a1wm1 + a2wm2 + ... anwmn.

Images