A method for precisely controlling oxidation channels of a VCSEL chip

Abstract

The application discloses a kind of VCSEL chip oxidation channel's accurate control method, including calculating the shrinkage stress distribution caused by material volume shrinkage in oxidation process, obtain its distribution parameter in different depth regions of oxidation channel;Simulate the influence of shrinkage stress on the shape and position of oxidation channel, get the predicted deformation variable and position offset of channel;With reducing predicted deformation variable and position offset as target, generate a set of optimized oxidation process parameters;According to position offset and shrinkage stress distribution, generate compensated oxidation path;Using optimized oxidation process parameters and compensated oxidation path to execute oxidation process, and in real time monitoring physical quantity reflecting stress state in oxidation process, dynamically adjust oxidation process parameters.The application can realize high-precision, high-stability control to channel shape and position by actively predicting stress distribution and feeding forward compensation, combined with real-time feedback regulation of oxidation process.

Description

Technical Field

[0001] This invention relates to the field of VCSEL chip fabrication technology, and in particular to a method for precise control of oxide channels in VCSEL chips. Background Technology

[0002] In the manufacturing of vertical-cavity surface-emitting lasers (VCSELs), precise control of oxide channels is a core technology determining the optoelectronic performance and reliability of the device. With the continuous expansion of VCSEL applications in high-speed optical communication, 3D sensing, and lidar, increasingly stringent requirements are being placed on the aperture consistency, positional accuracy, and sidewall morphology of oxide channels. However, existing technologies have long faced a fundamental challenge: during wet oxidation, the oxidation reaction of the aluminum gallium arsenide (AlGaAr) material components is accompanied by significant volume shrinkage. The resulting internal stress continuously acts on the unreacted semiconductor material, leading to uneven oxidation front advance speed, channel shape distortion, and even positional displacement. This uncontrollable deformation caused by stress accumulation has become a bottleneck restricting the performance improvement and yield control of VCSEL chips.

[0003] Currently, the industry mainly addresses this issue through empirical control of macroscopic parameters such as oxidation time and temperature, or through screening via electrical testing after oxidation. However, these methods have significant limitations: firstly, they cannot predict stress distribution and its impact before process execution, essentially remaining a trial-and-error approach; secondly, they treat the oxidation process as a static operation, ignoring the dynamic evolution of stress during oxidation; and thirdly, they lack the ability to perceive and compensate for stress states during oxidation in real time, making it difficult to guarantee channel uniformity across different batches or even within the same wafer. Summary of the Invention

[0004] Therefore, the technical problem to be solved by the present invention is to overcome the difficulty in controlling the deformation and positional displacement of the oxidation channels of VCSEL chips caused by uneven material oxidation shrinkage stress in the prior art, and to provide a precise control method for the oxidation channels of VCSEL chips, which can achieve high precision and high stability control of the channel shape and position by actively predicting stress distribution and performing feedforward compensation, combined with real-time feedback adjustment of the oxidation process.

[0005] To address the aforementioned technical problems, this invention provides a method for precise control of oxide channels in VCSEL chips, comprising the following steps:

[0006] Based on the material property data of the material to be oxidized, the distribution of shrinkage stress caused by the volume shrinkage of the material during the oxidation process is calculated, and its distribution parameters in different depth regions of the oxidation channels are obtained.

[0007] Based on the distributed parameter simulation, the influence of shrinkage stress on the shape and position of the oxidation channel is obtained, and the predicted deformation and position offset of the channel are obtained.

[0008] With the goal of reducing the predicted deformation and position offset, the oxidation process parameters are optimized to generate a set of optimized oxidation process parameters;

[0009] The initial oxidation path is corrected based on the position offset and shrinkage stress distribution to generate a compensated oxidation path;

[0010] The oxidation process is executed using optimized oxidation process parameters and a compensated oxidation path, and physical quantities reflecting stress state are monitored in real time during the oxidation process. The oxidation process parameters are dynamically adjusted according to the fluctuation trend of the monitoring data to stabilize the formation process of oxidation channels.

[0011] In one embodiment of the present invention, based on the material property data of the material to be oxidized, the distribution of shrinkage stress caused by the volume shrinkage of the material during the oxidation process is calculated, including:

[0012] Establish a graph showing the relationship between the oxidation process of materials and the volume shrinkage rate;

[0013] Based on the correspondence diagram and the three-dimensional structure of the target oxidation channels, the oxidation process is discretized into multiple continuous time stages;

[0014] For each time stage, the strain field caused by volume shrinkage is determined based on the current oxidation state of the material.

[0015] Based on the strain field and the material's ability to resist deformation under different strain states, the shrinkage stress distribution parameters corresponding to different depths of the oxide channels were analyzed.

[0016] In one embodiment of the present invention, if discontinuous jumps in stress parameters at different depths are detected during the analysis of shrinkage stress distribution parameters, the following steps are performed for smoothing:

[0017] A virtual stress buffer layer is constructed between adjacent depth regions, and the stress parameter value of the stress buffer layer is determined by weighted fusion of the parameter values ​​of its upper and lower adjacent regions.

[0018] Using a stress buffer layer as an intermediary, the overall stress distribution of the oxide pores from the surface to the bottom is recalculated to make the stress parameters transition smoothly along the depth direction.

[0019] The smoothed stress distribution parameters are used for subsequent prediction of the shape and location of the duct.

[0020] In one embodiment of the present invention, the influence of shrinkage stress on the shape and position of oxide channels is simulated based on distributed parameters to obtain the predicted deformation and position offset of the channels, including:

[0021] The shrinkage stress distribution parameters are mapped to equivalent loads acting on the boundary of the initial channel geometry model.

[0022] Under equivalent load, the structural response of the initial duct geometric model is calculated, and the displacement vectors of each point on the duct sidewall are identified.

[0023] Based on the displacement vector, the predicted three-dimensional shape of the channel is reconstructed and compared with the initial geometric model to quantify the overall deformation of the channel.

[0024] Based on the reconstructed predicted 3D topography, the deviation trajectory of its geometric central axis relative to the initial design position is calculated, thereby determining the position offset of the channel.

[0025] In one embodiment of the present invention, when calculating the structural response of the initial duct geometry model, if a drastic change in the displacement vector field is detected in space, causing distortion of the predicted topography, then a harmonization process is performed:

[0026] Virtual structural constraint links are introduced between adjacent regions where the displacement vector changes drastically in order to coordinate the inconsistency of their displacements.

[0027] Under the coordination of virtual structural constraint links, the displacement vectors of each point on the sidewall of the duct are recalculated to obtain a smoothed displacement field;

[0028] Based on the smoothed displacement field, the reconstruction and comparison of the predicted three-dimensional shape of the channel are re-executed to obtain deformation and position offset that are more consistent with physical reality.

[0029] In one embodiment of the present invention, with the goal of reducing predicted deformation and positional offset, the oxidation process parameters are optimized to generate a set of optimized oxidation process parameters, including:

[0030] Establish a mapping table relating oxidation process parameters to the predicted deformation and position offset;

[0031] Based on the association mapping table, the upper limits of allowable deformation and position offset are set as optimization boundaries;

[0032] Within the optimization boundary, multiple sets of alternative oxidation process parameter combinations are systematically generated;

[0033] For each set of alternative parameter combinations, the corresponding oxidation process is predicted, and the new stress distribution and the resulting new deformation and new position offset under this oxidation process are deduced.

[0034] From all the alternative combinations, select the set of parameters that yields the best combined result of the new deformation and the new position offset, and use it as the optimized oxidation process parameters.

[0035] In one embodiment of the present invention, the initial oxidation path is corrected based on the position offset and shrinkage stress distribution to generate a compensated oxidation path, including:

[0036] The position offset is decomposed into an offset vector in a plane perpendicular to the oxidation propulsion direction;

[0037] Analyze the spatial gradient of the shrinkage stress distribution to identify the dominant stress regions and their directions that cause positional shifts;

[0038] Based on the offset vector and the direction of the dominant stress region, a reverse path correction amount with opposite direction and related magnitude is applied at the corresponding spatial position of the initial oxidation path;

[0039] The reverse path correction is smoothly integrated into the entire initial oxidation path, forming a compensated post-oxidation path that transitions continuously from the starting point to the end point.

[0040] In one embodiment of the present invention, dynamically adjusting oxidation process parameters based on the fluctuation trend of monitoring data includes:

[0041] The real-time monitored physical quantity data is filtered to extract its trend components that change over time;

[0042] The trend component is compared with the preset stress state safety range to determine whether the current oxidation process deviates from the stable region.

[0043] When a deviation is detected, fine-tuning instructions for one or more oxidation process parameters are generated based on the direction and magnitude of the deviation.

[0044] Fine-tuning instructions are translated into executable parameter settings for process equipment, enabling online intervention in the ongoing oxidation process and bringing the monitored physical quantity data back to a safe range.

[0045] In one embodiment of the present invention, fine-tuning instructions for one or more oxidation process parameters are generated based on the direction and magnitude of the deviation, including:

[0046] If the monitored stress state trend component shows a continuous increase and deviates from the upper limit of the safe range, a fine-tuning instruction is generated to reduce the oxidation reaction rate. The fine-tuning instruction is to reduce the ambient humidity inside the oxidation furnace.

[0047] If the monitored stress state trend component shows a continuous decrease and deviates from the lower limit of the safe range, a fine-tuning instruction is generated to improve the uniformity of the oxidation reaction. The fine-tuning instruction is to increase the local airflow circulation intensity of the oxidation environment.

[0048] If the monitored stress state trend component shows violent oscillations at a high level, a fine-tuning instruction is generated to stabilize the reaction interface. The fine-tuning instruction is to slightly reduce the ambient temperature and maintain the current humidity.

[0049] In one embodiment of the present invention, when generating a fine-tuning instruction based on a deviation, if it is determined that multiple process parameters all need to be adjusted and have mutual coupling effects, coordinated control is performed, including:

[0050] Establish the correlation and sensitivity ranking among key oxidation process parameters;

[0051] Based on sensitivity ranking, prioritize adjusting the single dominant process parameter that is most sensitive to the current stress fluctuation;

[0052] After adjusting the dominant parameters, the trend of change in the monitoring data was reassessed.

[0053] If the stress state does not fully return to the safe range, then based on the reassessed state, the adjustment of the next minor parameter is initiated according to sensitivity, and so on, until the process returns to stability.

[0054] The technical solution of the present invention has the following advantages compared with the prior art:

[0055] The precise control method for oxide channels in VCSEL chips described in this invention significantly reduces the systematic deformation of oxide channels through stress simulation and path compensation in the early stage; it effectively suppresses the influence of random factors on channel uniformity through process monitoring and dynamic adjustment; and finally, while improving the performance consistency and yield of VCSEL chips, it reduces the time and material costs required by traditional trial and error methods, providing a reliable technical path for the large-scale manufacturing of high-precision VCSELs. Attached Figure Description

[0056] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein:

[0057] Figure 1 This is a flowchart of the steps of the method for precise control of the oxide channels of VCSEL chips according to the present invention;

[0058] Figure 2 This is a flowchart of the steps of the accurate stress prediction scheme of the present invention;

[0059] Figure 3 This is a flowchart of the deformation prediction technology solution of the present invention;

[0060] Figure 4 This is a flowchart of the steps of the multi-objective parameter optimization technique of the present invention;

[0061] Figure 5 This is a flowchart of the steps of the intelligent correction technology for oxidation pathways based on the concept of reverse compensation;

[0062] Figure 6 This is a flowchart of the steps of the closed-loop feedback control technology solution based on real-time data drive of the present invention;

[0063] Figure 7 This is a flowchart of the steps of generating fine-tuning instructions for one or more oxidation process parameters according to the present invention;

[0064] Figure 8 This is a flowchart of the steps of the sequence decoupling control strategy based on sensitivity ranking of the present invention. Detailed Implementation

[0065] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0066] Reference Figure 1 As shown, this invention proposes a precise control method for the oxidation channels of VCSEL chips, transforming the traditional passive response into active prediction and closed-loop control. Specifically, the method first calculates the distribution of shrinkage stress caused by the volume shrinkage of the material during the oxidation process based on the material property data of the material to be oxidized, and obtains its distribution parameters in different depth regions of the oxidation channels. This step solves the problem at its source, quantitatively and precisely revealing the source and three-dimensional spatial distribution of stress before the process is executed, providing a precise data foundation for all subsequent compensation measures.

[0067] Next, based on the aforementioned distribution parameters, the influence of shrinkage stress on the shape and position of the oxidation channels was simulated, yielding the predicted deformation and positional offset of the channels. This is equivalent to completing a process rehearsal in virtual space, providing advance insight into the possible deformation and offset of the channels under existing parameters, thus exposing problems before they actually occur.

[0068] Based on accurate predictions, proactive intervention is implemented. Aiming to reduce predicted deformation and positional offset, the oxidation process parameters are optimized, generating a set of optimized oxidation process parameters. Simultaneously, the initial oxidation path is corrected based on the positional offset and shrinkage stress distribution, generating a compensated oxidation path. These two steps constitute a complete feedforward compensation system, acting like a smart navigation path that automatically avoids stress traps during the oxidation process, preemptively mitigating the potential adverse effects of stress from both parameter and path perspectives.

[0069] Finally, the oxidation process is executed using optimized oxidation process parameters and a compensated oxidation path. During the process, physical quantities reflecting the stress state are monitored in real time, and the oxidation process parameters are dynamically adjusted based on the fluctuation trends of the monitored data. This step introduces real-time feedback control, forming a complete closed loop. It can effectively cope with unpredictable micro-fluctuations in actual production environments, ensuring that the entire oxidation process always proceeds in the direction of forming stable and precise channels.

[0070] In summary, the precise control method for oxide channels in VCSEL chips of this invention significantly improves the shape fidelity of oxide channels by predicting and compensating for stress at the source, making them closer to the ideal circle, thereby directly improving the light output quality and mode stability of VCSELs. Secondly, by precisely correcting the positional offset, the alignment accuracy of the channels with other functional structures is effectively guaranteed, improving the consistency and performance ceiling of the device. Finally, the prediction-compensation-monitoring closed-loop system constructed by the entire scheme greatly enhances the robustness to process fluctuations and material differences, which not only improves manufacturing yield but also provides a stable and reliable process platform for developing higher-performance VCSEL chips.

[0071] As mentioned earlier, the uneven volume shrinkage of materials during oxidation is the root cause of pore deformation and displacement. However, existing technologies often lack in-depth consideration of the dynamic evolution characteristics of the oxidation process when addressing this issue. Specifically, existing methods usually treat the oxidation process as a whole or only consider the final state, ignoring the physical fact that oxidation progresses gradually from the surface to the interior of the pore sidewall over time. This static or endpoint-oriented analytical perspective makes it impossible to accurately capture the differences in material shrinkage behavior at different oxidation times and depths, resulting in significant deviations in the prediction of internal stress. Consequently, subsequent compensation and control measures are unlikely to achieve the desired results.

[0072] To solve this problem, refer to Figure 2 As shown, this embodiment provides a layer-by-layer, spatiotemporally discretized, precise stress prediction scheme. It decomposes the continuous and complex oxidation physical process into a series of calculable and analyzable discrete stages, thereby achieving refined capture of the spatiotemporal evolution of contraction stress. Specifically, this scheme employs the following technical means:

[0073] First, through preliminary experiments or theoretical calculations, a correlation graph between the material oxidation process and the volume shrinkage rate is established. This graph serves as the foundational database for all subsequent calculations, quantitatively describing how much the material shrinks per unit volume as the degree of oxidation increases. For example, the correlation between the oxidation state and the shrinkage rate can be indirectly determined by measuring the changes in refractive index or density of the material at different oxidation times / depths.

[0074] Next, based on the aforementioned relationship diagram and the three-dimensional structure of the target oxidation channels, the oxidation process is discretized into multiple continuous time stages. The entire oxidation time can be divided into N equally or unequally spaced time periods (t1, t2, ..., tn) according to the total depth of the target channels and the oxidation rate. Each time period corresponds to a small depth into which the oxidation reaction front advances into the material. Through this discretization, a complex dynamic continuous problem is transformed into a series of quasi-static problems that occur sequentially on a static geometric structure and can be analyzed independently.

[0075] Then, for each time stage, based on the current oxidation state of the material, the strain field caused by volumetric shrinkage in that stage is determined. When calculating the i-th time stage, we assume that the newly oxidized thin layer of material undergoes complete volumetric shrinkage (the amount of shrinkage is obtained from the "relationship diagram"), while the dimensions of the previously oxidized layers remain unchanged. This newly added shrinkage strain will induce new displacement and deformation in the formed oxide layer structure. The strain field induced around the entire pore structure at the current time stage can be determined through physical and mechanical calculations.

[0076] Finally, based on the strain field and the material's resistance to deformation under different strain states (i.e., the material's constitutive relation), the shrinkage stress distribution parameters corresponding to different depths in the oxide pores are analyzed. The material's resistance to deformation is an inherent mechanical property, which can be expressed as a functional relationship between stress and strain (such as elastic modulus, plasticity model, etc.). Substituting the strain field calculated in the previous step into this constitutive relation, the stress field corresponding to that strain can be directly calculated. Finally, the stress values ​​at different locations along the depth direction on the sidewall of the oxide pores are extracted as the shrinkage stress distribution parameters for that time period.

[0077] In summary, the technical solution defined in this embodiment effectively solves the problem that existing technologies cannot accurately quantify the spatiotemporal evolution of internal stress during oxidation through an innovative, discretized dynamic stress analysis method, laying a solid and reliable foundation for the final realization of precise control of the oxidation channels of VCSEL chips.

[0078] Furthermore, during the implementation of the refined dynamic stress prediction scheme, a new technical problem was discovered in the calculation results: between adjacent depth regions, the analyzed stress distribution parameters may exhibit discontinuous and drastic jumps. These jumps are often physically unrealistic, usually not stemming from actual physical processes, but rather from numerical noise or spurious oscillations introduced by limitations in mathematical model discretization, material parameter settings, or the numerical calculation itself. Directly using such abrupt stress data for subsequent deformation simulations will result in unnatural jaggedness or distortion in the predicted duct profile, causing process optimization and path compensation based on this data to deviate from the correct direction, ultimately affecting the overall reliability and stability of the control method.

[0079] To address the issue of non-physical jumps in stress data, this solution, based on the aforementioned embodiments, introduces a virtual buffer layer as an intermediary to physically re-rationalize the theoretically discrete calculation results, ensuring the continuity of stress distribution and thus improving the accuracy of subsequent predictions. Specifically, this solution employs the following technical means:

[0080] The system first automatically identifies whether there are discontinuous jumps in stress parameters at different depths. This can be achieved by setting a threshold for the rate of change of stress between adjacent regions. For example, when the difference between the stress value of a certain depth region and the stress values ​​of its immediate upper and lower adjacent regions exceeds this threshold, it is determined that there is a discontinuous jump point at that location.

[0081] Once a jump is identified—that is, between two adjacent depth regions where the jump occurs—a virtual stress buffer layer is constructed. This buffer layer does not physically represent a real material layer, but rather serves as a mathematical intermediary for data smoothing. The stress parameter value of this buffer layer is determined by a weighted fusion of the parameter values ​​of its upper and lower adjacent regions. For example, a linear weighted average method can be used: the stress value of the buffer layer = (stress value of the upper region × weight A + stress value of the lower region × weight B) / (weight A + weight B). The weights can be adaptively adjusted based on the severity of the jump or the distance between regions.

[0082] After introducing a virtual buffer layer, the original discrete model was updated. Using this stress buffer layer as an intermediary, the system recalculated the overall stress distribution of the oxide pores from the surface to the bottom. This process forces the stress parameters to transition smoothly along the depth direction when passing through this region, thereby eliminating unreasonable numerical jumps.

[0083] Finally, the smoothed stress distribution parameters are used for subsequent predictions of the duct shape and location. This means that the input data upon which all subsequent steps rely is more continuous and stable stress information that has been physically validated and optimized.

[0084] In summary, through further optimization and supplementation, the introduction of a virtual buffer layer and stress smoothing processing cleverly solves the common problem of pseudo-oscillation in high-precision numerical calculations, ensuring the high quality and high reliability of the input data of the entire control process, which is an important guarantee for achieving the final precise control goal.

[0085] After obtaining high-precision shrinkage stress distribution parameters, it is necessary to achieve a precise mapping from stress to field and then to morphology, referring to... Figure 3 As shown, this embodiment constructs a complete physics-based deformation prediction technology solution, which equates internal stress to external load and intuitively reproduces the deformation process of the duct by simulating the mechanical response of the structure. The solution includes the following technical solutions:

[0086] First, the shrinkage stress distribution parameters are mapped as equivalent loads acting on the boundaries of the initial duct geometry model. The self-balancing stresses generated by volume shrinkage within the duct sidewalls and surrounding material are equivalently considered as a series of external pressures or surface forces applied to the boundaries of the initial perfect duct model. This equivalent treatment transforms the complex problem of intrinsic material stress into a standard structural mechanics problem, laying the foundation for subsequent analysis.

[0087] Under the equivalent load, the structural response of the initial duct geometric model is calculated, and the displacement vectors of each point on the duct sidewall are identified. In this step, the system solves the equilibrium equations of the duct structure under the equivalent load system. The result is not stress, but the movement of each calculated node on the duct sidewall due to the force, i.e., the displacement vector. This displacement vector includes not only the distance (magnitude) of the node movement, but also the direction of movement, thus fully describing the trend of structural deformation.

[0088] Based on the displacement vectors, the predicted 3D shape of the channel is reconstructed and compared with the initial geometric model to quantify the overall deformation of the channel. Applying the displacement vectors of all nodes to the initial channel model yields a new predicted 3D shape after deformation. This predicted shape is then compared with the initial design model in 3D, for example, calculating the change in channel diameter, the change in ellipticity, or the deviation value of the sidewall profile, thereby quantifying one or more deformation parameters (such as the dimension at the point of maximum deformation, the average deformation rate, etc.).

[0089] Based on the reconstructed predicted 3D topography, the deviation trajectory of its geometric central axis relative to the initial design position is calculated to determine the positional offset of the channel. This step separates shape changes from positional changes, fitting the geometric center lines (central axes) of the initial channel and the predicted topography separately. By calculating the relative positions of these two center lines in 3D space, a trajectory describing how the channel as a whole offsets in space can be obtained, and finally, positional offsets such as the end face center offset distance and offset direction angle are extracted.

[0090] When performing structural response calculations, due to numerical discretization, simplified boundary conditions, or the sensitivity of material models in certain local regions, the calculated displacement vector field may occasionally exhibit drastic spatial changes, such as adjacent nodes having opposite displacement directions or vastly different magnitudes. Such dramatic changes can lead to distortions in the reconstructed predicted 3D topography, such as physically impossible wrinkles, spikes, or discontinuities. These distortions are artifacts in numerical calculations, not actual physical deformations. Directly using these distorted topography for subsequent quantization and compensation will amplify the numerical errors, causing the entire control scheme to be based on incorrect data, ultimately misleading the direction of process optimization and severely impacting the reliability and practicality of the method.

[0091] To address the prediction distortion caused by drastic changes in the displacement field, this embodiment further constructs a displacement field coordination processing technology. By introducing virtual mechanical constraints, it forces incompatible local displacements to satisfy physical continuity and coordination, thereby obtaining a smoother and more realistic deformation field. This includes the following technical means:

[0092] After generating the initial displacement vector field, drastic changes in the field are automatically detected. For example, by calculating the difference or curl of the displacement vectors of adjacent nodes, if the value exceeds a preset reasonableness threshold, it is determined that there is inconsistency in displacement in that region. To address this issue, the solution introduces virtual structural constraint links between adjacent regions experiencing drastic changes. These links are not real structures, but mathematical constraints added to the computational model. Their function is to force the displacement behavior of these inconsistent regions to be interconnected and tend towards consistency, thereby reconciling their displacement inconsistencies.

[0093] After introducing virtual structural constraint links, the original calculation model was updated, and the displacement vectors of each point on the duct sidewall were recalculated under this coordination mechanism. Due to the existence of virtual constraints, the unreasonable, isolated, and drastic displacement changes were suppressed, and the displacement field was adjusted and redistributed under the action of constraints, ultimately resulting in a smoothed displacement field that is continuous and smoothly transitioned in space.

[0094] Finally, based on the smoothed displacement field, the reconstruction and comparison of the predicted three-dimensional morphology of the duct are re-executed. Since the numerical distortion of the input data has been removed, the reconstructed three-dimensional morphology can better reflect the continuous overall deformation trend of the duct under macroscopic stress. The deformation and positional offset obtained in this way have higher reliability and physical authenticity because they are based on a more physically reasonable deformation field.

[0095] Reference Figure 4As shown, after accurately predicting the deformation and positional offset of the channel, it is also necessary to reverse-engineer the optimal oxidation process parameters. To overcome the blindness and inefficiency of existing optimization methods, this embodiment constructs a systematic, data-driven multi-objective parameter optimization technology solution, establishes a clear parameter-objective relationship, and performs global search and evaluation within a preset boundary, thereby scientifically deciding on the optimal process formulation, including the following technical solutions:

[0096] First, a mapping table is established between oxidation process parameters and the predicted deformation and position offset. This mapping table is a core database constructed based on physical models, historical experimental data, or high-fidelity numerical simulations. It quantitatively records how different combinations of process parameters (inputs) affect the final deformation and position offset (output), clarifying the input-output relationship in the optimization process.

[0097] Based on the association mapping table, an upper limit is set for the allowable deformation and positional offset as the optimization boundary. This transforms product performance requirements into mathematical constraints for optimization problems. For example, according to chip design specifications, the variation of the aperture diameter must not exceed a certain value, and the center offset must not exceed another value. All searches will be conducted within the qualified solution space defined by these boundaries.

[0098] Within the optimization boundary, multiple sets of alternative oxidation process parameter combinations are systematically generated. This replaces random trial and error and allows for the use of experimental design methods such as full factorial design, partial factorial design, or Latin hypercube sampling to ensure that sampling points are distributed uniformly and efficiently in the multidimensional parameter space, fully covering the possible optimization region.

[0099] For each set of candidate parameters, the corresponding oxidation process is predicted, and the new stress distribution and the resulting new deformation and positional offset are deduced under this oxidation process. For each set of candidate parameters, the complete virtual simulation process from oxidation process to stress prediction and then to deformation simulation is re-executed to predict the new deformation and positional offset that can be achieved if this set of parameters is used.

[0100] From all candidate combinations, the set of parameters that optimizes the combined result of the new deformation and the new positional offset is selected as the optimized oxidation process parameters. The prediction results corresponding to all candidate schemes are evaluated, a comprehensive evaluation function is defined (e.g., a weighted summation of deformation and offset), and the set of process parameters that makes the comprehensive index reach its optimal value (e.g., minimum value) is selected as the final optimized scheme.

[0101] In summary, this embodiment, through a complete and systematic optimization process, successfully solves the problems of inefficiency and blindness in the optimization of multi-parameter and multi-objective processes in existing technologies, and provides a key technical bridge for transforming accurate theoretical predictions into executable high-performance process solutions.

[0102] In the manufacturing of VCSEL chips, the planning of the oxidation path directly determines the final shape and position of the oxidation channels. Existing technologies typically employ symmetrical oxidation paths based on an ideal geometric center. However, when there is uneven shrinkage stress within the material, this symmetrical path will be subjected to anisotropic stress, causing the actual formed channels to shift in position. Existing methods lack prior awareness of this shift and an active compensation mechanism, and can only passively accept the result of the shift or perform coarse correction by adjusting the overall process parameters. It is difficult to achieve precise cancellation of the shift in a specific direction within the surface, which has become a key bottleneck limiting further improvement in channel position accuracy.

[0103] To overcome the challenge of positional misalignment and achieve precise alignment of the oxide channels, reference was made. Figure 5 As shown, this embodiment further constructs an intelligent correction technology scheme for oxidation paths based on the reverse compensation concept. By analyzing the causal relationship between offset and stress, the initial path is pre-corrected with clear direction and magnitude correlation, thereby enabling the channel to eventually return to the design position under stress. This scheme adopts the following technical means:

[0104] The positional offset is decomposed into an offset vector in a plane perpendicular to the oxidation propulsion direction. This step transforms the final offset result into a physical quantity with a definite direction and magnitude. For example, if the predicted channel is offset by Δx in the positive X-axis direction and Δy in the negative Y-axis direction, the offset vector can be represented as (Δx, -Δy), which provides a precise target for subsequent reverse compensation.

[0105] By analyzing the spatial gradient of the shrinkage stress distribution, the dominant stress regions and their directions that cause the positional shift are identified, and the rate of change of stress in space (gradient) is analyzed. Through this analysis, it is possible to pinpoint which region(s) of stress asymmetry dominates the overall displacement of the channel and to clarify the dominant direction driving the channel movement.

[0106] Based on the offset vector and the direction of the dominant stress region, a reverse path correction, with the opposite direction and related magnitude, is applied at the corresponding spatial position of the initial oxidation path. The principle is negative feedback control: to counteract a positive offset, a reverse correction is applied. Specifically, at the initial path position corresponding to the identified dominant stress region, a compensation movement is planned that is opposite to the offset vector direction and related to the dominant stress direction. Its magnitude is related to the predicted offset and stress gradient amplitude to ensure sufficient compensation.

[0107] The reverse path correction is smoothly integrated into the entire initial oxidation path, forming a compensated post-oxidation path with a continuous transition from the start point to the end point. Local, discrete correction points may cause path discontinuities and introduce new problems. Therefore, the applied correction needs to be integrated into the entire oxidation path through a smoothing algorithm (such as spline interpolation) to ensure that the generated compensated post-oxidation path is a continuous and smooth trajectory that can be stably executed by the oxidation equipment.

[0108] In actual production, process fluctuations may still occur due to factors such as minor changes in equipment condition, environmental disturbances, or slight differences between batches of materials. To further achieve overall stability and precision in the oxidation process, reference is made to... Figure 6 As shown, this embodiment constructs a closed-loop feedback control technology solution based on real-time data driving, integrating online monitoring, intelligent decision-making, and immediate execution to form a dynamic balance system capable of adaptively suppressing disturbances and maintaining the process window. This solution adopts the following technical means:

[0109] Filtering the real-time monitored physical quantity data extracts its trend components that change over time. The raw signals from online monitoring (such as the shift rate of interference fringes reflecting stress state, slight changes in temperature or air pressure, etc.) usually contain high-frequency noise and random disturbances. By processing with digital filtering (such as low-pass filtering or moving average method), these irrelevant interferences can be filtered out, and smooth trend components that can truly reflect the macroscopic state of the oxidation process can be extracted.

[0110] The trend component is compared with a preset stress state safety range to determine whether the current oxidation process deviates from the stable region. The safety range is an ideal process window that can ensure high-quality channel formation, pre-determined based on a large amount of successful process data and theoretical models. By comparing the real-time extracted trend component with the upper and lower limits of this range, it is possible to objectively determine whether the current process is in a stable region or is developing towards a deviation.

[0111] When a deviation is detected, fine-tuning instructions for one or more oxidation process parameters are generated based on the direction and magnitude of the deviation. It not only determines whether a deviation has occurred, but also intelligently generates specific correction strategies based on the direction and magnitude of the deviation and preset control rules. The strategy clearly indicates which oxidation process parameter (such as humidity or temperature) needs to be adjusted, as well as the fine-tuning instructions (such as "slight increase" or "moderate decrease").

[0112] The fine-tuning instructions are translated into executable parameter settings for the process equipment, enabling online intervention in the ongoing oxidation process. The control system translates these decision-making fine-tuning instructions into specific parameter settings that the oxidation equipment controller can recognize and execute (e.g., converting "slightly increase humidity" into "increase the humidity setpoint by 0.5%)), and immediately sends the instructions to the actuators. This online intervention is real-time and continuous, ensuring timely correction of deviations and ultimately bringing the monitored physical quantity data back to a safe range, thereby ensuring that the oxidation process always operates on its optimal trajectory.

[0113] Implementing closed-loop feedback control presents two specific engineering challenges: First, how to translate abstract fine-tuning instructions into concrete, explicit, and effective process actions; a lack of specific mapping rules can lead to inaccurate control responses or even adverse effects. Second, when multiple process parameters (such as temperature, humidity, and airflow) are coupled together, how to avoid system oscillations or runaway caused by simultaneous adjustments of multiple parameters, i.e., solving the problem of multivariable coordinated control.

[0114] To solve the first problem mentioned above, namely the issue of specifying and refining control commands, refer to Figure 7 As shown, the following technical means are used in this embodiment:

[0115] For continuous increases exceeding limits: When it is detected that the stress trend is continuously rising and deviating from the upper limit of the safe range, it is determined that the root cause is that the oxidation reaction rate is too fast, resulting in excessive stress accumulation. At this time, a fine-tuning instruction is generated to reduce the oxidation reaction rate. According to the principle of oxidation process, reducing the humidity in the oxidation furnace is one of the most direct and effective means to suppress the reaction rate. This instruction is executed immediately to curb the further rapid increase of stress by slowing down the reaction.

[0116] For continuous stress decline exceeding the limit: When the stress trend continues to decline and deviates from the lower limit of the safe range, it usually means that the reaction is uneven or locally stagnant, which may lead to asymmetry in the channel shape. At this time, fine-tuning instructions are generated to improve the uniformity of the oxidation reaction. The solution is to increase the local airflow circulation intensity of the oxidation environment. By enhancing the consistency of the transport of reactants and by-products, the oxidation reaction front can be uniformly advanced, thereby correcting the trend of stress decline.

[0117] For severe high-level oscillations: When the stress trend oscillates severely at a high level, it indicates that the reaction interface is in an unstable state, which may be caused by factors such as an imbalance between exothermic reaction and heat dissipation. At this time, fine-tuning instructions are generated to stabilize the reaction interface. The strategy is to slightly reduce the ambient temperature and maintain the current humidity. The cooling aims to reduce the reaction activity and bring the system back to a smoother operating point. At the same time, the humidity is kept constant to avoid introducing another variable that may cause new disturbances. These measures work together to dampen the oscillations and restore stability.

[0118] To solve the second problem mentioned above, namely the coordination issue of multi-parameter adjustments, refer to Figure 8 As shown, this embodiment constructs a sequence decoupling control strategy based on sensitivity ranking, avoiding the simultaneous adjustment of multiple coupling parameters, and instead intervening in a time-sharing and orderly manner according to their influence. The specific implementation process is as follows:

[0119] First, the correlation between key oxidation process parameters (such as temperature, humidity, air pressure, and airflow) is established offline. Then, by analyzing historical data or models, the sensitivity ranking of each parameter to stress state is determined. For example, a sensitivity conclusion of "humidity > temperature > airflow" might be drawn.

[0120] When the system determines that multiple parameters need adjustment, it prioritizes the adjustment of the single dominant process parameter most sensitive to current stress fluctuations, based on their sensitivity. For example, humidity is adjusted first. This follows the principle of "grasping the main contradiction" and achieving initial correction using the most effective single method.

[0121] After adjusting the dominant parameters, the trend of the monitoring data is reassessed. The system will wait for a short response time to observe whether the stress trend begins to revert due to the adjustment of the dominant parameters.

[0122] If the price has returned to a safe range, the adjustment is complete.

[0123] If the stress state has not fully returned to the safe range, then based on the reassessed state, the next minor parameter adjustment is initiated according to sensitivity ranking. For example, based on humidity adjustment, temperature is then fine-tuned. This process is repeated until the process returns to stability.

[0124] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for precisely controlling the oxidation pore of a VCSEL chip, characterized in that, The method comprises the following steps: Based on the material property data of the material to be oxidized, the shrinkage stress distribution caused by the volume shrinkage of the material during the oxidation process is calculated, and the distribution parameters thereof in different depth regions of the oxidation channel are obtained; including: establishing a corresponding relationship map between the material oxidation process and the volume shrinkage rate; according to the corresponding relationship map and the three-dimensional structure of the target oxidation channel, the oxidation process is discretized into multiple continuous time stages; for each time stage, according to the current material oxidation state, the strain field caused by the volume shrinkage of the time stage is determined; based on the strain field, in combination with the internal resistance to deformation of the material under different strain states, the shrinkage stress distribution parameters corresponding to different depth regions of the oxidation channel are analyzed; Based on the distribution parameters, the influence of the shrinkage stress on the shape and position of the oxidation channel is simulated to obtain the predicted deformation variable and position offset of the channel; Taking the reduction of the predicted deformation variable and position offset as the target, the oxidation process parameters are optimized to generate a set of optimized oxidation process parameters; According to the position offset and the shrinkage stress distribution, the initial oxidation path is corrected to generate a compensated oxidation path; including: decomposing the position offset into an offset vector in a plane perpendicular to the oxidation advancing direction; analyzing the spatial gradient of the shrinkage stress distribution to identify the dominant stress region and its direction causing the position offset; according to the offset vector and the direction of the dominant stress region, an opposite path correction amount is applied to the corresponding spatial position of the initial oxidation path in the opposite direction and related to the size; the opposite path correction amount is smoothly integrated into the entire initial oxidation path to form a continuous transition compensation oxidation path from the starting point to the end point; The optimized oxidation process parameters and the compensated oxidation path are used to perform the oxidation process, and the physical quantity reflecting the stress state is monitored in real time during the oxidation process; according to the fluctuation trend of the monitoring data, the oxidation process parameters are dynamically adjusted to stabilize the formation process of the oxidation channel.

2. The method of claim 1, wherein: In the process of analyzing the shrinkage stress distribution parameters, if a non-continuous jump of the stress parameters in different depth regions is identified, the following steps are performed for smoothing processing: A virtual stress buffer layer is constructed between adjacent depth regions, and the stress parameter value of the stress buffer layer is determined by weighted fusion of the parameter values of the adjacent regions above and below; Using the stress buffer layer as an intermediary, the overall stress distribution of the oxidation channel from the surface to the bottom is recalculated to make the stress parameters present a smooth transition along the depth direction; The stress distribution parameters after smoothing are used for subsequent prediction of the shape and position of the channel.

3. The method of claim 1, wherein: Based on the distribution parameters, the influence of the shrinkage stress on the shape and position of the oxidation channel is simulated to obtain the predicted deformation variable and position offset of the channel, including: The shrinkage stress distribution parameters are mapped to equivalent loads acting on the boundary of the initial channel geometric model; Under the action of the equivalent loads, the structural response of the initial channel geometric model is calculated to identify the displacement vector of each point on the channel side wall; According to the displacement vector, the predicted three-dimensional topography of the channel is reconstructed, and compared with the initial geometric model to quantitatively obtain the overall deformation variable of the channel; Based on the reconstructed predicted three-dimensional morphology, the deviation trajectory of its geometric center axis relative to the initial design position is calculated to determine the position offset of the hole.

4. The method of claim 3, wherein: When calculating the structural response of the initial hole geometry model, if it is identified that the displacement vector field has a sharp change in space, resulting in distortion of the predicted morphology, a coordination process is performed: Between adjacent regions where the displacement vector has a sharp change, a virtual structural constraint link is introduced to coordinate the inconsistency of the displacement; Under the coordination of the virtual structural constraint link, the displacement vector of each point on the hole side wall is recalculated to obtain a smoothed displacement field; Based on the smoothed displacement field, the reconstruction and comparison of the predicted three-dimensional morphology of the hole are performed again to obtain a deformation and position offset that is more consistent with physical reality.

5. The method of claim 1, wherein: To reduce the predicted deformation and position offset, the oxidation process parameters are optimized to generate a set of optimized oxidation process parameters, including: Establishing a correlation mapping table between the oxidation process parameters and the predicted deformation and position offset; Based on the correlation mapping table, setting the allowed upper limit of the deformation and position offset as the optimization boundary; Systematically generating multiple sets of alternative oxidation process parameter combinations within the optimization boundary; For each set of alternative parameter combinations, the corresponding oxidation process is estimated, and the new stress distribution and the resulting new deformation and position offset under this oxidation process are deduced; From all the alternative combinations, the set of parameters that optimizes the comprehensive results of the new deformation and position offset is selected as the optimized oxidation process parameters.

6. The method of claim 1, wherein: According to the fluctuation trend of the monitoring data, the oxidation process parameters are dynamically adjusted, including: Filtering the real-time monitored physical quantity data to extract the trend component over time; Comparing the trend component with the pre-set stress state safety interval to determine whether the current oxidation process deviates from the stable region; When it is determined that it deviates, a fine-tuning instruction for one or more oxidation process parameters is generated according to the direction and amplitude of the deviation; The fine-tuning instruction is converted into a parameter setting value executable by the process equipment to implement online intervention on the ongoing oxidation process, so that the monitored physical quantity data returns to the safety interval.

7. The method of claim 6, wherein: According to the direction and amplitude of the deviation, a fine-tuning instruction for one or more oxidation process parameters is generated, including: If the monitored stress state trend component shows a continuous rise and deviates from the upper limit of the safety interval, a fine-tuning instruction for reducing the oxidation reaction rate is generated, and the fine-tuning instruction is to reduce the environmental humidity in the oxidation furnace; If the monitored stress state trend component shows a continuous decline and deviates from the lower limit of the safety interval, a fine-tuning instruction for improving the oxidation reaction uniformity is generated, and the fine-tuning instruction is to increase the local airflow circulation intensity of the oxidation environment; If the monitored stress state trend component shows a sharp oscillation at a high level, a fine-tuning instruction for stabilizing the reaction interface is generated, and the fine-tuning instruction is to slightly reduce the environmental temperature and maintain the current humidity.

8. The method of claim 6, wherein: When generating a fine-tuning instruction based on the deviation, if it is determined that multiple process parameters need to be adjusted and there is mutual coupling effect, coordination control is performed, including: Establishing the correlation and sensitivity ranking between key oxidation process parameters; According to the sensitivity ranking, the single dominant process parameter that is most sensitive to the current stress fluctuation is adjusted first; After adjusting the dominant parameter, the change trend of the monitoring data is re-evaluated; If the stress state does not completely return to the safe interval, the adjustment of the next secondary parameter is started based on the re-evaluated state according to the sensitivity ranking, and so on until the process is restored to stability.

Images