Optical waveguide optimization design method based on equivalent adiabatic degree criterion and optical waveguide

Abstract

The invention belongs to the technical field of integrated photoelectronics, and particularly relates to an optical waveguide optimization design method based on an equivalent adiabatic degree criterion and an optical waveguide. The method comprises the following steps: step 1, initial setting and MCTE curve acquisition; 2, defining a globally unified equivalent adiabatic factor (EAF); 3, based on an EAF criterion, iteratively correcting width division; and 4, iterative convergence is carried out, and a final structure is determined. The optical waveguide is formed by connecting N sections in series, and the combination of the width difference delta W and the length L of each section accurately enables the mode conversion performance index of each section to fall on the same preset equivalent adiabatic factor (EAF) working point located in a non-oscillation area of an MCTE curve.

Description

Technical Field

[0001] This invention belongs to the field of integrated optoelectronics technology, specifically relating to an optical waveguide optimization design method and an optical waveguide based on the equivalent adiabaticity criterion. Background Technology

[0002] In integrated optical circuits, segmented thermally adiabatic waveguides are the core components for achieving low-loss optical connections between different functional regions. The core challenge in their design lies in the "efficiency-size" contradiction: achieving high-efficiency mode thermal evolution typically requires a longer device size; while chip integration requires the device to be as compact as possible.

[0003] The current mainstream design method is an iterative optimization method, and its general process is as follows:

[0004] (1) Preset width profile: The designer first determines a fixed waveguide width partitioning scheme along the length based on experience, analytical functions (such as linear, exponential) or intuition.

[0005] (2) Optimize the length of each segment: Under this fixed width profile, the length of each segment is iteratively adjusted through numerical simulation (such as FDTD, EME, etc.) to find the shortest total length that meets the total loss requirements.

[0006] This traditional approach suffers from a fundamental and unavoidable flaw: the final performance of the design is severely constrained by the initial, empirically determined width profile. A suboptimal width segmentation scheme fundamentally limits the upper limit of the device's performance. No matter how finely the length is subsequently optimized, the result is only a "local optimum" within this constrained framework, far from a true global optimum. For example, if the width variation (ΔW) of a certain segment is preset too large, mode evolution at that point becomes very "difficult," requiring an extremely long length (L) to compensate, resulting in an overly long device size; conversely, if ΔW is preset too small, it leads to wasted length.

[0007] Existing technologies lack a clear, physically meaningful global guideline that can guide how to correct the width profile itself, resulting in a design process that is highly dependent on trial and error, inefficient, and makes it difficult to simultaneously achieve optimal performance and manufacturing tolerance of the final product. Summary of the Invention

[0008] This invention aims to completely solve the fundamental problem in existing technologies where the reliance on experience for width profiles leads to local optima in design. This invention provides a novel design paradigm that introduces a unified physical constraint that balances performance and robustness to derive and determine the optimal width partitioning scheme in reverse engineering. This allows for the synergistic optimization of optical waveguide transmission efficiency, device compactness, and manufacturing tolerance during the design phase.

[0009] This invention discloses an optical waveguide optimization design method and an optical waveguide based on the equivalent adiabatic criterion. This method overturns the traditional approach of "fixed width, optimized length," and its core is to mandate that each independent segment of the waveguide must meet a completely identical local performance standard designed to ensure process robustness. Finally, the width difference (ΔW) of each segment is iteratively adjusted until all segments can "fairly" meet this globally unified standard.

[0010] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: an optical waveguide optimization design method based on the equivalent adiabatic criterion, comprising the following steps: Step 1: initial setting and MCTE curve acquisition; Step 2: defining a globally unified equivalent adiabatic factor EAF; Step 3: iteratively correcting the width division based on the EAF criterion; Step 4: iterative convergence to determine the final structure.

[0011] As a further preferred embodiment of the present invention, step one specifically involves: dividing the waveguide to be designed into N (N ≥ 2) segments along the optical transmission direction. An arbitrary initial width division scheme is set (such as simple linear interpolation). For each of these N segments, a numerical simulation tool is used independently to perform parameter scanning, scanning its length L, calculating and plotting the mode conversion transmission efficiency (MCTE) curve of that segment as a function of its length, i.e., the MCTE(L) curve.

[0012] Furthermore, as a preferred embodiment of the present invention, in step two, a globally unified EAF is defined based on the following two criteria for all MCTE(L) curves of all segments:

[0013] Robustness-first criterion: On the MCTE(L) curve, the EAF operating point must be selected in the top saturation region, but clearly below and avoiding the interference oscillation region near the peak. This oscillation region represents the Fabry-Perot interference effect, which is sensitive to process errors. Therefore, this invention actively abandons the pursuit of the theoretically limiting peak efficiency, sacrificing minimal theoretical performance (e.g., choosing 99% of the peak efficiency instead of 100%) in exchange for robustness of device performance to manufacturing errors.

[0014] Equivalence Criterion: This EAF operating point (e.g., efficiency = 99%) must be used as a globally uniform standard and enforced on all N segments. The physical meaning of this criterion is: requiring that the "mode evolution task difficulty" of each segment is equivalent, ensuring that the thermal insulation performance of the entire device is smoothly and evenly distributed, and that no segment is a performance bottleneck or design redundancy.

[0015] Furthermore, as a preferred embodiment of the present invention, under the current width division scheme, the coupling length required for each segment to reach the defined unified EAF standard is calculated, denoted as L. reqCompare L across all segments. req value:

[0016] If L in a certain segment req The difference is significantly longer than the average, indicating that the current width difference ΔW is too large, making model evolution difficult (insufficient adiabaticity).

[0017] If L in a certain segment req The fact that it is significantly shorter than the average indicates that its current ΔW is too small, the design is too conservative, and the length budget is not being fully utilized.

[0018] Based on the above judgment, the width division scheme was directly adjusted and optimized: for L req For excessively long segments, reduce their ΔW; for L req For segments that are too short, increase their ΔW. Essentially, this is a spatial redistribution of the "task difficulty" of pattern evolution.

[0019] Furthermore, as a preferred embodiment of the present invention, a modified new width division scheme is adopted, and steps one to three are repeated. When the L of all segments req The iteration ends when all segments converge to a balanced and reasonable range. The resulting width partitioning scheme is the optimal solution. The precise length of each segment of the final device is then determined by the converged L... req Determined, the total length is L for each req sum.

[0020] An optical waveguide based on the equivalent adiabatic factor criterion is proposed. The optical waveguide is composed of N segments connected in series. The combination of the width difference ΔW and the length L of each segment precisely ensures that its mode conversion performance index (such as transmission efficiency) falls on the same preset equivalent adiabatic factor (EAF) operating point located in the non-oscillating region of its MCTE curve.

[0021] The optical waveguide optimization design method and optical waveguide based on the equivalent adiabaticity criterion described in this invention have the following technical advantages compared with the prior art:

[0022] (1) Breaking through local optima and approaching physical limits: This invention breaks through the performance bottleneck brought about by traditional methods by directly optimizing the "width profile" which has a higher degree of design freedom and a more fundamental impact, so that the final design can be closer to the physical "efficiency-size" Pareto front.

[0023] (2) Design takes manufacturing into account, inherent high robustness: This invention takes "manufacturing tolerance" as the core design principle and actively avoids performance-sensitive areas by selecting EAF, so that the designed device is inherently insensitive to process deviations (such as etching errors), which can significantly improve chip yield.

[0024] (3) Clear physical guidance, say goodbye to the "black box": The "equivalent adiabatic" criterion of this invention provides designers with a clear physical picture. The imbalance of Lreq directly and quantitatively points out the unreasonableness of the width profile and the direction of modification, making the optimization process transparent and efficient.

[0025] (4) High efficiency and fast convergence: Since the optimization objective (to equalize the Lreq of all segments) is very clear, the iteration process usually converges quickly in 2-4 times. Compared with the traditional global optimization algorithm that often requires tens of thousands of calculations, the design efficiency is improved by orders of magnitude. Attached Figure Description

[0026] Figure 1 This is a flowchart of the method of the present invention in an embodiment of the present invention;

[0027] Figure 2 This is a typical schematic diagram of the MCTE(L) curve of a single waveguide segment in an embodiment of the present invention, with segment 6 as an example.

[0028] Figure 3 This is a comparison diagram of the optimal width profile (solid line) and the traditional linear profile (dashed line) in the embodiments of the present invention. Detailed Implementation

[0029] 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.

[0030] This embodiment aims to design an ultra-compact, low-loss tapered waveguide for use on a 1550 nm wavelength SOI (silicon-on-insulator) platform, transitioning from a 1.5 μm wide multimode waveguide at the input to a 5.0 μm wide single-mode waveguide at the output. The waveguide core layer thickness is 300 nm.

[0031] Reference Figure 1 and Figure 2 Design process:

[0032] 1. Initial settings and MCTE curve acquisition (Step 1):

[0033] We divide the tapered waveguide of unknown total length into N = 14 segments. As the starting point for iteration, we use simple linear interpolation to generate an initial width profile. Using the EME (Eigenmode Expansion) method, we scan the length L of each of these 14 segments with known widths, obtaining 14 independent MCTE(L) curves, as shown below. Figure 2 As shown in segment 6.

[0034] 2. Define the global EAF (Step Two):

[0035] We observed that all MCTE(L) curves exhibited significant Fabry-Perot interference oscillations after reaching a peak efficiency of approximately 99.99%. Figure 3 (The "oscillation region" in the text). Directly targeting the peak value will make the device extremely sensitive to manufacturing errors in length and width.

[0036] Based on the robustness priority criterion and equivalence criterion of this invention, we set the globally uniform EAF as: an operating point with a transmission efficiency of 99.6%. This operating point is located in the efficient saturation region, but completely avoids the unstable oscillation region.

[0037] 3. Iteratively correct the width profile (step three):

[0038] (1) First iteration: We calculate the length L required for each segment to achieve 99.6% efficiency under a linear width profile. req The results showed a severe imbalance: in the first few sections where the width changed gradually, L... req < 4 μm; while in the last few sections where the width changes drastically, L req > 1 μm.

[0039] (2) Adjust the width profile: Based on the above results, we reduce the width difference ΔW in the front section and increase the width difference ΔW in the back section. The purpose is to make the "task difficulty" of the pattern evolution more evenly distributed in all sections.

[0040] (3) Second and third iterations: Repeat the above process using the new width profile. After the third iteration, all nine segments achieve the L required to reach 99.6% of the EAF standard. req All converged to an equilibrium range of 6 μm ± 2.0 μm. Iterative convergence was achieved.

[0041] 4. Determine the final structure (Step Four):

[0042] The width profile obtained after iterative convergence is determined to be the optimal solution. This profile exhibits a nonlinear, counterintuitive shape (such as...). Figure 3 The solid line shown is significantly different from the simple linear (dashed line) cross-section. The final segment length of the device is L after the third iteration converges. req The total length of the device is approximately 14 × 6 μm = 84 μm.

[0043] The 100 μm long tapered waveguide designed using the method of this invention has been verified by simulation to have an insertion loss of less than 0.05 dB at a wavelength of 1550 nm, and a loss variation of less than 0.1 dB under a width manufacturing deviation of ±20 nm, demonstrating excellent performance and high robustness, which is far superior to linear or exponential tapered waveguides of the same length.

[0044] 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 method for optimizing design of an optical waveguide based on an equivalent adiabaticity criterion, characterized in that, The method comprises the following steps: Step 1: initial setting and MCTE curve acquisition; Step 2: defining a globally unified equivalent adiabaticity factor EAF; Step 3: based on the EAF criterion, iteratively correcting the width division; Step 4: iterative convergence to determine the final structure.

2. The method of claim 1, wherein, Step 1 specifically comprises the following steps: dividing the waveguide to be designed into N segments along the light transmission direction, N ≥ 2; setting an arbitrary initial width division scheme; for each of the N segments, independently using a numerical simulation tool to perform parameter scanning, scanning the length L, and calculating and plotting the mode conversion transmission efficiency MCTE curve of the segment, i.e. MCTE(L) curve.

3. The method of claim 2, wherein, In step 2, the MCTE(L) curves of all segments are defined based on the following two criteria: Robustness priority criterion: on the MCTE(L) curve, the EAF working point must be selected in the top saturation region, but must be clearly below and avoid the interference oscillation region near the peak; Equivalent criterion: this EAF working point must be used as a globally unified standard and be applied to all N segments.

4. The method of claim 3, wherein, In step three, the coupling length required for each section to reach the defined uniform EAF criteria is calculated, denoted as L req ; the L req values of all sections are compared: If the L req Significantly longer than the average value, indicating that the current width difference ΔW is too large, the pattern evolution is difficult, and the adiabaticity is insufficient; If the L req Significantly shorter than the average, indicating that its current ΔW is too small, the design is too conservative, and the length budget is not fully utilized; According to the above judgment, directly adjust and optimize the width division scheme: for L req Too long section, reduce its ΔW; for L req Too short section, increase its ΔW.

5. The method of claim 4, wherein the method further comprises: In step four, repeat steps one to three with the modified new width division scheme. When all L req converge to a balanced and reasonable range, the iteration ends. The width division scheme at this time is the optimal solution. The final lengths of the segments are determined by the converged L req , and the total length is the sum of all L req .

6. The optical waveguide based on the method of optical waveguide design optimization based on the equivalent adiabaticity criterion according to any one of claims 1-5, characterized in that, The optical waveguide is composed of N segments in series, wherein the combination of the width difference ΔW and the length L of each segment is precisely such that its mode conversion performance index falls on the same preset equivalent adiabaticity factor EAF working point located in the non-oscillation region of its MCTE curve.

Images