Fully vertical grating coupler and method of manufacturing the same

Abstract

The application provides a fully vertical grating coupler and a preparation method thereof, which maps structure parameters of a grating unit to three physically discrete heights; a high-performance initial solution is obtained by optimizing convergence, iteration number and discrete degree requirements corresponding to a wideband coupling efficiency target function; global discrete correction is performed on insufficiently discrete variables, so that single-mode fiber incident light is efficiently and widely coupled to a TE0 mode of a waveguide on a chip under the condition that manufacturing process constraints are strictly met; the optimized grating unit is locked at the three discrete heights, without complex gray-scale lithography, and only relying on mature processes to flow the chip, which can effectively improve the yield of large-scale manufacturing; the discrete advantage of the algorithm is utilized and high-dimensional optimization disaster is avoided, so that the calculation efficiency is improved and the design cycle is reduced.

Description

Technical Field

[0001] This application relates to the field of grating couplers, and more particularly to a fully vertical grating coupler and its fabrication method. Background Technology

[0002] Silicon-based optoelectronics technology, due to its CMOS process compatibility and large-scale integration potential, has become an important technological route for realizing high-speed optical interconnects and on-chip optical networks. Efficient coupling between optical fibers and on-chip waveguides is a key issue. Since the mode field size of a single-mode fiber is much larger than the lateral size of a silicon-based single-mode waveguide, a significant mode field mismatch exists between the two. Therefore, specialized coupling structures are typically required to achieve mode field conversion. Current mainstream solutions mainly include end-face couplers and grating couplers: the former usually has higher coupling efficiency and wider bandwidth, but requires end-face dicing and polishing and can only be placed at the chip edge; the latter allows for wafer-level testing and flexible layout, and is therefore widely used in silicon photonic chips.

[0003] To improve the performance of vertical or near-vertical grating couplers, various solutions have been proposed in the prior art. For example, some solutions combine chirped blazed grating regions with uniform blazed grating regions and reverse-engineer parameters such as grating period, etching depth, groove width, and fiber position through particle swarm optimization; others use tilted fully etched gratings combined with subwavelength structures and apodization design to improve upward diffraction directionality and reduce back reflection; still others employ reverse-engineering of free-form / random-form grating structures on 220 nm SOI or similar platforms, improving vertical coupling efficiency and bandwidth by defining objective functions and optimization variables. The disclosed solutions include both parameterized structure optimization and free-form reverse design.

[0004] While the aforementioned approaches have made some progress in coupling efficiency, directionality, and compactness, several shortcomings remain: some approaches rely on blazing profiles, tilted etching, subwavelength recombination periods, or multi-level etching structures, which place high demands on sidewall angles, critical dimension consistency, and mask alignment accuracy, making fabrication difficult; while reverse-engineered structures are typically complex and, although they can achieve high performance, are prone to center wavelength drift, bandwidth narrowing, or decreased coupling efficiency when corrected for actual manufacturing; furthermore, optimization in high-dimensional parameter spaces often incurs significant computational costs, and the process tolerance and repeatability of complex free-form structures remain limited. Therefore, existing technologies struggle to guarantee fully vertical coupling performance on a standard 220 nm SOI platform while simultaneously addressing strict discrete manufacturing constraints and fabrication feasibility. Summary of the Invention

[0005] The following is an overview of the topics described in detail in this article.

[0006] The purpose of this application is to at least partially solve one of the technical problems existing in the related technologies. The embodiments of this application provide a fully vertical grating coupler and its fabrication method.

[0007] According to an embodiment of the first aspect of this application, a method for fabricating a fully vertical grating coupler is provided. The fully vertical grating coupler comprises, from bottom to top, a silicon substrate layer, a silicon buried oxide layer, a silicon grating layer, and a silicon top cover layer. The silicon grating layer includes a plurality of grating units. The method includes: The initial structural parameters of the grating unit are mapped to three preset physical discrete heights through a projection function to obtain the first candidate structural parameters. The three preset physical discrete heights include the bottom height of full etching, the middle height of shallow etching, and the maximum height of zero etching. The broadband coupling efficiency corresponding to the first candidate structure parameters is calculated as the objective function. The first candidate structure parameters are optimized according to the convergence, iteration number and dispersion requirements corresponding to the objective function, and the second candidate structure parameters are determined from the optimized first candidate structure parameters. The second candidate structural parameter whose error is within the tolerance range is snapped to the standard discrete height and locked to obtain the third candidate structural parameter. The undiscrete second candidate structural parameter whose error is not within the tolerance range is discretized to obtain the fourth candidate structural parameter. The third candidate structural parameter and the fourth candidate structural parameter are merged to obtain the fifth candidate structural parameter. The discrete height of the fifth candidate structural parameter is replaced for optimization. The objective function of the fifth candidate structural parameter with the replaced discrete height is calculated to obtain a new objective function. When the new objective function is less than the objective function, the acceptance probability is calculated based on the current annealing temperature. The acceptance probability is used to determine whether to accept the replacement of the discrete height. The target structural parameter is obtained after the optimization is completed. A fully vertical grating coupler was fabricated based on the target structural parameters.

[0008] According to certain embodiments of the first aspect of this application, the projection function is: ; in, ρ is the output of the projection function, β is the continuous variable normalized by the initial structure parameters, β is the parameter controlling the projection steepness, and η is the jump threshold.

[0009] According to certain embodiments of the first aspect of this application, the first candidate structural parameter is represented as follows: ; in, The actual physical height of the i-th grating unit. The continuous variable is the normalized initial structural parameters of the i-th grating unit. The first transition threshold, The second transition threshold, The bottom height of the full etching. This represents the intermediate height of the shallow etching. The maximum height for zero etching. This is the output of the projection function.

[0010] According to certain embodiments of the first aspect of this application, the broadband coupling efficiency is expressed as: ; in, ; Let T(λ) represent the broadband coupling efficiency, where λ is the coupling efficiency at different wavelengths. For flat-top Gaussian broadband weights, λ c λ is the center wavelength, Δλ is the target flat bandwidth, and M is the super-Gaussian order.

[0011] According to certain embodiments of the first aspect of this application, optimizing the first candidate structure parameters based on the convergence, iteration number, and dispersion requirements corresponding to the objective function includes: If the current iteration does not converge and the maximum number of iterations has not been reached, perform gradient optimization and update the initial structural parameters, then return to the step of mapping the initial structural parameters of the grating unit to three preset physical discrete heights through a projection function to obtain the first candidate structural parameters and continue to the next iteration; If the current iteration converges or reaches the maximum number of iterations, but the discreteness requirement is not met, return to the step of mapping the initial structural parameters of the grating unit to three preset physical discrete heights through the projection function to obtain the first candidate structural parameters, and increase the parameter controlling the projection steepness of the projection function to add noise perturbation to the initial structural parameters. If the current iteration converges or reaches the maximum number of iterations and meets the discreteness requirement, save the current first candidate structure parameters.

[0012] According to certain embodiments of the first aspect of this application, adding noise perturbation to the initial structural parameters includes: Identify initial structural parameters that are stuck near a transition threshold, and add normally distributed random noise to these parameters.

[0013] According to certain embodiments of the first aspect of this application, the undiscreteized second candidate structural parameters whose discretization error is not within the tolerance range are used to obtain a fourth candidate structural parameter, including: A genetic algorithm is used to optimize the undiscrete second candidate structure parameters whose errors are outside the tolerance range; If the preset optimization stopping condition is not met, perform selection, crossover and mutation operations on the undiscrete second candidate structural parameters whose error is not within the tolerance range after optimization by the genetic algorithm, and return to the step of optimizing the undiscrete second candidate structural parameters whose error is not within the tolerance range using the genetic algorithm. When the preset optimization stopping condition is reached, the undiscrete second candidate structural parameter whose error after optimization by the genetic algorithm is not within the tolerance range is taken as the fourth candidate structural parameter.

[0014] According to certain embodiments of the first aspect of this application, the acceptance probability is expressed as: ;in, To accept probability, The difference between the new objective function and the original objective function. The annealing temperature decreases with the number of optimization iterations.

[0015] According to certain embodiments of the first aspect of this application, after obtaining the target structural parameters, the target structural parameters are subjected to three-dimensional finite-time domain difference verification.

[0016] According to a second aspect of this application, a fully vertical grating coupler is prepared according to the method for preparing a fully vertical grating coupler as described in the first aspect of this application.

[0017] The above-mentioned scheme has at least the following beneficial effects: it locks the optimized grating unit at three discrete heights, avoiding the continuous etching height caused by the traditional adjoint method, eliminating the need for complex grayscale lithography, and allowing for wafer fabrication using only mature processes, thus effectively improving the yield of large-scale manufacturing; it utilizes the discrete advantages of the algorithm and avoids the curse of high-dimensional optimization, improving computational efficiency and reducing the design cycle; the fabricated fully vertical grating coupler breaks through the back reflection bottleneck caused by second-order diffraction when fully vertically coupled at 90 degrees, achieving a peak coupling efficiency of -2.13 dB for the quasi-TE0 mode and a 1 dB broadband flat response at 91 nm in the 2-μm band, which can improve the performance of the device in broadband optical communication and other scenarios. Attached Figure Description

[0018] The accompanying drawings are used to provide a further understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.

[0019] Figure 1 This is a step-by-step diagram of the fabrication method of a fully vertical grating coupler; Figure 2This is a flowchart of the fabrication method for a fully vertical grating coupler; Figure 3 This is a schematic diagram of the XZ cross-section structure of a fully vertical grating coupler; Figure 4 This is a diagram showing the height distribution of the grating units in a fully vertical grating coupler. Figure 5 This is a broadband coupling efficiency spectrum of a fully vertical grating coupler embodiment. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0021] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, or the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0022] First, the abbreviations and key terms are defined as follows.

[0023] Complementary Metal Oxide Semiconductor (CMOS): The mainstream manufacturing process for integrated circuits. This invention is compatible with standard CMOS foundry processes.

[0024] Single-mode fiber (SMF) is an optical fiber with a small central glass core that can only transmit one mode. The core diameter is typically 9-10 μm.

[0025] Mode field diameter (MFD) is used to characterize the distribution of fundamental mode light in the core region of a single-mode fiber. It is generally defined as 1 / e of the maximum intensity at the axis when the light intensity decreases. 2 The maximum distance between any two points.

[0026] Edge Coupler (EC) is a coupler located at the edge of an optical chip that connects optical fibers to the chip via end-face coupling.

[0027] A grating coupler (GC) is a device that uses the diffraction effect of a grating to couple diffracted light into an optical waveguide for propagation, enabling signal transmission from fiber to chip and from chip to fiber.

[0028] Silicon on Insulator (SOI) refers to the fabrication of silicon semiconductor devices in a layered silicon-insulator-silicon substrate to reduce parasitic capacitance within the device and thus improve performance.

[0029] Transverse electric (TE) refers to a propagation mode in which the longitudinal component of the electric field is zero and the longitudinal component of the magnetic field is not zero in the direction of electromagnetic wave propagation.

[0030] Transverse magnetic (TM) refers to a propagation mode in which the longitudinal component of the magnetic field is zero and the longitudinal component of the electric field is not zero in the direction of electromagnetic wave propagation.

[0031] Inverse design is a design method that uses algorithms to automatically deduce the structural parameters of a device by specifying target performance parameters (such as coupling efficiency and bandwidth).

[0032] To address the technical challenges of existing fully vertical grating couplers in reverse engineering, such as the difficulty in fabricating the continuous etching structure generated by the adjoint method, the performance drop caused by forced discretization, and the extreme time consumption of pure heuristic algorithms for high-dimensional parameters, embodiments of this application provide a fully vertical grating coupler and its fabrication method.

[0033] By combining continuous adjoint optimization with the introduction of smooth projection and dynamic penalty mechanism, and discretization fine-tuning based on parallel genetic algorithm and direct binary search with simulated annealing strategy, the bottleneck of difficulty in balancing algorithm optimization efficiency and physical processing limitations is effectively overcome. This provides an efficient and accurate solution for achieving high-performance fully vertical fiber-chip coupling that is both highly coupled, has large bandwidth, and fully complies with ternary (third-order height) manufacturing process.

[0034] The embodiments of this application will be further described below with reference to the accompanying drawings.

[0035] For fully vertical grating couplers, the device is built entirely on a 220 nm silicon-on-insulator (SOI) platform or a similar silicon-based optoelectronic platform, using only Si and SiO2, without any auxiliary materials.

[0036] Reference Figure 3The physical structure of this fully vertical grating coupler, from bottom to top, includes a silicon substrate, a silicon oxide (BOX) layer, a silicon grating layer, and a silicon cladding layer. The silicon substrate serves as the bottom support of the entire device; the BOX layer limits downward leakage of the optical field; and the cladding layer protects the grating structure and provides refractive index matching with the upper optical fiber. The silicon grating layer is the core region for beam deflection and wavefront modulation, and its total grating length is denoted as L. t It is divided into N equally spaced grating units, and the length of each grating segment in each grating unit is L. g The period length remains constant. At the fiber coupling end, a standard single-mode fiber (SSMF) is suspended vertically directly above the capping layer, with its optical axis strictly parallel to the chip surface normal (Z-axis), thus achieving a 90-degree perfectly vertical coupling.

[0037] To ensure perfect compatibility with existing multilayer mask semiconductor photolithography and etching processes, the grating height H of each segment of each grating unit within the silicon grating layer is... g It abandons the traditional continuous surface morphology and is required to strictly follow three preset physical discrete heights (i.e., the bottom height h of the full etching). low The intermediate height h of shallow etching mid and the maximum height h of zero etch high The values ​​are selected from ) to make the entire grating present a ternary digital form composed of these three step heights.

[0038] The fully vertical grating coupler was fabricated using the following method.

[0039] Reference Figure 1 The fabrication method of a fully vertical grating coupler includes the following steps: Step S100: The initial structural parameters of the grating unit are mapped to three preset physical discrete heights through a projection function to obtain the first candidate structural parameters; Step S200: Calculate the broadband coupling efficiency corresponding to the first candidate structure parameters as the objective function; Step S300: Optimize the first candidate structure parameters according to the convergence, iteration number and dispersion requirements corresponding to the objective function, and determine the second candidate structure parameters from the optimized first candidate structure parameters; Step S400: The second candidate structural parameter whose error is within the tolerance range is snapped to the standard discrete height and locked to obtain the third candidate structural parameter. The undiscrete second candidate structural parameter whose error is not within the tolerance range is discretized to obtain the fourth candidate structural parameter. The third candidate structural parameter and the fourth candidate structural parameter are merged to obtain the fifth candidate structural parameter. Step S500: Optimize by replacing the discrete height of the fifth candidate structural parameter, calculate the objective function of the fifth candidate structural parameter with the replaced discrete height to obtain a new objective function. If the new objective function is less than the objective function, calculate the acceptance probability based on the current annealing temperature, and determine whether to accept the replacement of the discrete height based on the acceptance probability. After optimization, the target structural parameter is obtained. Step S600: A fully vertical grating coupler is prepared according to the target structural parameters.

[0040] Among them, steps S100, S200 and S300 are the first stage - the adjoint method continuous optimization and initial discretization stage; steps S400, S500 and S600 are the second stage - the heuristic global discretization fine-tuning stage.

[0041] Reference Figure 2 For step S100, the initial structural parameters, parameter optimization range, and maximum number of iterations of the grating unit are set.

[0042] The initial structural parameters of the grating unit are normalized to continuous variables ρ∈[0,1].

[0043] The initial structural parameters of the grating unit are mapped to three preset physical discrete heights through a projection function to obtain the first candidate structural parameters. The three preset physical discrete heights include the bottom height of full etching, the middle height of shallow etching, and the maximum height of zero etching.

[0044] Specifically, the projection function is a double-threshold projection function based on a smooth step function.

[0045] The mathematical expression for the projection function is: ; in, ρ is the output of the projection function, β is the continuous variable normalized by the initial structure parameters, β is the parameter controlling the projection steepness, and η is the jump threshold.

[0046] Two thresholds are set: the first threshold is η1 = 1 / 3, and the second threshold is η2 = 2 / 3.

[0047] The actual physical height H of the i-th grating unit is calculated by superposition mapping. i The first candidate structure parameters are obtained.

[0048] The first candidate structure parameter is expressed as: ; in, The actual physical height of the i-th grating unit. The continuous variable is the normalized initial structural parameters of the i-th grating unit. The first transition threshold, The second transition threshold, The bottom height of the full etching. This represents the intermediate height of the shallow etching. The maximum height for zero etching. This is the output of the projection function.

[0049] For step S200, the objective function is calculated using the two-dimensional finite-time difference method, and the broadband coupling efficiency corresponding to the first candidate structural parameters is used as the objective function.

[0050] The objective function introduces a flat-top Gaussian broadband weight W(λ).

[0051] Broadband coupling efficiency is expressed as: ; in, ; Let T(λ) represent the broadband coupling efficiency, where λ is the coupling efficiency at different wavelengths. For flat-top Gaussian broadband weights, λ c λ is the center wavelength, Δλ is the target flat bandwidth, and M is the super-Gaussian order.

[0052] For step S300, the parameters of the first candidate structure are optimized according to the convergence, iteration number, and dispersion requirements corresponding to the objective function, including the following steps: If the current iteration does not converge and the maximum number of iterations has not been reached, perform gradient optimization and update the initial structural parameters, then return to the step of mapping the initial structural parameters of the grating unit to three preset physical discrete heights through the projection function to obtain the first candidate structural parameters and continue to the next iteration; If the current iteration converges or reaches the maximum number of iterations, but the discreteness requirement is not met, return to the step of mapping the initial structural parameters of the grating unit to three preset physical discrete heights through the projection function to obtain the first candidate structural parameters, and increase the parameter controlling the projection steepness of the projection function to add noise perturbation to the initial structural parameters. If the current iteration converges or reaches the maximum number of iterations and meets the discreteness requirement, save the current first candidate structure parameters.

[0053] Then, the second candidate structural parameters are determined from the optimized first candidate structural parameters.

[0054] Specifically, it determines whether the current iteration has converged or reached the maximum number of iterations. If the current iteration has not converged and has not reached the maximum number of iterations, it calculates the gradient using the adjoint method and updates the initial structure parameters using the L-BFGS-B algorithm (Limited-memory Broyden–Fletcher–Goldfarb–Shanno with Bounds, a memory-constrained quasi-Newton method). It then returns to the step of mapping the initial structure parameters of the grating unit to three preset physical discrete heights through a projection function to obtain the first candidate structure parameters and continues to the next iteration.

[0055] If the current iteration converges or the maximum number of iterations is reached, determine whether the current first candidate structure parameter meets the discreteness requirement.

[0056] If the current iteration converges or reaches the maximum number of iterations, but the discreteness requirement is not met, increase the parameter controlling the projection steepness of the projection function, add noise perturbation to the initial structural parameters, and return with the updated steepness and perturbed parameters to the step of mapping the initial structural parameters of the grating unit to three preset physical discrete heights through the projection function to obtain the first candidate structural parameters, and restart the optimization.

[0057] Understandably, in order to prevent parameters from getting stuck near the transition threshold and falling into local optima, structural parameters that are stuck near η1 or η2 are identified and normalized random noise is injected into them to escape the local optimum trap.

[0058] If the current iteration converges or reaches the maximum number of iterations, and the structural parameters meet the discreteness requirements, the first candidate structural parameters after the final projection function mapping are saved as the second candidate structural parameters, thus ending the first stage of continuous optimization.

[0059] For step S400, the good and bad point separation strategy is used to attract the second candidate structural parameter whose error is within the tolerance range to the standard discrete height and lock it to obtain the third candidate structural parameter. The locked third candidate structural parameter is the "good point".

[0060] Undiscrete second candidate structural parameters whose discretization error is outside the tolerance range are extracted separately as "bad points".

[0061] The fourth candidate structure parameter is obtained by taking the undiscreteized second candidate structure parameter whose discretization error is outside the tolerance range, including the following steps: A genetic algorithm is used to optimize the undiscrete second candidate structure parameters whose errors are outside the tolerance range; If the preset optimization stopping condition is not met, perform selection, crossover and mutation operations on the undiscrete second candidate structural parameters whose error is not within the tolerance range after optimization by the genetic algorithm, and return to the step of optimizing the undiscrete second candidate structural parameters whose error is not within the tolerance range using the genetic algorithm. When the preset optimization stopping condition is reached, the undiscrete second candidate structural parameter whose error after optimization by the genetic algorithm is not within the tolerance range is taken as the fourth candidate structural parameter.

[0062] Specifically, a genetic algorithm is used to optimize undiscrete variables. This is achieved by performing selection, crossover, and mutation operations on the undiscrete variables, within a gene pool limited by the target discrete height {h}. low ,h mid ,h high Under the premise of continuous propagation, by judging whether the optimization stopping condition is met, the best strictly discrete combination of the bad points is finally obtained, eliminating the parasitic reflection caused by forced discretization.

[0063] After the bad pixel repair is completed, the discretized variable (i.e., the fourth candidate structural parameter) is used as the new optimization variable. That is, the successfully discretized grating cell height is merged with the previously locked grating cell height to form a set of all variables that fully satisfy the third-order discretization requirements. In other words, the fifth candidate structural parameter is obtained by merging the third and fourth candidate structural parameters.

[0064] Among them, the Genetic Algorithm (GA) is a heuristic global optimization algorithm that simulates the biological evolution process. It iteratively updates candidate solutions through operations such as selection, crossover, and mutation to search for better solutions in a complex parameter space.

[0065] For step S500, the parameters of all fifth candidate structures are discretely optimized using a direct binary search with a simulated annealing strategy.

[0066] The discrete height of the fifth candidate structural parameter is replaced for optimization. The objective function of the fifth candidate structural parameter with the replaced discrete height is calculated to obtain the new objective function. When the new objective function is less than the objective function, the acceptance probability is calculated based on the current annealing temperature. The acceptance probability is used to determine whether to accept the replacement of the discrete height. The target structural parameter is obtained after the optimization is completed.

[0067] Simulated Annealing (SA) is a stochastic optimization algorithm inspired by the solid annealing process. During the optimization process, it accepts candidate solutions with deteriorating performance with a certain probability, thereby helping the algorithm to escape local optima and gradually converge to a better solution as the annealing temperature decreases.

[0068] Direct Binary Search (DBS) is a local search optimization method that involves flipping discrete variables point by point and comparing changes in the objective function. It typically optimizes discrete structures by sequentially changing the values ​​of each design unit while retaining modifications that improve performance.

[0069] Specifically, the discrete height of the fifth candidate structure parameter is randomly traversed and attempted to be replaced one by one. According to the formula... The objective function for calculating the fifth candidate structure parameter with the discrete height is used to obtain the new objective function FOM. new .

[0070] When determining whether the optimization stopping condition has been met, if the new objective function is less than the objective function (i.e., ΔFOM = FOM), then... new FOM old <0), calculate the acceptance probability based on the current annealing temperature T.

[0071] The probability of acceptance is expressed as: ;in, To accept probability, The difference between the new objective function and the original objective function. The annealing temperature decreases with the number of optimization iterations.

[0072] By generating random numbers, with this probability Decide whether to accept the variation in simulation results, thus escaping the local optimum trap. After each round of traversal, reduce the annealing temperature (T=T×cooling_rate) and continue the loop to find the optimal solution.

[0073] When the temperature drops to extremely low levels and the search parameters remain unchanged throughout the entire round, thus reaching the optimization stopping condition, the final optimization result is saved as the target structure parameter.

[0074] After obtaining the target structural parameters, the target structural parameters are verified using three-dimensional finite-time difference.

[0075] For step S600, a fully vertical grating coupler is prepared according to the target structural parameters.

[0076] In one specific embodiment, a fully vertical grating coupler is designed for 2-μm band communication scenarios. The device is based on a standard 220 nm silicon-on-insulator (SOI) platform. The SiO2 thickness of both the capping layer and the buried oxide layer is 3 μm. The silicon grating layer is divided into N = 100 equally spaced grating units, with a fixed period length of 0.10 μm for each unit, resulting in a total grating length Lt of 10 μm. To meet the requirements of ternary single-step or two-step etching fabrication processes, three physical discrete heights are defined as: bottom height h... low = 0 nm (fully etched to buried oxide layer), intermediate height h mid = 150 nm (shallow etching), and maximum height h high = 220 nm (zero etching, i.e., retaining the original silicon thickness of the top layer).

[0077] During operation, incident light from the single-mode fiber perpendicularly illuminates the grating region and undergoes diffraction and mode conversion with the ternary discrete-height grating elements distributed along the grating length. Each grating element, through coordinated modulation of its local equivalent refractive index, coupling strength, and radiation phase, gradually converts the incident light field and couples it into the TE0 mode of the on-chip waveguide for propagation. By using continuous-discrete hybrid optimization to design the overall height of each grating element, comprehensive optimization of coupling directionality, field distribution matching, and reflection suppression can be achieved while meeting the constraints of ternary discrete manufacturing, thereby obtaining high coupling efficiency and a wide operating bandwidth. Unless otherwise specified, the coupling efficiency results given in subsequent embodiments refer to the coupling efficiency of the single-mode fiber fundamental mode coupled to the TE0 mode of the on-chip waveguide.

[0078] Reference Figure 4 In the grating element height distribution diagram, the horizontal axis represents the grating element number (from 1 to 100), and the vertical axis represents the optimized physical height of the grating. According to the grating element height distribution diagram, after a two-stage hybrid optimization process involving continuous optimization using the adjoint method (an efficient numerical method combining forward and adjoint electromagnetic field simulations to calculate the gradient of all parameters in the entire design domain with respect to the objective function) and heuristic discrete fine-tuning, the height of all grating elements is strictly limited to three preset discrete values: 0 nm, 150 nm, and 220 nm. There are no continuous intermediate heights that do not conform to standard fabrication processes.

[0079] Reference Figure 5In the broadband coupling efficiency spectrum, the horizontal axis represents the test wavelength (1900 nm to 2100 nm), and the vertical axis represents the coupling efficiency between the grating coupler and the standard single-mode fiber. According to the test curve of the broadband coupling efficiency spectrum, a peak coupling efficiency of -2.13 dB was achieved at a wavelength of 1979 nm; more importantly, its 1 dB bandwidth is as high as 91 nm. Due to the introduction of a flat-top Gaussian broadband weight into the objective function during the continuous optimization phase, the device's efficiency spectrum exhibits excellent "flat-top" characteristics at the top (i.e., maintaining a flat response with high efficiency over an extremely wide wavelength range). This not only completely solves the narrowband resonance problem easily caused by the adjoint method but also successfully overcomes the bottleneck of high reflection and low efficiency caused by second-order diffraction when coupled at a perfectly perpendicular 90-degree angle.

[0080] By combining the grating unit height distribution map and the broadband coupling efficiency spectrum, this method for fabricating a fully vertical grating coupler can efficiently design and fabricate a high-performance fully vertical grating coupler with fabrication feasibility (strictly ternary discrete height), high coupling efficiency (peak -2.13 dB), and large operating bandwidth (1 dB bandwidth 91 nm) in a vast high-dimensional parameter space, which has high practical engineering application value.

[0081] This fully vertical grating coupler fabrication method combines the high efficiency of gradient optimization using the adjoint method with the strict discrete advantages of heuristic algorithms (genetic algorithm + simulated annealing). By employing a "good and bad point separation" strategy, it avoids the high-dimensional optimization curse of purely heuristic algorithms, improving computational efficiency and reducing design cycle time. This ensures that the optimized grating cells are 100% strictly locked at three discrete heights (e.g., 0 nm, 150 nm, 220 nm), avoiding the continuous etching heights produced by traditional adjoint methods. It eliminates the need for complex grayscale lithography, relying solely on a mature "double etching" process for fabrication, effectively improving the yield of large-scale manufacturing. It overcomes the back reflection bottleneck caused by second-order diffraction in 90-degree fully vertical coupling, achieving a peak coupling efficiency of -2.13 dB in the quasi-TE0 mode and a 1 dB broadband flat response at 91 nm in the 2-μm band, enhancing the device's performance in broadband optical communication and other scenarios.

[0082] The above is a detailed description of the preferred embodiments of this application, but this application is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for fabricating a fully vertical grating coupler, characterized in that, The fully vertical grating coupler comprises, from bottom to top, a silicon substrate layer, a silicon buried oxide layer, a silicon grating layer, and a silicon top cover layer, wherein the silicon grating layer includes multiple grating units; The method includes: The initial structural parameters of the grating unit are mapped to three preset physical discrete heights through a projection function to obtain the first candidate structural parameters. The three preset physical discrete heights include the bottom height of full etching, the middle height of shallow etching, and the maximum height of zero etching. The broadband coupling efficiency corresponding to the first candidate structure parameters is calculated as the objective function. The first candidate structure parameters are optimized according to the convergence, iteration number and dispersion requirements corresponding to the objective function, and the second candidate structure parameters are determined from the optimized first candidate structure parameters. The second candidate structural parameter whose error is within the tolerance range is snapped to the standard discrete height and locked to obtain the third candidate structural parameter. The undiscrete second candidate structural parameter whose error is not within the tolerance range is discretized to obtain the fourth candidate structural parameter. The third candidate structural parameter and the fourth candidate structural parameter are merged to obtain the fifth candidate structural parameter. The discrete height of the fifth candidate structural parameter is replaced for optimization. The objective function of the fifth candidate structural parameter with the replaced discrete height is calculated to obtain a new objective function. When the new objective function is less than the objective function, the acceptance probability is calculated based on the current annealing temperature. The acceptance probability is used to determine whether to accept the replacement of the discrete height. The target structural parameter is obtained after the optimization is completed. A fully vertical grating coupler was fabricated based on the target structural parameters.

2. The method of claim 1, wherein the method further comprises: The projection function is: ; where, is the output of the projection function, p is a continuous variable normalized by the initial structure parameters, β is a parameter that controls the steepness of the projection, and η is a jump threshold.

3. The method for fabricating a fully vertical grating coupler according to claim 1, characterized in that, The first candidate structure parameter is expressed as follows: ； in, The actual physical height of the i-th grating unit. The continuous variable is the normalized initial structural parameters of the i-th grating unit. The first transition threshold, The second transition threshold, The bottom height of the full etching. This represents the intermediate height of the shallow etching. The maximum height for zero etching. This is the output of the projection function.

4. The method of claim 1, wherein the method further comprises: The broadband coupling efficiency is expressed as: ； wherein ; T(λ) is the coupling efficiency at different wavelengths, λ is the wavelength of light, T(λ) is the coupling efficiency at different wavelengths, λ is the wavelength of light, c λ0 is the center wavelength, Δλ is the target flat bandwidth, M is the super-Gaussian order.

5. The method of claim 1, wherein the method further comprises: The optimization of the first candidate structure parameters based on the convergence, iteration number, and dispersion requirements corresponding to the objective function includes: If the current iteration does not converge and the maximum number of iterations has not been reached, perform gradient optimization and update the initial structural parameters, then return to the step of mapping the initial structural parameters of the grating unit to three preset physical discrete heights through a projection function to obtain the first candidate structural parameters and continue to the next iteration; If the current iteration converges or reaches the maximum number of iterations, but the discreteness requirement is not met, return to the step of mapping the initial structural parameters of the grating unit to three preset physical discrete heights through the projection function to obtain the first candidate structural parameters, and increase the parameter controlling the projection steepness of the projection function to add noise perturbation to the initial structural parameters. If the current iteration converges or reaches the maximum number of iterations and meets the discreteness requirement, save the current first candidate structure parameters.

6. The method of claim 5, wherein the method further comprises: The addition of noise perturbation to the initial structural parameters includes: Identify initial structural parameters that are stuck near a transition threshold, and add normally distributed random noise to these parameters.

7. The method of claim 1, wherein the method further comprises: The undiscreteized second candidate structural parameters whose discretization error is outside the tolerance range are used to obtain the fourth candidate structural parameters, including: A genetic algorithm is used to optimize the undiscrete second candidate structure parameters whose errors are outside the tolerance range; If the preset optimization stopping condition is not met, perform selection, crossover and mutation operations on the undiscrete second candidate structural parameters whose error is not within the tolerance range after optimization by the genetic algorithm, and return to the step of optimizing the undiscrete second candidate structural parameters whose error is not within the tolerance range using the genetic algorithm. When the preset optimization stopping condition is reached, the undiscrete second candidate structural parameter whose error after optimization by the genetic algorithm is not within the tolerance range is taken as the fourth candidate structural parameter.

8. The method of claim 1, wherein the method further comprises: The acceptance probability is expressed as: ;in, To accept probability, The difference between the new objective function and the original objective function. The annealing temperature decreases with the number of optimization iterations.

9. The method of claim 1, wherein the method further comprises: After obtaining the target structural parameters, the target structural parameters are verified using three-dimensional finite-time domain difference.

10. A fully vertical grating coupler, characterized by, It was prepared according to the method for preparing a fully vertical grating coupler as described in any one of claims 1 to 9.

Images