Dynamic dimmable mode selection control device and method

Abstract

The invention provides a dynamic adjustable light mode selection control device and method, and belongs to the technical field of optical communication, the method comprises a silicon waveguide layer and a PWB layer, each coupling area of the silicon waveguide layer and the PWB layer is provided with a stack absorption layer, the stack absorption layers are deposited in the PWB layer, a space exists between the stack absorption layers and the silicon waveguide layer, and the silicon waveguide layer is arranged in the stack absorption layers. The top of the stack absorption layer is connected with an MEMS cantilever beam, the MEMS cantilever beam is arranged outside the PWB layer, and the MEMS cantilever beam is used for adjusting the distance between the stack absorption layer and the silicon waveguide layer; and the distance between the stack absorption layer and the silicon waveguide layer is kept in a set critical range. The distance between the stack absorption layer and the silicon waveguide layer is adjusted through the MEMS cantilever beam, so that selective absorption of a specific mode is achieved, the needed coupling efficiency is selected for mode transmission, mode selective absorption of the silicon-based photonic device is better carried out, and optimization of the absorption efficiency is facilitated.

Description

Technical Field

[0001] This application relates to the field of optical communication technologies, and particularly to a dynamic tunable optical mode selection control device and method. Background Art

[0002] In the fields of optical communication and photonic computing, an optical mode selection control device is a core component for realizing multimode fiber communication, vortex optical field regulation, and photonic neural networks.

[0003] In traditional silicon-based photonic devices, there are significant technical bottlenecks in mode absorption control. Existing solutions usually adopt an absorption layer design with a fixed position or directly integrated on the waveguide surface, resulting in ineffective differentiation between the substrate mode and the target mode. For example, the distance between the conventional absorption layer and the waveguide is limited within 50 nanometers. Although it can improve the light absorption efficiency, its broadband absorption characteristics will loss all modes without discrimination. When the waveguide width is reduced to 200 nanometers, the leakage rate of the high-order mode can reach more than 90%, but the substrate TE mode will still be synchronously absorbed, causing at least 15% of non-selective loss.

[0004] The root cause of this non-selective absorption lies in the lack of spatial resolution ability of the traditional design for the mode field distribution. When the waveguide enters the tapered region, the mode coupling process presents complex three-dimensional field reconstruction characteristics: there are significant differences in the effective refractive index between the substrate mode and the leakage mode, but the position of the fixed absorption layer cannot match the spatial attenuation length of a specific mode. In addition, the refractive index drift caused by temperature changes will change the mode coupling path, further exacerbating the non-selective interference of the absorption layer. Research shows that the mode extinction ratio of traditional devices is generally lower than 10 dB, making it difficult to meet the requirements of high-speed optical communication for an extinction ratio above 25 dB.

[0005] In summary, in order to better perform mode selection absorption and optimize the absorption efficiency of silicon-based photonic devices, there is an urgent need for a new mode selection control technology to achieve the coupled absorption of specific modes and select the coupling efficiency. Summary of the Invention

[0006] Based on the above technical problems, the present invention provides a dynamic tunable optical mode selection control device, which adjusts the distance between the stack absorption layer and the silicon waveguide layer through a MEMS cantilever beam, so as to achieve the selective absorption of specific modes and realize the mode transmission with the selected required coupling efficiency, and further better perform the mode selection absorption of silicon-based photonic devices and help optimize the absorption efficiency. The present invention also provides a dynamic tunable optical mode selection control method to realize the automatic control of the device and improve the accuracy of control.

[0007] On the one hand, a dynamic adjustable light mode selection control device provided by the present application includes a silicon waveguide layer and a PWB layer. A stack absorption layer is provided in each coupling region of the silicon waveguide layer and the PWB layer. The stack absorption layer is deposited inside the PWB. There is a spacing between the stack absorption layer and the silicon waveguide layer. The top of the stack absorption layer is connected to a MEMS cantilever beam. The MEMS cantilever beam is provided outside the PWB layer and is used to adjust the spacing between the stack absorption layer and the silicon waveguide layer. Among them, the spacing between the stack absorption layer and the silicon waveguide layer is maintained within a set critical range.

[0008] In one embodiment, the silicon waveguide layer is a silicon-based SWG waveguide.

[0009] In one embodiment, the stack absorption layer includes a TiN layer, a HfO2 layer, an ITO layer, and a Cu layer from bottom to top.

[0010] In one embodiment, a low refractive index cladding is provided outside the PWB layer.

[0011] In one embodiment, a piezoelectric thin film is provided on the top of each MEMS cantilever beam.

[0012] On the other hand, a method for a dynamic adjustable light mode selection control device provided by the present application includes: Obtaining sample data of parameters affecting optical mode coupling and corresponding coupling modes and coupling efficiencies of the modes, and generating a data set; Performing normalization processing on the data in the data set; Constructing an optical mode selection control model based on a fully connected feedforward neural network; Training the optical mode selection control model using the normalized data set; Inputting the actual parameters, the target mode, and the target coupling efficiency of the mode into the trained optical mode selection control model to generate a target spacing between the stack absorption layer and the silicon waveguide layer; Adjusting the height position of the stack absorption layer according to the target spacing using the MEMS cantilever beam.

[0013] In one embodiment, obtaining sample data of parameters affecting optical mode coupling and corresponding coupling modes and coupling efficiencies of the modes to generate a data set includes: determining the parameters affecting optical mode coupling. During the process of optical mode selection control testing, each time one parameter is controlled to change while the other parameters remain unchanged. When the measured coupling efficiency of the actual mode meets the set conditions, record the corresponding parameter data, mode, and coupling efficiency of the mode. After arranging them in a preset order, use them as a vector data. Until all parameters are traversed, a data set composed of multiple vector data is formed.

[0014] In one embodiment, the parameters affecting optical mode coupling include geometric parameters, environmental parameters, and the refractive indices of various materials. Among them, the geometric parameters include the width and length of the silicon waveguide layer, the horizontal position of the stacked absorption layer, the spacing between the stacked absorption layer and the silicon waveguide layer, and the length and thickness of each layer material of the stacked absorption layer. The environmental parameters include the temperature, humidity, and pressure of the environment.

[0015] In one embodiment, training the optical mode selection control model using the normalized dataset includes: Dividing the normalized dataset into a training set and a validation set; Taking the spacing between the stacked absorption layer and the silicon waveguide layer as the output, and all other parameters, the coupling mode, and the coupling efficiency of the mode as the input, training the optical mode selection control model using the training set, and updating and optimizing the number of hidden layers and nodes in the model using the control variable method. When the loss function used to evaluate the error between the predicted value and the true value of the model for the test set tends to converge, the model training is completed.

[0016] In one embodiment, after adjusting the height position of the stacked absorption layer using the MEMS cantilever according to the target spacing, it further includes: detecting the coupled mode and the coupling efficiency of the mode after adjustment, and adjusting the spacing between the stacked absorption layer and the silicon waveguide layer using the piezoelectric film in the case of an error.

[0017] A dynamic adjustable optical mode selection control device and method provided by the present application have the following beneficial effects: 1. Dynamically and mainly adjusting the spacing between the stacked absorption layer and the silicon waveguide layer using the MEMS cantilever according to the actual mode selection and coupling efficiency requirements, realizing the selective absorption of specific modes and the mode transmission with the required coupling efficiency by adjusting the spacing, thereby better performing the mode selection absorption of silicon-based photonic devices and helping to optimize the absorption efficiency.

[0018] 2. When adjusting the spacing between the stacked absorption layer and the silicon waveguide layer using the MEMS cantilever, the piezoelectric effect is utilized to cause a slight deformation of the piezoelectric film, thereby driving a slight movement of the MEMS cantilever, realizing the use of the piezoelectric film as a secondary adjustment to cooperate with the MEMS cantilever for control, which helps to offset the mechanical hysteresis brought by the MEMS cantilever regulation.

[0019] 3. Selecting a silicon-based SWG waveguide and adopting a cascaded stepped structure to form a tapered silicon-based SWG waveguide structure helps to prevent optical field leakage, reduce the structural transmission loss, and reduce the structural size.

[0020] 4. A low refractive index cladding is provided outside the PWB layer, which can not only protect the PWB layer, but also form a refractive index difference with the buried oxide layer, generate structural asymmetry in the PWB layer, induce the mode hybridization effect, convert the mode in the PWB layer into a mode matching that in the few-mode fiber, and improve the subsequent coupling efficiency with the few-mode fiber.

[0021] 5. A fully connected feedback neural network is introduced to establish a target optical mode selection control model. By taking the pitch as the output and other parameters, modes, and the coupling efficiency of the modes as the inputs, the automatic adjustment of the pitch can be realized, which helps to improve the accuracy of pitch control.

[0022] 6. Taking the temperature as one of the input parameters, considering the actual temperature when adjusting the pitch helps to offset the refractive index drift caused by temperature changes, and further improves the accuracy of mode selection. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 FIG. is a schematic structural diagram of a dynamically adjustable optical mode selection control device provided by an embodiment of the present invention; Figure 2 FIG. is a schematic structural diagram showing the positional relationship between the stacked absorption layer and the MEMS cantilever beam provided by an embodiment of the present invention; Figure 3 FIG. is a schematic structural diagram of a silicon-based SWG waveguide with a stepwise reduced width provided by an embodiment of the present invention; Figure 4 FIG. is a schematic diagram of the stacked absorption layer and its connection to the power supply provided by an embodiment of the present invention; Figure 5 FIG. is a schematic diagram of the connection of the MEMS cantilever beam and the piezoelectric thin film to the power supply provided by an embodiment of the present invention; Figure 6 FIG. is a neural network architecture diagram of the present invention.

[0024] Description of the reference numerals in the figure: 1. Silicon substrate; 2. Buried oxide layer; 3. PWB layer; 4. Stacked absorption layer; 401. TiN layer; 402. HfO2 layer; 403. ITO layer; 404. Cu layer; 5. Silicon waveguide layer; 6. MEMS cantilever beam; 7. Piezoelectric thin film. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] In order to make the objectives, technical solutions, and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0026] In the present application, PWB stands for Photonic Wire Bonding; MEMS stands for Micro-Electro-Mechanical System.

[0027] Referring to Figures 1 to 5 , this application provides a mode selection control device, including a silicon waveguide layer 5 and a PWB layer 3. Each coupling region of the silicon waveguide layer 5 and the PWB layer 3 is provided with a stack absorption layer 4. The top of the stack absorption layer 4 is connected with a MEMS cantilever beam 6. The MEMS cantilever beam 6 is arranged outside the PWB layer 3 and is used to adjust the distance between the stack absorption layer 4 and the silicon waveguide layer 5. Among them, the distance between the stack absorption layer 4 and the silicon waveguide layer 5 is maintained within a set critical range.

[0028] Specifically, a buried oxide layer 2 is provided on the upper surface of the silicon substrate 1 layer. The silicon waveguide layer 5, the PWB layer 3 and the stack absorption layer 4 are all arranged on the upper surface of the buried oxide layer 2. Before building the mode selection control device, first generate the geometric parameters of the stack absorption layer 4 according to the width of the silicon waveguide layer 5, the refractive indices of the materials of the stack absorption layer 4 and the theoretical environmental factors, and then build the mode selection control device according to the geometric parameters, so that the stack absorption layer 4 of the device has a set thickness and length, and there is a set distance between the stack absorption layer 4 and the silicon waveguide layer 5, so that the stack absorption layer 4 has a certain height compared with the silicon waveguide layer 5, and the height of the silicon waveguide layer 5 is within the set critical range, so as to achieve the absorption of a specific mode.

[0029] In one embodiment, the range of the set thickness is 50 nm to 200 nm; the range of the set length is 10 nm to 50 nm; the set critical range is 100 nm to 500 nm.

[0030] When absorbing a specific mode, it is necessary to use the MEMS cantilever beam 6 to dynamically adjust the distance between the stack absorption layer 4 and the silicon waveguide layer 5 according to the actual situation, so as to help offset the refractive index drift caused by temperature change on the basis of absorbing the specific mode.

[0031] In one embodiment, the vertical displacement range of the MEMS cantilever beam 6 is 0 to 200 nm.

[0032] In one embodiment, this application can achieve the best coupling efficiency for a specific mode by adjusting the distance between the stack absorption layer 4 and the silicon waveguide layer 5, and can also achieve the transmission of a specific coupling efficiency for a mode.

[0033] In one embodiment, the critical range defined in this application is as follows: when the spacing value between the stack absorption layer 4 and the silicon waveguide layer 5 is less than the minimum value of the critical range, all modes in the silicon waveguide layer 5 are absorbed by the stack absorption layer 4; when the spacing value between the stack absorption layer 4 and the silicon waveguide layer 5 is greater than the maximum value of the critical value, the target mode cannot be effectively absorbed; when the spacing value between the stack absorption layer 4 and the silicon waveguide layer 5 is within the critical range, the absorption of the target mode and the adjustment of the absorption efficiency can be achieved by adjusting the size of the spacing value.

[0034] Through the adjustment of the MEMS cantilever beam 6, the absorption of non-target modes in the silicon waveguide is ≤ 0.05, and the absorption rate of the modes in the PWB layer 3 is < 0.05.

[0035] In one embodiment, the silicon waveguide layer 5 is prepared by the following steps: Step 1: Use a single-crystal silicon wafer with a thickness of 5 μm as the substrate material, and remove surface particles and organic residues through ultrasonic cleaning and chemical treatment. Utilize the plasma-enhanced chemical vapor deposition (PECVD) process to generate a 2-μm buried oxide layer 2 (BOX) on the silicon substrate surface to achieve bottom insulation isolation during optical signal transmission.

[0036] Step 2: Deposit a 220-nm-thick single-crystal silicon thin film as the optical waveguide core layer on the BOX layer by PECVD. Use the centrifugal spin-coating process (rotation speed 3000 - 5000 rpm) to uniformly cover the photoresist layer on the silicon surface to ensure that the film thickness uniformity error is ≤ ±5%.

[0037] Step 3: Fabricate a 1:1 ratio optical path mask according to the preset optical path design, and use an ultraviolet lithography system for contact exposure. Light with a specific wavelength passes through the mask pattern area, triggering a photochemical reaction of the photoresist to accurately replicate the waveguide pattern onto the silicon layer surface.

[0038] Step 4: Spray the exposed substrate with an alkaline developer to remove the photoresist in the uncured area. Maintain the development process in the range of 21 - 23°C ± 0.5°C through a closed-loop temperature control system, and use an optical microscopy imaging system to detect the pattern integrity and line width accuracy after development.

[0039] Step 5: Apply the reactive ion beam etching (RIE) technology. Through the synergistic effect of physical ion bombardment and chemically active radicals, selectively remove the silicon material not protected by the photoresist to finally form a sub-micron-level high-precision waveguide structure with a sidewall perpendicularity ≥ 85°.

[0040] In one embodiment, the PWB layer 3 is prepared by the following steps: Step 1: First, thoroughly clean the chip surface with acetone and alcohol, and then deposit a photoresist on the buried oxide layer 2. The photoresist is a TPP acrylate polymer material doped with a certain proportion of TiO2 high refractive index inorganic nanoparticles.

[0041] Step 2: Use femtosecond laser as the excitation light source, set the wavelength of the femtosecond laser to 532 nm, the repetition frequency to 50 MHz, and the pulse width to 55 fs.

[0042] Step 3: After the beam of the femtosecond laser passes through the beam expander, it is focused by the objective lens into the acrylate polymer hybrid material. The laser is focused through the lens to print the PWB layer 3 above the silicon waveguide layer 5.

[0043] In one embodiment, the MEMS layer is prepared through the following steps: Step 1: Spin-coat a photoresist (thickness 1.5 μm) on the silicon substrate 1, and define the cantilever beam profile (length 200 - 500 μm, width 50 - 150 μm) through mask exposure. Perform deep silicon etching (etching rate 5 μm / min) using the Bosch process to form a hollow cantilever structure with a sidewall perpendicularity > 88°.

[0044] Step 2: Deposit a scandium-doped aluminum nitride (Sc-AlN) piezoelectric thin film 7 on the surface of the cantilever beam through radio frequency magnetron sputtering.

[0045] Step 3: Use a buffered hydrofluoric acid solution (BHF) to etch the sacrificial oxide layer (thickness 2 μm), release the movable structure of the cantilever beam, and apply a bias voltage of 0 - 50 V to verify the displacement linearity (0 - 200 nm, non-linear error < 3%).

[0046] In one embodiment, the silicon waveguide layer 5 is a silicon-based SWG waveguide.

[0047] The evanescent field and coupling strength of the SWG waveguide are large, which can greatly shorten the structural length required for mode evolution between the upper and lower waveguide layers, thereby reducing the structural transmission loss.

[0048] In one embodiment, referring to Figure 3 , the silicon waveguide layer 5 is a geometric structure with a stepwise decreasing width, and the decreasing step size matches the effective refractive index difference of the target mode.

[0049] The silicon-based SWG waveguide is a tapered waveguide with a stepwise decreasing width, which is used to generate a mode hybridization effect and couple the high-order modes in the multimode chip into the PWB layer 3. The ideal width of the end of the silicon-based SWG waveguide is an infinitely thin tip, but in actual production, it is limited by the processing accuracy, and finally the tip width is controlled within the range of 60 - 120 nm.

[0050] In this embodiment, a silicon-based SWG waveguide is selected, and a cascaded stepped structure is adopted to form a tapered silicon-based SWG waveguide structure, which helps to prevent optical field leakage, reduce the transmission loss of the structure, and reduce the size of the structure.

[0051] In one embodiment, referring to Figure 2 and Figure 4 , the stack absorption layer 4 is successively a TiN layer 401, a HfO2 layer 402, an ITO layer 403, and a Cu layer 404 from bottom to top. The vertical distance between each layer is adjustable, and each layer is successively deposited and embedded in the PWB layer 3.

[0052] The Cu layer 404 and the TiN layer 401 are electrodes, the HfO2 layer 402 is a high dielectric constant gate dielectric, and the ITO layer 403 is a semiconductor layer. Among them, the ITO layer 403 can change its own electron concentration after being applied with an electric field, thereby changing the permittivity; the function of the HfO2 layer 402 is to enhance the local regulation ability of the electric field of the ITO layer 403 and improve the thermal stability of the structure.

[0053] It can be understood that when the permittivity is negative, the ITO layer 403 exhibits metallic characteristics; when the permittivity is positive, the ITO layer 403 exhibits dielectric characteristics.

[0054] Each absorption layer is independently connected to an external power supply through the Cu layer 404 and the TiN layer 401. By controlling the carrier concentration, the ITO layer 403 is regulated to have metallic characteristics or electrolyte characteristics, thereby achieving a switching effect. When the ITO layer 403 has dielectric characteristics, the device is in the ON state, the refractive index of the stack absorption layer 4 decreases and the absorption decreases, allowing the mode to pass through; when the ITO layer 403 exhibits metallic characteristics, the device is in the OFF state, the refractive index of the stack absorption layer 4 increases and the absorption increases, blocking mode coupling.

[0055] In this embodiment, the design standard of the device is that when the device is in the ON state, it can achieve the best coupling efficiency transmission or a specific coupling efficiency transmission, and when the device is in the OFF state, it can achieve a coupling efficiency transmission lower than 5%.

[0056] In one embodiment, the refractive index of the PWB layer 3 ranges from 1.5 to 1.59, preferably 1.57. The refractive index of the silicon waveguide layer 5 is 3.47.

[0057] The refractive index difference between the PWB layer 3 and the silicon waveguide layer 5 is large. Through the structural asymmetry, the mode hybridization effect is induced to achieve the coupling of the high-order mode to the PWB layer 3.

[0058] In one embodiment, a low refractive index cladding is provided outside the PWB layer 3, and the refractive index of the low refractive index cladding is 1.3.

[0059] The low refractive index cladding can not only protect the PWB layer 3 but also form a refractive index difference with the buried oxide layer 2, generating structural asymmetry in the PWB layer 3, inducing the mode hybridization effect, converting the mode in the PWB layer 3 into a mode matching that in the few-mode optical fiber, and improving the subsequent coupling efficiency with the few-mode optical fiber.

[0060] In one embodiment, referring to Figure 5 , a piezoelectric thin film 7 is provided on the top of each MEMS cantilever beam 6.

[0061] When using the MEMS cantilever beam 6 to adjust the spacing between the stack absorption layer 4 and the silicon waveguide layer 5, the piezoelectric thin film 7 undergoes a slight deformation by using the piezoelectric effect, thereby driving the MEMS cantilever beam 6 to undergo a slight movement, and further offsetting the refractive index drift caused by the mechanical hysteresis and temperature change induced by the regulation of the MEMS cantilever beam 6, realizing the collaborative control using the piezoelectric thin film 7 and the MEMS cantilever beam 6, which helps to offset the temperature drift and mechanical hysteresis of the device.

[0062] In one embodiment, the MEMS cantilever beam 6 can achieve vertical displacement within a set range through electrostatic drive.

[0063] In one embodiment, the material of the piezoelectric thin film 7 is AlN.

[0064] The piezoelectric thin film 7 made of AlN material can generate a deformation of ±1 nm by using the piezoelectric effect, thereby making a slight adjustment to the MEMS cantilever beam 6, which helps to offset the temperature drift and mechanical hysteresis.

[0065] On the other hand, referring to Figure 6 , a dynamic adjustable optical mode selection control method provided by the present application includes: S100, obtaining sample data of parameters affecting optical mode coupling and the corresponding coupling modes and mode coupling efficiencies, and generating a data set.

[0066] Specifically, step S100 includes: determining the parameters affecting optical mode coupling. During the process of optical mode selection control testing, each time one parameter is controlled to change while the other parameters remain unchanged. When the measured actual mode coupling efficiency meets the set conditions, the corresponding parameter data, mode, and mode coupling efficiency are recorded, and after being arranged in a preset order, they are used as a vector data. Until all parameters are traversed, a data set composed of multiple vector data is formed.

[0067] In one embodiment, the set conditions set in the present application are that the non-target mode coupling efficiency threshold is set to ≤0.05, and if it exceeds this value, the set of data is discarded; only the combination of structural parameters that meet the threshold conditions and the target mode coupling efficiency is within the range of 0-1 is recorded to generate a parameter data set. For example: the non-target mode efficiency is 0.12 and the target mode efficiency is 0.5, which is regarded as invalid data; the non-target mode efficiency is 0.03 and the target mode efficiency is 0.45, then it is recorded, and the data collection and storage are completed according to this rule.

[0068] In one embodiment, the parameters affecting optical mode coupling include geometric parameters, environmental parameters, and the refractive indices of various materials; among them, the geometric parameters include the width and length of the silicon waveguide layer 5, the horizontal position of the stack absorption layer 4, the spacing between the stack absorption layer 4 and the silicon waveguide layer 5, and the length and thickness of each layer of material of the stack absorption layer 4; the environmental parameters include the temperature, humidity, and pressure of the environment.

[0069] S200, perform normalization processing on the data in the data set.

[0070] S300, construct an optical mode selection control model based on a fully connected feedforward neural network.

[0071] S400, use the normalized data set to train the optical mode selection control model.

[0072] Specifically, step S400 includes: dividing the normalized data set into a training set and a validation set; using the spacing between the stack absorption layer 4 and the silicon waveguide layer 5 as the output, and all other parameters, coupling modes, and the coupling efficiency of the modes as inputs, training the optical mode selection control model with the training set, and using the control variable method to update and optimize the number of hidden layers and nodes in the model. When the loss function used to evaluate the error between the predicted value and the true value of the model for the test set tends to converge, the model training is completed.

[0073] S500, input the actual parameters, target mode, and target coupling efficiency of the mode into the trained optical mode selection control model to generate the target spacing between the stack absorption layer 4 and the silicon waveguide layer 5.

[0074] S600, adjust the height position of the stack absorption layer 4 using the MEMS cantilever beam 6 according to the target spacing.

[0075] Introduce a fully connected feedback neural network to establish a target optical mode selection control model. By using the spacing as the output and other parameters, modes, and the coupling efficiency of the modes as inputs, the automatic adjustment of the spacing can be realized, which helps to improve the accuracy of spacing regulation. And taking the temperature as one of the input parameters, considering the actual temperature when adjusting the spacing helps to offset the refractive index drift caused by temperature changes, and further improves the accuracy of mode selection.

[0076] In one embodiment, after adjusting the height position of the stack absorption layer 4 according to the target spacing by using the MEMS cantilever beam 6, it further includes: detecting the coupled mode and the coupling efficiency of the mode after adjustment, and adjusting the spacing between the stack absorption layer 4 and the silicon waveguide layer 5 by using the piezoelectric thin film 7 in the case of errors.

[0077] When adjusting the spacing between the stack absorption layer 4 and the silicon waveguide layer 5 by using the MEMS cantilever beam 6, the piezoelectric effect is used to cause a slight deformation of the piezoelectric thin film 7, thereby driving a slight movement of the MEMS cantilever beam 6, so as to realize the collaborative control with the MEMS cantilever beam 6 by using the piezoelectric thin film 7 as a secondary adjustment, which helps to offset the mechanical hysteresis caused by the regulation of the MEMS cantilever beam 6.

[0078] When the device is prepared and the environmental parameters are stable in the preset working range, the host computer system obtains the actual temperature, humidity and air pressure data in real time through the environmental perception module integrating multi-source sensors, inputs the above data, the geometric parameters of the device, and the mode and coupling efficiency to be coupled into the target optical mode selection control model, and performs the main adjustment of the height of the stack absorption layer 4 by using the MEMS cantilever beam 6 according to the predicted spacing output, so as to adjust the spacing between the stack absorption layer 4 and the silicon waveguide layer 5, and further perform mode selection absorption and achieve absorption with the set efficiency. At the same time, a stress feedback network is constructed by using the piezoelectric thin film 7 integrated with the MEMS cantilever beam 6, and dynamic compensation is performed through the piezoelectric effect to form a master-slave collaborative control architecture. This architecture realizes the dynamic mapping between the environmental parameters and the driving voltage through the fuzzy logic algorithm, and suppresses the refractive index drift caused by temperature fluctuations, and finally realizes the precise regulation of the closed-loop adaptive displacement within the tolerance range of the standard MEMS manufacturing process.

[0079] In this application, after obtaining the data set through parameter traversal, the parameters in the data set can provide guidance for the structural parameters of the manufacturing device, so that relevant personnel can select the preferred parameters in the data set for device manufacturing, thereby realizing the static adjustment of the geometric parameters of the device, and further helping to improve the accuracy of mode selection.

[0080] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0081] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A dynamic adjustable light mode selection control device, comprising a silicon waveguide layer and a PWB layer. Each coupling region of the silicon waveguide layer and the PWB layer is provided with a stack absorption layer, and the stack absorption layer is deposited inside the PWB. It is characterized in that, There is a spacing between the stack absorption layer and the silicon waveguide layer. A MEMS cantilever beam is connected to the top of the stack absorption layer. The MEMS cantilever beam is disposed outside the PWB layer and is used to adjust the spacing between the stack absorption layer and the silicon waveguide layer. Among them, the spacing between the stack absorption layer and the silicon waveguide layer is maintained within a set critical range.

2. The device according to claim 1, characterized in that, The silicon waveguide layer is a silicon-based SWG waveguide.

3. The device according to claim 1, characterized in that, The stack absorption layer is, from bottom to top, a TiN layer, a HfO2 layer, an ITO layer, and a Cu layer.

4. The device according to claim 1, characterized in that, There is a low refractive index cladding outside the PWB layer.

5. The device according to claim 1, characterized in that, A piezoelectric thin film is provided on the top of each MEMS cantilever beam.

6. A dynamic dimming mode selection control method, applied to the device according to any one of claims 1 to 5, characterized in that, Including: Obtaining sample data of parameters affecting optical mode coupling, the corresponding coupling modes, and the coupling efficiency of the modes to generate a data set; Performing normalization processing on the data in the data set; Constructing an optical mode selection control model based on a fully connected feedforward neural network; Training the optical mode selection control model using the normalized data set; Inputting the actual parameters, the target mode, and the target coupling efficiency of the mode into the trained optical mode selection control model to generate the target spacing between the stack absorption layer and the silicon waveguide layer; Adjusting the height position of the stack absorption layer using the MEMS cantilever beam according to the target spacing.

7. The method according to claim 6, wherein Obtaining sample data of parameters affecting optical mode coupling, the corresponding coupling modes, and the coupling efficiency of the modes to generate a data set, including: determining the parameters affecting optical mode coupling. During the optical mode selection control test, each time one parameter is controlled to change while the other parameters remain unchanged. When the measured coupling efficiency of the actual mode meets the set conditions, record the corresponding parameter data, the mode, and the coupling efficiency of the mode. After arranging them in a preset order, they are used as a vector data. Until all parameters are traversed, a data set composed of multiple vector data is formed.

8. The method according to claim 6, wherein The parameters affecting optical mode coupling include geometric parameters, environmental parameters, and the refractive indices of various materials. Among them, the geometric parameters include the width and length of the silicon waveguide layer, the horizontal position of the stack absorption layer, the spacing between the stack absorption layer and the silicon waveguide layer, and the length and thickness of each layer of materials in the stack absorption layer. The environmental parameters include the temperature, humidity, and pressure of the environment.

9. The method according to claim 8, wherein Training the optical mode selection control model using the normalized data set, including: Dividing the normalized data set into a training set and a validation set; Taking the spacing between the stack absorption layer and the silicon waveguide layer as the output, and taking all the other parameters, the coupling modes, and the coupling efficiency of the modes as the input. Using the training set to train the optical mode selection control model, and using the control variable method to update and optimize the number of hidden layers and nodes in the model. When the loss function used to evaluate the error between the predicted value and the true value of the model for the test set tends to converge, the model training is completed.

10. The method according to claim 6, characterized in that, After adjusting the height position of the stack absorption layer using the MEMS cantilever beam according to the target spacing, it further includes: detecting the coupled mode and the coupling efficiency of the mode after adjustment. In the case of errors, using the piezoelectric thin film to adjust the spacing between the stack absorption layer and the silicon waveguide layer.

Images