Heat insulation photonic device design method based on constant adiabaticity principle

Abstract

The invention belongs to the technical field of integrated photoelectronics, and particularly relates to a heat insulation photonic device design method based on a constant heat insulation principle. The method comprises the following steps: step 1, system modeling and parameter acquisition; 2, analyzing and extracting core physical factors; and step 3, geometric generation based on a constant adiabaticity principle PCA. The invention aims to solve the technical problems of lengthy device size, dependence on a large amount of numerical iteration in the design process, high calculation cost, sensitivity of performance to parameters, lack of physical interpretability and the like in the existing design method of the heat insulation photonic device. According to the method, traditional geometric tuning and numerical iteration are abandoned, and the optimal structure is deterministically generated by establishing an analytical relation between physical invariants and device geometry.

Description

Technical Field

[0001] This invention belongs to the field of integrated optoelectronics technology, specifically relating to a design method for adiabatic photonic devices based on the principle of constant thermal insulation. Background Technology

[0002] With the rapid development of data centers, artificial intelligence, quantum computing, and other fields, photonic integrated circuits (PICs) have become a key technology for overcoming the "von Neumann bottleneck" in electronic chips due to their high speed and low power consumption. In PICs, thermally adiabatic photonic devices (such as mode converters, couplers, and power dividers) are fundamental components for achieving stable optical signal transmission and processing. They guide the smooth evolution of the optical mode field through gradually varying waveguide structures, thereby achieving high efficiency and high fault tolerance.

[0003] However, the design of traditional thermally adiabatic devices has long been constrained by the fundamental contradiction between size and performance, the so-called "thermal limit." To ensure sufficiently low mode crosstalk (i.e., high thermal insulation), the device's geometry must change very slowly, resulting in lengths typically reaching hundreds of micrometers or even millimeters, severely limiting the integration density and system complexity of photonic chips. To address this issue, existing technologies mainly employ the following two approaches:

[0004] (1) Traditional geometric gradient scheme (such as linear tapered structure) This is the simplest and most common scheme. For example, a linear tapered mode converter, whose waveguide width changes linearly from the input to the output. The core idea of ​​this method is to achieve adiabatic evolution of the mode through a globally uniform, slow rate of geometric change.

[0005] ① Example of technical literature: D. Dai, Y. Tang, and JEBowers, “Mode conversion intapered submicron silicon ridge optical waveguides,” Opt. Express, vol. 20, no. 12, pp. 13425–13439, Jun. 2012. This literature discusses mode conversion in tapered silicon waveguides in detail and is a representative work in this field.

[0006] ② Defects: The fundamental flaw of this scheme lies in its "geometry-driven" rather than "physically-driven" design philosophy. It ignores the local physical characteristics of mode coupling. In some regions of the waveguide, the coupling between different modes is very weak (large propagation constant mismatch), and even rapid geometric changes will not cause crosstalk, but the linear scheme still allocates too long a distance, resulting in wasted space. In other regions (such as near the anti-crossing point of modes), the mode coupling is extremely strong, and the slow rate of change provided by the linear scheme may still be insufficient to suppress crosstalk, leading to performance degradation. The end result is a redundant device and low space efficiency. The manuscript data shows that to achieve a conversion efficiency of 90%, the linear tapered scheme requires a length of 443 micrometers.

[0007] (2) Numerical Optimization Iterative Schemes (such as Slope Loss Algorithm - SLA) To overcome the limitations of linear schemes, researchers have developed iterative optimization algorithms based on numerical simulation. SLA is a representative of these algorithms. This method divides the device into multiple small segments, and then uses a large number of simulation tools such as intrinsic mode extension (EME) to calculate and optimize the length or taper of each segment in order to find an optimal structure for a given total length.

[0008] ① Example of technical literature: T.-L. Liang et al., “A fully numerical method for designing efficient adiabatic mode evolution structures...,” IEEE J. Lightw. Technol., vol. 39, no. 17, pp. 5531-5547, Sep. 2021. This literature is a prior art directly compared in the manuscript, which elaborates on the design process of SLA.

[0009] ② Drawbacks: High computational cost: This method is essentially a "brute-force search," requiring dozens or even hundreds of cyclic simulations, resulting in an extremely long design cycle. Manuscript data shows that designing a single device requires 58 EME cycles, taking approximately 295 minutes; Unstable results and unclear physical meaning: The optimization results are highly sensitive to initial parameters and discretization accuracy, often yielding oscillating performance curves with parasitic resonances. This indicates that the structure is sensitive to manufacturing errors and has poor robustness. Furthermore, the complex structure output typically lacks a clear physical picture, representing a "black box" design; It doesn't address the essence: While methods like SLA shorten device length to some extent, their iterative optimization approach remains within the framework of "geometric tuning," failing to establish a direct analytical relationship between the physical model and device geometry based on first principles.

[0010] In summary, existing technologies are either too simple, resulting in cumbersome devices, or too complex, leading to high design costs and unstable results, failing to fundamentally solve the size-performance bottleneck of thermally adiabatic devices. Therefore, there is an urgent need in this field for a new design paradigm that can analytically and deterministically obtain thermally adiabatic photonic devices with ultra-compact size, ultra-wide operating bandwidth, and high robustness in a low-cost and highly efficient manner. Summary of the Invention

[0011] This invention aims to solve the technical problems of existing thermally adiabatic photonic device design methods, such as lengthy device size, reliance on a large number of numerical iterations in the design process, high computational cost, sensitivity of performance to parameters, and lack of physical interpretability.

[0012] This invention provides a design method for adiabatic photonic devices based on the principle of constant thermal insulation. This method abandons the traditional geometric optimization and numerical iteration, and deterministically generates the optimal structure by establishing an analytical relationship between physical invariants and device geometry.

[0013] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: a design method for adiabatic photonic devices based on the principle of constant thermal insulation, comprising the following steps: Step 1, system modeling and parameter acquisition; Step 2, core physical factor analysis and extraction; Step 3, geometric generation based on the principle of constant thermal insulation PCA.

[0014] Furthermore, as a preferred embodiment of the present invention, step 1 includes the following steps:

[0015] Step 1.1: Determine the device function and geometric variation range: Based on the target function TE1-TM0 mode conversion, determine the operating waveguide width range of the device, starting from the initial width w. start To the final width w end ;

[0016] Step 1.2, Discretize the width variable: Set the total width range w start ,w end Discretized into N width points w i This forms N-1 infinitesimal elements with a width step size Δw. i =w i+1 -w i ;

[0017] Step 1.3, Calculate the effective refractive index of the supermode: Use any standard mode solver to calculate the index over the entire width w. i Within this model, the effective refractive index n of the two coupled core supermodes, ground-state supermode SM1 and higher-order supermode SM2, is calculated. eff,1 (w) and n eff,2 (w). And obtain their difference Δn. eff (w)=neff,1 (w)-n eff,2 (w).

[0018] Furthermore, as a preferred embodiment of the present invention, step 2 includes the following steps:

[0019] Step 2.1: Fitting the Dispersion Curve of Uncoupled Modes: In the width range far from the anti-crossing point of the modes, i.e., the region of strongest coupling, the dispersion curve of the supermode asymptotically approaches that of its uncoupled constant mode; the dispersion curves of the component modes, namely the independent TE1 mode and TM0 mode. Utilizing this physical property, a higher-order polynomial, specifically a quadratic polynomial, is used to fit the asymptotic behavior of the supermode dispersion curve, thereby reconstructing with high precision the effective refractive index n of the virtual, full-width uncoupled component modes. const,TE (w) and n const,TM (w); and calculate the difference Δn. const (w)=n const,TE (w)-n const,TM (w);

[0020] Step 2.2, Calculate the mode hybridization factor H(w): According to the mode anti-crossover theory, the effective refractive index difference Δn of the coupled supermodes eff (w) Effective refractive index difference Δn with uncoupled component modes const There exists a definite physical relationship between w and the mode hybridization factor H(w); w is analytically calculated at each discrete width point using the following relationship. i The mode hybridization factor H(w) on i ):

[0021]

[0022] Where k0 is the free space wavenumber; H(w) is the propagation constant (rad / m), which quantitatively describes the intrinsic coupling strength between two component modes at a specific width w.

[0023] Furthermore, as a preferred embodiment of the present invention, step 3 includes the following steps:

[0024] Step 3.1: Set the adiabatic factor: Based on the trade-off between the total device length and performance, select a dimensionless global control parameter, the adiabatic factor Cadiab. A larger Cadiab means smaller allowable losses per step, a longer total device length, and higher performance. Step 3.2: Calculate the segment length using the PCA analytical formula: Based on the adiabatic perturbation theory and the principle of constant adiabaticity, the length L of any i-th segment... i， Corresponding width change Δw i It is determined deterministically by the following analytical formula:

[0025]

[0026] This formula utilizes the trapezoidal rule for second-order precision to ensure the accuracy of the calculation;

[0027] Step 3.3, Construct the final device geometry: Combine all calculated segment lengths L i Its corresponding width {w i ,w i+1 Sequential connections yield the final, complete geometric description of the adiabatic photonic device with the optimal non-monotonic taper profile.

[0028] The thermally adiabatic photonic device design method based on the principle of constant thermal insulation described in this invention has the following technical advantages compared with the prior art:

[0029] (1) Performance Breakthrough – Breaking the “Adiabatic Limit”: The devices designed using this method far surpass existing technologies in terms of size and bandwidth. Taking the TE1-TM0 mode converter as an example, its length is only 115μm when achieving 90% efficiency, which is 3.8 times shorter than the traditional linear tapered scheme (443μm). At the same time, its operating bandwidth is as high as 1.28μm (1370-2650nm), which is 43% higher than the SLA numerical optimization scheme (0.895μm), and the device length is also shortened by 7.7%.

[0030] 2. Efficiency Revolution – Design Time Reduced by >59 Times: This method transforms the design process from time-consuming numerical iterations to one-time analytical calculations. The entire design time is reduced from approximately 295 minutes for SLA to only about 5 minutes for verification simulation, an efficiency improvement of over 59 times. This significantly accelerates the R&D cycle and enables real-time parameter optimization and large-scale circuit co-design.

[0031] 3. Determinism and Robustness: This method is deterministic; given the input of physical parameters (dispersion curve) and design objectives (Cadiab), a unique and optimal geometry can be obtained. Its output performance curve is smooth and free of parasitic resonances, indicating stronger robustness to length variations and manufacturing errors.

[0032] 4. Physical Interpretability: Unlike "black box" numerical optimization, each step of this invention has a clear physical picture. The final generated non-monotonic taper structure (gradually changing in strongly coupled regions and rapidly changing in weakly coupled regions) is a direct manifestation of the physical law of "constant adiabaticity," providing profound physical insights for the design.

[0033] 5. Universality: Based on universal supermode theory and perturbation theory, this method can be widely applied to the design of various thermally adiabatic photonic devices, such as couplers, power dividers, polarization rotators, mode speckle converters, etc., and has extremely strong versatility. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the TE1–TM0 mode converter in an embodiment of the present invention;

[0035] Figure 2 This is a schematic diagram of the supermodel dispersion curve, uncoupled mode fitting curve, and mode anti-crossing extracted from the core physical factors in this embodiment of the invention.

[0036] Figure 3 This is a performance comparison chart showing the dependence of conversion efficiency on device length between the device designed by the method of the present invention and existing technologies (linear scheme, SLA scheme) in the embodiments of the present invention. Detailed Implementation

[0037] The present invention will be further explained in detail below with reference to the accompanying drawings, so that those skilled in the art can better understand and implement the present invention. However, the following examples are only used to explain the present invention and are not intended to limit the present invention.

[0038] A design method for adiabatic photonic devices based on the principle of constant thermal insulation includes the following steps:

[0039] Step 1: System Modeling and Parameter Acquisition

[0040] Determine the device function and geometric variation range: Based on the target function (e.g., TE1-TM0 mode conversion), determine the operating waveguide width range of the device, starting from the initial width w. start To the final width w end .

[0041] Discretize the width variable: set the total width range w start ,w end Discretized into N width points w i This forms N-1 infinitesimal elements with a width step size Δw. i =w i+1 -w i .

[0042] Calculate the effective refractive index of the supermode: using any standard mode solver (such as Lumerical MODE), over the entire width w i Within this framework, the effective refractive index n of two coupled core supermodes (e.g., ground-state supermode SM1 and higher-order supermode SM2) is calculated. eff,1 (w) and n eff,2 (w). And obtain their difference Δn. eff (w)=n eff,1 (w)-n eff,2 (w).

[0043] Step 2: Core Physics Factor Analysis and Extraction:

[0044] This step is one of the key innovations of this invention. It ingeniously extracts the core physical quantity that was previously difficult to obtain—the mode hybridization factor H(w)—from data that can be directly simulated.

[0045] Fitting the dispersion curve of the uncoupled mode: In a wide region far from the anti-crossing point (i.e., the region of strongest coupling), the dispersion curve of the supermode asymptotically approaches the dispersion curve of its uncoupled constant mode (component modes, such as independent TE1 and TM0 modes). Utilizing this physical property, a higher-order polynomial (such as a quadratic polynomial) is used to fit the asymptotic behavior of the supermode dispersion curve, thereby reconstructing with high precision the effective refractive index n of the virtual, full-width uncoupled component modes. const,TE (w) and n const,TM (w). And calculate its difference Δn. const (w)=n const,TE (w)-n const,TM (w).

[0046] Calculate the mode hybridization factor H(w): Based on the mode anti-crossover theory, the effective refractive index difference Δn of the coupled supermodes. eff (w) Effective refractive index difference Δn with uncoupled component modes const There exists a defined physical relationship between w and the mode hybridization factor H(w). The following relationship allows for the analytical calculation of w at each discrete width point. i The mode hybridization factor H(w) on i ):

[0047]

[0048] Where k0 is the free-space wavenumber. The unit of H(w) is the propagation constant (rad / m), which quantitatively describes the intrinsic coupling strength between two component modes at a specific width w.

[0049] Step 3: Geometry generation based on the constant adiabatic property (PCA) principle:

[0050] This step is the core idea of ​​the invention, which is to force the non-adiabatic power loss generated by the device in each step of geometric micro-element change to be a tiny constant.

[0051] Set the adiabatic factor:

[0052] Based on a trade-off between the overall length and performance of the device, the designer selects a dimensionless global control parameter—the adiabatic factor C. adiab C adiabThe larger the value, the smaller the allowable loss per step, the longer the total length of the device, and the higher the performance.

[0053] Calculating segment length using PCA analytical formula: Based on the theory of adiabatic perturbation and the principle of constant adiabaticity, the length L of any i-th segment is... i (corresponding width change Δw) i It can be deterministically calculated using the following analytical formula:

[0054]

[0055] This formula utilizes the trapezoidal rule for second-order precision to ensure the accuracy of the calculation.

[0056] Constructing the final device geometry: Calculate all segment lengths L i Its corresponding width {w i ,w i+1 By sequentially connecting these components, a complete geometric description of the adiabatic photonic device with an optimal non-monotonic taper profile can be obtained.

[0057] In practical implementation, this embodiment aims to design a TE1–TM0 mode converter on an SOI (Silicon-on-Insulator) platform. The platform structure is as follows: Figure 1 As shown, it includes a 220 nm thick bottom silicon planar layer and a 280 nm thick top silicon ridge waveguide, both embedded in silicon dioxide. This vertical asymmetric structure is necessary for achieving coupling between TE and TM modes.

[0058] Step 1: System Modeling and Parameter Acquisition

[0059] Based on the device's functional requirements, the width of the ridge waveguide was set to vary from an initial value of 1.18 μm to a final value of 1.70 μm. This width range was discretized into 58 segments. Then, using Lumerical MODE software, the following values ​​were calculated and obtained: Figure 2 The effective refractive index dispersion curves of the two supermodes, shown by the solid red line and the dashed black line, are n. eff,1 (w) and n eff,2 (w).

[0060] Step 2: Core Physics Factor Analysis and Extraction:

[0061] observe Figure 2 At a distance far from the central anti-crossing region (approximately 1.44 μm), the red curve exhibits the dispersive characteristics of the TM0 mode, while the black curve exhibits the characteristics of the TE1 mode. We performed quadratic polynomial fitting on the asymptotic portions of these two curves, obtaining the following results: Figure 2 The uncoupled mode dispersion curve n is shown by the blue and pink dashed lines. const,TE (w) and n const,TM (w).

[0062] Then, the Δn obtained from the simulation eff (w) and the fitted Δn const Substituting (w) into formula (1), the mode hybridization factor H(w) at all discrete width points can be calculated. The calculation results show that H(w) reaches a sharp peak near the anti-crossing point, which is completely consistent with the physical expectation.

[0063] Step 3: Geometry generation based on PCA:

[0064] We select a suitable adiabatic factor C. adiab For example, set a target of 0.25% power crosstalk in each segment. Then, take the Δn obtained in the previous step... eff Substituting H(w) and H(w) into the core analytical formula (2), the length L of each segment can be directly calculated. i For example, Table 1 lists the calculation results for all 58 segments. Connecting these segments sequentially yields the final device geometry.

[0065] Table 1 Design parameters of the PCA-based TE1–TM0 mode converter

[0066]

[0067]

[0068] As shown in Table 1, in the central region (approximately 1440 nm wide) where mode coupling is strongest, the taper (Δw) i / L i The taper reaches a minimum value (approximately 0.4 nm / μm), meaning the structural change is the slowest. However, in regions with weak coupling at both ends, the taper is much larger (e.g., 9.71 nm / μm in the first segment), more than 20 times the minimum value. This "adaptive" structure, directly derived from physical laws, is the key to achieving ultra-compactness and high performance in this invention.

[0069] The final simulation results are attached. Figure 3 As shown, the device designed in this embodiment is significantly superior to the prior art in all performance indicators, proving the effectiveness and advancement of the method of the present invention.

[0070] The specific implementation schemes described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific implementation schemes of the present invention and are not intended to limit the scope of the present invention. Any equivalent changes and modifications made by those skilled in the art without departing from the concept and principles of the present invention should fall within the scope of protection of the present invention.

Claims

1. A design method for adiabatic photonic devices based on the principle of constant thermal insulation, characterized in that, Includes the following steps: Step 1: System modeling and parameter acquisition; Step 2: Core physical factor analysis and extraction; Step 3: Geometric generation of PCA based on the principle of constant adiabatic performance.

2. The design method for adiabatic photonic devices based on the principle of constant thermal insulation according to claim 1, characterized in that, Step 1 includes the following steps: Step 1.1: Determine the device function and geometric variation range: Based on the target function TE1-TM0 mode conversion, determine the operating waveguide width range of the device, starting from the initial width w. start To the final width w end ; Step 1.2, Discretize the width variable: Set the total width range w start ,w end Discretized into N width points w i This forms N-1 infinitesimal elements with a width step size Δw. i =w i+1 -w i ; Step 1.3, Calculate the effective refractive index of the supermode: Use any standard mode solver to calculate the index over the entire width w. i Within this model, the effective refractive index n of the two coupled core supermodes, ground-state supermode SM1 and higher-order supermode SM2, is calculated. eff,1 (w) and n eff,2 (w). And obtain their difference Δn. eff (w)=n eff,1 (w)-n eff,2 (w).

3. The design method for adiabatic photonic devices based on the principle of constant thermal insulation according to claim 2, characterized in that, Step 2 includes the following steps: Step 2.1: Fitting the Dispersion Curve of Uncoupled Modes: In the width range far from the anti-crossing point of the modes, i.e., the region of strongest coupling, the dispersion curve of the supermode asymptotically approaches that of its uncoupled constant mode; the dispersion curves of the component modes, namely the independent TE1 mode and TM0 mode. Utilizing this physical property, a higher-order polynomial, specifically a quadratic polynomial, is used to fit the asymptotic behavior of the supermode dispersion curve, thereby reconstructing with high precision the effective refractive index n of the virtual, full-width uncoupled component modes. const,TE (w) and n const,TM (w); and calculate the difference Δn. const (w)=n const,TE (w)-n const,TM (w); Step 2.2, Calculate the mode hybridization factor H(w): According to the mode anti-crossover theory, the effective refractive index difference Δn of the coupled supermodes eff (w) Effective refractive index difference Δn with uncoupled component modes const There exists a definite physical relationship between w and the mode hybridization factor H(w); w is analytically calculated at each discrete width point using the following relationship. i The mode hybridization factor H(w) on i ): Where k0 is the free space wavenumber; H(w) is the propagation constant (rad / m), which quantitatively describes the intrinsic coupling strength between two component modes at a specific width w.

4. The adaptive segmented design method for thermally adiabatic optical waveguide couplers based on physical sensitivity according to claim 3, characterized in that, Step 3 includes the following steps: Step 3.1: Set the adiabatic factor: Based on the trade-off between the total length of the device and its performance, select a dimensionless global control parameter, the adiabatic factor Cadiab; the larger the Cadiab, the smaller the allowable loss per step, the longer the total length of the device, and the higher the performance. Step 3.2: Calculate the segment length using the PCA analytical formula: Based on the adiabatic perturbation theory and the principle of constant adiabaticity, the length L of any i-th segment is... i， Corresponding width change Δw i It is determined deterministically by the following analytical formula: This formula utilizes the trapezoidal rule for second-order precision to ensure the accuracy of the calculation; Step 3.3, Construct the final device geometry: Combine all calculated segment lengths L i Its corresponding width {w i ,w i+1 Sequential connections yield the final, complete geometric description of the adiabatic photonic device with the optimal non-monotonic taper profile.

Images