A micro-crack optical waveguide multi-stage radial bionic nerve branch topology
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-16
- Publication Date
- 2026-08-11
AI Technical Summary
[0002]现有微裂纹光波导的末梢分支多为无规则随机生成,无统一的层级化拓扑设计,无法实现信号的分级处理与差异化存储,类脑仿生程度低;同时随机分支的应力分布无规律,信号留存时长差异大,无法构建稳定的非易失性存储单元
Smart Images

Figure FT_1
Abstract
Description
Technical Field
[0001] This invention discloses a multi-level radial biomimetic neural branch topology of a microcracked optical waveguide, which is disposed in the terminal storage region of the waveguide and extends radially in at least three levels. The number, size, and extension angle of the branches vary gradient along the direction away from the main trunk, and the residual stress increases synchronously with the gradient. It is matched with a quartz-plastic optical fiber composite multi-material collaborative transmission optical path. This configuration balances high-speed signal transmission and long-term signal storage, adapts to the architectural requirements of brain-like photonic computing, and can be stably mass-produced. Background Technology
[0002] Existing microcracked waveguides often have randomly generated terminal branches, lacking a unified hierarchical topology design. This makes it impossible to achieve hierarchical signal processing and differentiated storage, resulting in low neuromorphic biomimicry. Furthermore, the stress distribution of these random branches is irregular, leading to significant variations in signal retention time and hindering the construction of stable, non-volatile storage units. Conventional tree-like topologies in the industry employ a uniform branching design, which only enables signal transmission and cannot address the differentiated needs of both transmission and storage. Moreover, the structural parameters are incompatible with stress-induced cracking processes, making stable mass production through stress-induced methods difficult. Existing topology solutions cannot simultaneously meet the three core requirements of brain-like bionic architecture, differentiated in-memory computing functions, and process adaptability, becoming a key architectural weakness that restricts the improvement of in-memory computing performance of microcracked optical waveguides. Summary of the Invention
[0003] This invention provides a multi-level radial biomimetic neural branch topology, which draws on the hierarchical extension logic of biological neural synapses and combines gradient residual stress design to achieve high-speed signal transmission and long-term signal storage within the same topology. Moreover, the structural parameters are fully compatible with the gradient directional stress-induced cracking process, enabling stable batch production. To achieve the above-mentioned objectives, the present invention adopts the following complete technical solution: A multi-level radial biomimetic nerve branch topology of a microcracked optical waveguide is disposed in the terminal region of the optical waveguide storage medium, extending in a multi-level radial manner, including at least three levels of branches, the number, size, and extension angle of the branches changing in a gradient along the direction away from the main trunk, and the residual stress increasing synchronously in a gradient. Preferably, each branch corresponds to a different extension angle range; Preferably, the residual stress difference between adjacent layers is ≥15MPa; Preferably, the branch hierarchy and extension method can be flexibly adapted to different application scenarios; Preferably, the inner walls of each branch are compositely arranged with quartz optical fiber and plastic optical fiber transmission paths to form a multi-material collaborative transmission architecture. Attached image description:
[0004] Figure 1 is a schematic diagram of the planar layout of the multi-level radial bionic nerve branch topology in an embodiment of the present invention. Detailed Implementation Plan
[0005] This topology is based on the aforementioned multilayer microcracked optical waveguide storage medium. The branch parameters are iteratively optimized through finite element stress simulation to determine the matching relationship between the extension angle, branch density and stress loading parameters of each branch. The optimized topology path is then transformed into the writing trajectory of the femtosecond laser prefabricated three-dimensional guided etching channel. The formed structure extends from the main transmission optical path into first-level, second-level, and third-level branches, with the number of branches increasing progressively, forming a biomimetic neural radial structure. First-level branches are few in number, have large cross-sections, and small extension angles, corresponding to a fully annealed process, resulting in low residual stress and low optical transmission loss, and undertaking the high-speed distribution function of the main signal. Second-level branches serve as transitional layers. Third-level and above terminal branches are numerous, have small cross-sections, and large extension angles, corresponding to a short-term annealing process, resulting in high residual stress and long retention time of the optical signal interference state, undertaking non-volatile storage and weighted memory functions. This embodiment adopts a three-level planar divergent structure, with the first-level branch extending at an angle of approximately 20° and the second-level branch extending at an angle of approximately 35°; the residual stress of the first-level branch is approximately 20 MPa, and the residual stress of the third-level terminal branch is approximately 50 MPa. The difference between adjacent levels is greater than 15 MPa, forming a stable gradient stress distribution, corresponding to differentiated signal retention time. Depending on storage capacity and computing power requirements, it can be expanded to 4 to 6 levels of branches, or adopt a three-dimensional spatial divergent structure to further improve device integration. Preferred Implementation Example 1 (Level 5 High-Capacity Solution) It adopts a 5-level three-dimensional spatial divergent topology structure, with the first-level branch extension angle of 15°, the second-level branch extension angle of 25°, the third-level branch extension angle of 35°, and the fourth and fifth-level terminal branches arranged in a fully spatial divergent pattern; the residual stress difference between adjacent levels is about 20MPa, and the residual stress of the terminal branch is 3 times that of the main branch; the storage capacity is increased by 2.2 times compared with the 3-level scheme, making it suitable for large-capacity storage application scenarios. Preferred Implementation Example 2 (4-Level Hybrid Scheme) It adopts a 4-level planar + spatial hybrid divergence topology, with the first two levels being planar divergence and the last two levels being three-dimensional spatial divergence; it is suitable for edge computing scenarios, balancing transmission efficiency and storage capacity. Instructions for implementing new features This solution introduces a new multi-material collaborative transmission optical path combining quartz fiber and plastic fiber. Quartz fiber transmission paths are compositely deployed on the inner walls of microcracks in the main branches to ensure long-distance, low-loss transmission. Plastic fiber transmission paths are compositely deployed on the inner walls of microcracks in the terminal branches to meet the needs of high curvature and low cost terminal transmission. The multi-material collaborative transmission architecture can reduce overall transmission loss by 15% and reduce manufacturing costs by 30%. Working principle The core working principle of this topology is a hierarchical functional differentiation matching mechanism: by mapping gradient structural parameters to gradient residual stress fields, functional differentiation of different level branches is achieved—the large-section, low-stress near-trunk branches have low transmission loss characteristics, adapting to the needs of high-speed signal distribution; the small-section, high-stress far-trunk branches have long-term phase-locking characteristics, adapting to the needs of non-volatile data storage; the quartz-plastic composite fiber transmission architecture takes into account both the low loss of trunk transmission and the low cost and high adaptability of terminal transmission; the biomimetic radial topology structure greatly improves the branch density and storage capacity, while all structural parameters are deeply coupled with the stress-induced cracking process, and can be stably batch-formed through laser pre-etched channels. Beneficial effects
[0006] It adopts a multi-level radial biomimetic neural topology, which has both signal distribution and storage memory functions. It has a high degree of brain-like biomimicry and is more suitable for the architectural requirements of brain-like photonic computing. The gradient residual stress design corresponds one-to-one with the hierarchical structure. The signal retention characteristics of different levels can be precisely controlled by process parameters without additional doping or secondary processing. The topological parameters are perfectly matched with the gradient-oriented stress-induced cracking process, and it can be stably formed through laser prefabrication channels. It has good batch consistency and is suitable for large-scale preparation. The newly added quartz-plastic composite fiber optic collaborative transmission architecture balances low loss and low cost, significantly improving the cost-effectiveness and adaptability of the solution to various scenarios.
Claims
1. A multi-level radial biomimetic neural branch topology of a microcracked optical waveguide, disposed in the terminal storage region of the optical waveguide storage medium, characterized in that: The structure is multi-level radially extended, with at least three levels of branch microcracks extending from the main transmission optical path. Along the direction away from the main trunk, the number of branches increases and the cross-sectional size decreases step by step. The extension angle and branch density of each level of branch are distributed in a gradient along the radial direction. The further away from the main trunk the branch is, the higher the residual stress value, forming a continuous gradient residual stress distribution.
2. The multi-level radial biomimetic neural branch topology according to claim 1, characterized in that, The extension angle of the first-level branches ranges from 15° to 30°, the extension angle of the second-level branches ranges from 30° to 45°, and the terminal branches of the third level and above are arranged in a divergent pattern.
3. The multi-level radial biomimetic neural branch topology according to claim 1, characterized in that, The residual stress difference between adjacent branches is ≥15MPa, and the residual stress of the terminal branch is 2 to 3 times that of the main branch.
4. The multi-level radial biomimetic neural branch topology according to claim 1, characterized in that, The number of branch levels ranges from 3 to 6.
5. The multi-level radial biomimetic neural branch topology according to claim 1, characterized in that, The branch extension method can be selected from either two-dimensional planar divergence or three-dimensional spatial divergence.
6. The multi-level radial biomimetic neural branch topology according to claim 1, characterized in that, The microcracks at each level of the branch are reinforced with a composite arrangement of quartz fiber and plastic fiber transmission paths. The main optical path uses a quartz fiber path, while the terminal branches use plastic fiber paths, forming a multi-material collaborative transmission optical path.