Compact photonic device design method based on reverse design algorithm idea
By introducing a reverse design optimization method with a random translation mechanism in photonic device design, the problem that DBS algorithm is prone to fall into local optimal solutions is solved, and more efficient device performance optimization is achieved, improving the compactness and performance of the device.
Patent Information
- Application Number
- CN202410706605.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-03
- Publication Date
- 2024-11-08
AI Technical Summary
In device design, DBS algorithms are prone to fall into local optimal solutions and are difficult to break out of the problem of local optimal solutions, resulting in insufficient device performance.
The reverse design optimization method using a random translation mechanism is adopted. By randomly changing the state and position of pixel points in the space to be optimized, the target performance is evaluated using the FOM function, and the device structure is gradually optimized to achieve the global optimal solution.
Effectively breaking out of local optimal solutions improves device performance, especially in device design, which can be closer to the global optimal solution, improving device compactness and performance.
Smart Images

Figure CN118915306A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to intelligent design of silicon-based photonic devices, and in particular to a method for optimizing compact photonic devices using an inverse design algorithm. Background Art
[0002] Compared with traditional forward design, the reverse design method does not consider the specific structure of the design area when designing a device, but only considers the input and output of the device. It takes meeting the design goals as the first priority and solves the structural parameters of the design area through optimization algorithms.
[0003] As one of the important algorithms for reverse design, the DBS algorithm is widely used in device design, such as wavelength demultiplexers and power dividers based on the DBS algorithm. However, according to our experience, the DBS algorithm is easy to fall into the local optimal solution instead of the global optimal solution. Although the initial mode can be manually set to avoid falling into the local optimal solution as much as possible, this method cannot fundamentally change this shortcoming from the algorithm. In order to get as close to the optimal solution as possible, a unique new method is proposed, an improved version of the DBS algorithm that can be randomly shifted to reduce the device area and improve device performance. Summary of the invention
[0004] The purpose of the present invention is to overcome the shortcomings of the DBS algorithm and provide a new method for photonics inverse design optimization using a random translation mechanism. The method is applicable to any linear or nonlinear photonic device optimization problem; when trapped in a local optimal solution, random translation is used to achieve the problem of jumping out of the local optimal solution; therefore, the method proposed by the present invention is better than the DBS algorithm in its ability to approach the global optimal solution.
[0005] The objective of the present invention is achieved through the following technical solutions:
[0006] A compact photonic device design method based on the inverse design algorithm includes:
[0007] S1: Set the size of the space to be optimized and the size of the input and output waveguides, select the material type of the waveguide, divide the space to be optimized into a number of pixels, and each pixel is composed of air material or waveguide material;
[0008] S2: Construct the initial structure of the space to be optimized. All pixels in the initial structure are a single structure composed of air material or waveguide material, or a known integrated photonic device structure consistent with the target performance. For example, if you want to design a device with a size of 2μm×2μm, then use this 2μm×2μm area as the optimization space, and then randomly generate an n×n matrix with only 0 or 1, and use this matrix to generate the distribution of pixels in the space to be optimized, and use this structure as the initial structure.
[0009] S3: Based on the electromagnetic field simulation method, a FOM function is established to evaluate the target performance of integrated photonic devices. The larger the value of the FOM function, the better the target performance index.
[0010] S4: Select pixel points in order, and use the FOM function to calculate the function value FOM of the initial photonic device structure, and store the calculated FOM in the array FOM_sum in order;
[0011] S5: Change the state of the pixel selected in S4 and calculate the function value FOM now , if FOM now >FOM, the changed pixel state is retained, otherwise the change is canceled and step S4 is executed again;
[0012] S6: After all pixels in the optimization space are traversed, the difference between the first FOM and the last FOM in the array FOM_sum is calculated. If the calculated result is less than 1×10 -6 , then execute step S7, otherwise enter the second iteration and re-execute step S4;
[0013] S7: Randomly select a pixel point. If the pixel point state is "1", randomly translate it and then calculate FOM new Function value, if FOM new If there is an improvement over the previous FOM, the pixel changes are retained;
[0014] S8: If the number of second-stage iterations is reached, the optimization stops and the final design structure is obtained, otherwise, step S7 is executed.
[0015] Furthermore, the space to be optimized refers to the space occupied by the compact photonic device to be optimized.
[0016] Furthermore, the material property of the pixel refers to the type of material filled with the corresponding pixel, and the material type includes air material or waveguide material.
[0017] Furthermore, when executing step S6, it is determined whether to enter the second stage according to the value of FOM:
[0018] (1) Generate a random integer, which is randomly selected from the range of 0 to the maximum number of pixels;
[0019] (2) Select the pixel point corresponding to the random integer. If the state of the pixel point is "0", skip it. Re-execute S7 until the state of the selected pixel point is "1", and then randomly translate it in the X direction and the Y direction.
[0020] (3) After the pixel position is changed for the first time, the FOM value of the objective function of the modified photonic device is calculated based on FDTD. If the FOM value is greater than the last value of the array FOM_sum, the change is retained. Otherwise, the pixel position is restored and step S7 is executed again. After the pixel position is changed for the second time and thereafter, the calculated FOM value is compared with the previous one.
[0021] Furthermore, the changing of the pixel state means that when the current material of the pixel is air material, the current material of the pixel is changed to waveguide material; when the current material of the pixel is waveguide material, the current material of the pixel is changed to air material.
[0022] Furthermore, the electromagnetic field simulation method includes three-dimensional time-domain finite difference method FDTD and finite element analysis method.
[0023] Compared with the prior art, the technical solution of the present invention has the beneficial effect of: taking random translation to jump out of the local optimal solution when trapped in the local optimal solution, while the traditional DBS algorithm does not have this ability. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 The flowchart of the compact photonic device design method based on the inverse design algorithm concept of the present invention.
[0025] Figure 2 In a specific embodiment of the present invention, the initial structure of the space is optimized.
[0026] Figure 3 Initial structure of the device in a specific embodiment of the present invention
[0027] Figure 4 This is the execution process of random search in a specific embodiment of the present invention.
[0028] Figure 5 This is the final optimization result of the specific embodiment of the present invention.
[0029] Figure 6 For the specific embodiment of the present invention Figure 5 Transmittance results of the output port in the design. DETAILED DESCRIPTION
[0030] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0031] This specific embodiment provides a compact photonic device design method based on the inverse design algorithm concept. The specific process is shown in Figure 1, Lumerical software and its built-in script function are used to verify the method, where Lumerical script is used to perform FDTD calculation to obtain FOM function value and perform iterative optimization; the details are as follows:
[0032] First, if Figure 2 As shown, in the Lumerical software, the length, width and height of the space to be optimized are set to 2.0μm×2.0μm×220nm, the air material is (black), and the waveguide material is silicon (white). The length, width and height of the divided pixel points are 100nm×100nm×220nm, that is, the length and width distribution divides the space to be optimized into 20 parts, and a circle with a radius of 45nm is divided out in each pixel point to facilitate subsequent processing. The divided pixel points can be represented by a 20×20 matrix containing only 0 or 1, 1 represents that the material of the pixel point is air, and 0 represents that the material of the pixel point is a waveguide. The general three-dimensional space optimization problem proposed by the present invention is simplified to a two-dimensional plane optimization problem in this specific embodiment. In addition, Figure 2 The input port and output port of the device are given here, both of which are waveguides with a width of 400nm, and the center spacing between input ports 1 and 2 is 800nm. The ultimate goal of this compact photonic device is to realize the XOR function. The laser with a wavelength of 1500nm to 1600nm enters the optimization space from the input port according to the XOR gate truth table. When the input is 1, there is laser incidence, and when the input is 0, there is no laser incidence. Its truth table is shown in Table 1:
[0033] Table 1 XOR gate truth table
[0034] Input Port 1 Input Port 2 Output Port 0 0 0 0 1 1 1 0 1 1 1 0
[0035] Figure 3 The initial structure of the space to be optimized in this specific embodiment is given. The initial structure is a simple XOR gate. Under this structure, laser energy with a wavelength of 1500nm to 1600nm will be emitted from the output port under any circumstances, which obviously does not meet the function of XOR logic and needs to be improved through optimization methods. According to the target performance that the device needs to achieve, the index function is defined as:
[0036] FOM=P 01 +P 10 -1.5*P 11 -3*|P 11 | (1)
[0037] Where P 01 , P 10 , P 11They represent the laser power output from the output port when the input ports are 01, 10, and 11, respectively. From the formula, we can see that only when the input ports are 01 and 10, the laser power is fully output from the output port. When the input port is 11, there is no laser power output from the output port, and FOM=2 reaches the maximum value.
[0038] Figure 4 The execution process of random search in this specific embodiment is given. Figure 4 (a) The randomly selected pixel A is a circle with a radius of 45 nm. If the pixel state is 0, the next pixel is randomly searched; if the pixel state is 1, the pixel needs to be randomly translated in the X and Y directions. The random translation distance is randomly generated by the system and the value of the distance is between 0 and 0.02×10 -6 Directly generate randomly. Then calculate the function value FOM of the device after the pixel is moved. Because the FOM value after the pixel is moved is less than the FOM value calculated last time, the pixel needs to be reset and the above random search process needs to be executed again.
[0039] The above random search process is repeated until the set number of iterations is reached and the optimization stops. After random search optimization, the final optimization result is obtained. The final structure of the device is as follows Figure 5 As shown. Figure 3 , you can see Figure 5 The pixels in the image are not neatly arranged, and many of them show signs of being moved.
[0040] Figure 6 This is the final result of the device output port optimized by random search. When the logic value of the device output port is 1, the transmittance of the laser with a wavelength of 1500nm to 1600nm at the output port is between 36% and 42%; when the logic value of the device output port is 0, the transmittance of the laser with a wavelength of 1500nm to 1600nm at the output port is between 0.4% and 0.8%. The contrast of the device reaches 16.5dB.
[0041] The above description is of a specific embodiment of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in a different order than that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or sequential order shown to achieve the desired results.
[0042] The present invention is not limited to the embodiments described above. The above description of the specific embodiments is intended to describe and illustrate the technical solution of the present invention. The above specific embodiments are merely illustrative and not restrictive. Without departing from the scope of the present invention and the scope of protection of the claims, a person of ordinary skill in the art can also make many forms of specific changes under the guidance of the present invention, which all fall within the scope of protection of the present invention.
Claims
1. A compact photonic device design method based on the inverse design algorithm, characterized in that: The following steps are involved: S1: Set the size of the space to be optimized and the size of the input and output waveguides, select the material type of the waveguide, divide the space to be optimized into a number of pixels, and each pixel is composed of air material or waveguide material; S2: construct an initial structure of the optimization space, in which all pixels are a single structure composed of air material or waveguide material, or a known integrated photonic device structure consistent with the target performance; S3: Based on the electromagnetic field simulation method, a FOM function is established to evaluate the target performance of integrated photonic devices. The larger the value of the FOM function, the better the target performance index. S4: Select pixel points in order, and use the FOM function to calculate the function value FOM of the initial photonic device structure, and store the calculated FOM in the array FOM_sum in order; S5: Change the state of the pixel selected in S4 and calculate the function value FOM now , if FOM now >FOM, the changed pixel state is retained, otherwise the change is canceled and step S4 is executed again; S6: After all pixels in the optimization space are traversed, the difference between the first FOM and the last FOM in the array FOM_sum is calculated. If the calculated result is less than 1×10 -6 , then execute step S7, otherwise enter the second iteration and re-execute step S4; S7: Randomly select a pixel point. If the pixel point state is "1", randomly translate it and then calculate FOM new Function value, if FOM new If there is an improvement over the previous FOM, the pixel changes are retained; S8: If the number of second-stage iterations is reached, the optimization stops and the final design structure is obtained, otherwise, step S7 is executed.
2. A compact photonic device design method based on the inverse design algorithm concept according to claim 1, characterized in that: The space to be optimized refers to the space occupied by the compact photonic device to be optimized.
3. A compact photonic device design method based on the inverse design algorithm concept according to claim 1, characterized in that: The material property of the pixel point refers to the type of material filled with the corresponding pixel point, and the material type includes air material or waveguide material.
4. A compact photonic device design method based on the inverse design algorithm concept according to claim 1, characterized in that: When executing step S6, it is determined whether to enter the second stage according to the value of FOM: (1) Generate a random integer, which is randomly selected from the range of 0 to the maximum number of pixels; (2) Select the pixel corresponding to the random integer. If the state of the pixel is "0", skip it. Re-execute S7 until the state of the selected pixel is "1". Then randomly translate it in the X and Y directions. (3) After the pixel position is changed for the first time, the FOM value of the objective function of the modified photonic device is calculated based on FDTD. If the FOM value is greater than the last value of the array FOM_sum, the change is retained; otherwise, the pixel position is restored and step S7 is re-executed; after the pixel position is changed for the second time and thereafter, the calculated FOM value can be compared with the previous one.
5. The compact photonic device design method based on the inverse design algorithm concept according to claim 1, characterized in that: Changing the pixel state means that when the current material of the pixel is air material, the current material of the pixel is changed to waveguide material; when the current material of the pixel is waveguide material, the current material of the pixel is changed to air material.
6. A compact photonic device design method based on the inverse design algorithm concept according to claim 1, characterized in that: The electromagnetic field simulation method includes three-dimensional time domain finite difference method FDTD and finite element analysis method.
