A controllable mode optical filter and optimization method

Abstract

The present invention provides a controllable mode optical filter, comprising a TE0 pass / resistance control component and a TE1 pass / resistance control component based on a silicon waveguide, wherein the TE0 pass / resistance control component comprises a plurality of first fins arrayed along the direction of light incidence, and the TE1 pass / resistance control component comprises a second fin extending along the direction of light incidence, wherein the outer side surface of the second fin is coplanar with the side surface of the silicon waveguide. The first fin and the second fin are both made of phase change material. By adjusting the size, number, and position of the first fins, a short fin array is formed in which the TE0 pass / resistance control component meets a predetermined first output target. By adjusting the size and number of the second fins, a long fin structure is formed in which the TE1 pass / resistance control component meets a predetermined second output target. The controllable mode optical filter provided by the present invention significantly reduces the number of devices that need to be designed in coordination, which is conducive to large-scale on-chip integration of photonic devices.

Description

Technical Field

[0001] The present invention belongs to the technical field of planar optical waveguide integrated devices, and in particular relates to a controllable mode optical filter and an optimization method based on a multi-algorithm hybrid optimization design. Background Art

[0002] With the development of human society, people's demand for information is rapidly increasing. Consequently, modern communication systems are facing increasing demands for bandwidth and capacity. Optical interconnect technology is considered a very promising high-speed communication solution, and it can further increase capacity by applying advanced photon multiplexing techniques. In recent decades, various multiplexing technologies have emerged to increase the communication capacity of optical communication systems, such as wavelength division multiplexing (WDM) and space division multiplexing (SDM). These technologies have been developing rapidly to support the rapid growth of optical transmission capacity. In addition, optical mode division multiplexing (MDM) technology has also gradually entered the field of vision. Mode division multiplexing technology multiplexes different modes of light onto a single multimode or few-mode fiber for transmission and demultiplexes the different modes into their corresponding signals at the receiving end. In recent years, silicon-on-insulator integrated mode division multiplexing (MDM) systems have attracted widespread attention due to their small size, compatibility with CMOS manufacturing processes, and scalability to the established wavelength division multiplexing (WDM) systems.

[0003] Optical modes, as a dimension of light, can serve as an important form of multiplexing in optical communications, just like wavelengths. The existence of different modes of light has long been recognized, and light from different modes can experience dispersion and crosstalk. Therefore, in mode-division multiplexing systems, mode filters are essential components, filtering out unwanted modes and allowing only the desired modes to pass, similar to the function of wavelength filters in wavelength-division multiplexing systems. In practice, the same device may need to filter out different modes of light while retaining another mode in different situations. Relying solely on passive devices cannot meet this requirement. Using a single-function mode-pass / blocking control device typically requires the design of multiplexed devices to split and recombine the light. This increases the number of components and the area occupied, hindering large-scale on-chip integration of photonic devices. Furthermore, devices designed using traditional methods often suffer from excessive size. To ensure the integration of the entire system, the size of the mode filter should be kept to a minimum. Summary of the Invention

[0004] The technical problems to be solved by the present invention are: 1. It is difficult for on-chip passive optical control devices to change the control results in real time according to changes in demand; 2. Single-function mode optical pass / resistance control devices generally require the design of multi-channel devices, which is not conducive to large-scale on-chip integration of photonic devices; 3. Devices designed by traditional design methods generally have the disadvantages of being too large and having complex processing technology.

[0005] In order to achieve the above-mentioned object, the present invention provides a controllable mode optical filter, comprising a TE0 pass / resistance control component and a TE1 pass / resistance control component based on a silicon waveguide, wherein the TE0 pass / resistance control component comprises a plurality of first fins arrayed along the light incident direction, and the TE1 pass / resistance control component comprises a second fin extending along the light incident direction, wherein the outer side surface of the second fin is coplanar with the side surface of the silicon waveguide, and the first fin and the second fin are both made of phase change material, and a TE0 pass / resistance control component is formed by adjusting the size, number and position of the first fin to meet the predetermined first A short fin array of output targets, wherein the first output target is the difference between the sum of the transmittances of the phase change material in the A state, the TE0 mode light in the A state, and the TE1 mode light in the C state, and the transmittance of the phase change material in the C state, and the TE0 mode light. By adjusting the size and number of the second fins, a long fin structure that meets the predetermined second output target in the TE1 pass / resistance control component is formed. The second output target is the difference between the sum of the transmittances of the phase change material in the A state, the TE1 mode light in the A state, and the TE0 mode light in the C state, and the transmittance of the phase change material in the C state, and the TE1 mode light.

[0006] Preferably, the two second fins are symmetrically arranged about the central axis of the upper surface of the silicon waveguide.

[0007] Preferably, the phase change material is germanium antimony selenium telluride.

[0008] Preferably, the width of the silicon waveguide is 1 μm and the height is 220 nm. The first fins are six rectangular blocks of equal size. The distance between the centers of two adjacent first fins is 208 nm. The length of each first fin along the direction of light incidence is 130 nm, the width is 385 nm, and the height is 340 nm. The axis of the first fin as a whole along the direction of light incidence is arranged parallel to the axis of the silicon waveguide.

[0009] Preferably, the second fin has a length of 8 μm, a width of 75 nm, and a height of 340 nm along the light incident direction.

[0010] Preferably, the second fin has a length of 3.52 μm, a width of 112 nm, and a height of 170 nm along the light incident direction.

[0011] The present invention also provides a method for optimizing a controllable mode optical filter, comprising the following steps:

[0012] Determine the initial optimized structure: The initial optimized structure includes a silicon waveguide with a width of 1 μm, five first fins arrayed along the light incident direction, the first fins are evenly spaced on the central axis of the upper surface of the silicon waveguide, and second fins are symmetrically arranged on both sides of the central axis of the upper surface of the silicon waveguide. Both the first fin and the second fin are rectangular parallelepipeds;

[0013] Determine the parameters to be optimized: the parameters to be optimized for the first fin are length, width, height, and the center distance between two adjacent first fins; the parameters to be optimized for the second fin are length, width, and height;

[0014] TE0 pass / resistance control component optimization: determine the ranges of the four parameters that need to be optimized for the first fin, and the first parameter space can be formed by the four parameter ranges. A certain number of points are selected in a subspace of the first parameter space, that is, the parameter combinations of this number are selected. Each parameter combination of the first fin is input into the simulation model to obtain the structure of a TE0 pass / resistance control component. The transmittance of the TE0 / TE1 mode light corresponding to each parameter combination in the A state and the C state is calculated by the finite time-domain difference method. The first output target of the device optimization process is calculated from the four transmittances. The first output target is the difference between the sum of the transmittances of the phase change material in the A state, the TE1 mode light in the A state, and the TE1 mode light in the C state, and the transmittance of the phase change material in the C state.

[0015] TE1 pass / resistance control component optimization: determine the three parameter ranges that need to be optimized for the second fin, and the three parameter ranges can constitute a second parameter space. Select a certain number of points in a subspace of the second parameter space, that is, select the parameter combinations of this number. Input each parameter combination of the second fin into the simulation model to obtain a TE1 pass / resistance control component. Calculate the transmittance of the TE0 / TE1 mode light corresponding to each parameter combination in the A state and the C state respectively by the finite time-domain difference method. Calculate the second output target of the device optimization process from the four transmittances. The second output target is the difference between the sum of the transmittances of the phase change material in the A state, the TE0 mode light in the A state, and the TE0 mode light in the C state, and the transmittance of the TE1 mode light when the phase change material is in the C state.

[0016] Obtaining a final structure: determining a structure of the controllable mode optical filter based on an optimal parameter combination corresponding to the first output target convergence value and the second output target convergence value.

[0017] Preferably, a mixing optimization step is further included before the step of obtaining the final structure, and the mixing optimization step includes:

[0018] The one-to-one correspondence between parameter combinations and transmittance in a subspace of the first parameter space / second parameter space is taken as a data set. All data sets are then fed into the neural network as training sets. The neural network is trained to obtain a general correspondence between "parameter combinations" and "transmittance" in the parameter subspace, thus obtaining a neural network proxy model in the parameter space.

[0019] The neural network proxy model is substituted into the optimization algorithm, and the optimization algorithm is used to calculate the optimal parameter combination that maximizes the first output target / second output target in this parameter space.

[0020] Preferably, if one or more parameter values ​​in the optimal parameter combination are at the boundary position, the parameter space is changed to the other side of the boundary, and the TE0 pass / resistance control component optimization, TE1 pass / resistance control component optimization and mixed optimization steps are repeated again until the values ​​of the four parameters in the "optimal parameter combination" are not on the boundary. At this time, the "optimal parameter combination" can be used as the final optimization result.

[0021] Preferably, before obtaining the final structure step, a fin number optimization step is further included, and the fin number optimization step includes:

[0022] After determining the optimal parameter combination, the optimal number of fins under the optimal parameter combination is obtained by adjusting the number of the first fins or the second fins and performing simulation.

[0023] The beneficial effects of the present invention are as follows: the present invention uses the phase change material GSST to control the pass / resistance of two modes of light. Since the phase change response time of GSST can reach the order of 10ns, the controllable mode optical filter can change the control result in real time according to changes in demand; the controllable mode optical filter can realize four functions of "TE0 pass and TE1 resistance", "TE1 pass and TE0 resistance", "TE0 and TE1 full resistance", and "TE0 and TE1 full pass". By separately regulating the phase state of GSST in the "TE0 pass / resistance control component" and the "TE1 pass / resistance control component", it can effectively realize the separate pass / resistance control of TE0 and TE1 mode light at the same time, thereby greatly reducing the number of devices that need to be designed in coordination, and facilitating large-scale on-chip integration of photonic devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a structural diagram of one embodiment of the present invention.

[0025] Figure 2 for Figure 1 The side view structural diagram of the TE0 pass / resistance control component is shown.

[0026] Figure 3 for Figure 1The top view structural diagram of the TE0 pass / resistance control component is shown.

[0027] Figure 4 It is the pass / resistance condition of TE0 and TE1 mode light in the TE0 pass / resistance control component; the curve marked with "TE0,A" is the transmittance of TE0 mode light in the component when GSST is in the A state; "TE1,A" is the transmittance of TE1 mode light in the component when GSST is in the A state; "TE0,C" is the transmittance of TE0 mode light in the component when GSST is in the C state; "TE1,C" is the transmittance of TE1 mode light in the component when GSST is in the C state.

[0028] Figure 5 for Figure 1 The side view structural diagram of the TE1 pass / resistance regulating component is shown.

[0029] Figure 6 for Figure 1 The schematic diagram of the top view of the TE1 pass / resistance control component is shown.

[0030] Figure 7 It is the pass / resistance condition of TE0 and TE1 mode light in the TE1 pass / resistance control component; the curve marked with "TE0,A" is the transmittance of TE0 mode light in the component when GSST is in the A state; "TE1,A" is the transmittance of TE1 mode light in the component when GSST is in the A state; "TE0,C" is the transmittance of TE0 mode light in the component when GSST is in the C state; "TE1,C" is the transmittance of TE1 mode light in the component when GSST is in the C state.

[0031] Figure 8 for Figure 1The calculation results of the overall control performance of the embodiment shown; the curve marked with "TE0,AC" is the transmittance of the TE0 mode light in the component when the GSST of the first part is in state A and the GSST of the second part is in state C; the curve marked with "TE0,AA" is the transmittance of the TE0 mode light in the component when the GSST of the first part is in state A and the GSST of the second part is in state A; the curve marked with "TE0,CA" is the transmittance of the TE0 mode light in the component when the GSST of the first part is in state C and the GSST of the second part is in state A; the curve marked with "TE0,CC" is the transmittance of the TE0 mode light in the component when the GSST of the first part is in state C and the GSST of the second part is in state C. The transmittance of light in this component; the curve marked with "TE1,AC" is the transmittance of TE1 mode light in this component when the GSST of the first part is in state A and the GSST of the second part is in state C; the curve marked with "TE1,AA" is the transmittance of TE1 mode light in this component when the GSST of the first part is in state A and the GSST of the second part is in state A; the curve marked with "TE1,CA" is the transmittance of TE1 mode light in this component when the GSST of the first part is in state C and the GSST of the second part is in state A; the curve marked with "TE1,CC" is the transmittance of TE1 mode light in this component when the GSST of the first part is in state C and the GSST of the second part is in state C.

[0032] Figure 9 It is a structural schematic diagram of another embodiment of the invention.

[0033] Figure 10 for Figure 9 The side view structural diagram of the TE1 pass / resistance regulating component is shown.

[0034] Figure 11 for Figure 9 The schematic diagram of the top view of the TE1 pass / resistance control component is shown.

[0035] Figure 12 for Figure 1 The calculation results of the overall control performance of the embodiment shown in FIG. 1; the TE1 pass / resistance control component in this embodiment is designed under the condition that the height of the long fin is variable; the meaning of each mark in the figure is shown in FIG. Figure 6 Description.

[0036] In the figure, 100, silicon waveguide; 1, first fin; 2, second fin. DETAILED DESCRIPTION

[0037] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0038] Example 1

[0039] Please also refer to Figure 1-8 The controllable mode optical filter provided in this embodiment includes a TE0 pass / resistance control component and a TE1 pass / resistance control component based on a silicon waveguide 100, wherein the TE0 pass / resistance control component includes six first fins 1 arrayed along the light incident direction, and the TE1 pass / resistance control component includes a second fin 2 extending along the light incident direction, wherein the outer side surface of the second fin 2 is coplanar with the side surface of the silicon waveguide 100, and the first fin 1 and the second fin 2 are both made of germanium antimony selenium tellurium (Ge2Sb2Se4Te, GSST) material.

[0040] GSST is a phase-change material, as described above, that has two phases: A (amorphous) and C (crystalline). When exposed to external influences (such as light, electricity, or heat), its optical properties, such as its refractive index and extinction coefficient, undergo dramatic changes, enabling significant manipulation of light. Unlike traditional phase-change materials like germanium antimony telluride (GST), GSST replaces some of the tellurium (Te) element with selenium (Se). This allows it to maintain a large refractive index difference while achieving infrared transparency in both its crystalline and amorphous states, significantly reducing optical loss. GSST also boasts a larger volume within which it can undergo phase transitions. To achieve complete amorphization, the thickness limit of a GST nanostructure is approximately 100 nanometers. Compared to traditional phase-change materials like GST, GSST is inherently more susceptible to amorphization, making it less susceptible to the cooling rate during the melt-quench phase. This provides greater design flexibility for various structural parameters within the device.

[0041] In the TE0 pass / resistance control component, the GSST fin controls the pass / resistance of TE0 mode light by changing its own phase state. When the GSST fin is in the A state, TE0 mode light can pass through the component with almost no loss, while when the GSST fin is in the C state, TE0 mode light is almost completely absorbed and dissipated by the component. In the TE1 pass / resistance control component, the GSST fin controls the pass / resistance of TE1 mode light by changing its own phase state. When the GSST fin is in the A state, TE1 mode light can pass through the component with almost no loss, while when the GSST fin is in the C state, TE1 mode light is almost completely absorbed and dissipated by the component.

[0042] By adjusting the size and position of the first fin 1, a short fin array is formed in the TE0 pass / resistance control component to meet the predetermined first output target FOM1. The first output target FOM1 is the difference between the sum of the transmittance of the phase change material for TE0 mode light in the A state, TE1 mode light in the A state, and TE1 mode light in the C state, and the transmittance of the phase change material for TE0 mode light in the C state. The size and position of the first fin are first optimized using an algorithm to maximize the FOM1. The number of fins is then adjusted based on the optimal FOM1 to achieve the final result.

[0043]

[0044] in, is the transmittance of TE0 mode light in the component when GSST is in state A; is the transmittance of TE1 mode light in the component when GSST is in state A; is the transmittance of TE0 mode light in the component when GSST is in C state; is the transmittance of TE1 mode light in the component when GSST is in C state.

[0045] By adjusting the size of the second fin 2, a long fin structure that meets the predetermined second output target FOM2 is formed in the TE1 pass / resistance control component. The second output target FOM2 is the difference between the sum of the transmittances of TE0 mode light in the A state, TE1 mode light in the A state, and TE0 mode light in the C state of the phase change material and the transmittance of TE1 mode light when the phase change material is in the C state.

[0046]

[0047] in, is the transmittance of TE0 mode light in the component when GSST is in state A; is the transmittance of TE1 mode light in the component when GSST is in state A; is the transmittance of TE0 mode light in the component when GSST is in C state; is the transmittance of TE1 mode light in the component when GSST is in C state.

[0048] The present invention uses phase-change material GSST to control the pass / resistance of two modes of light. Since the phase change response time of GSST can reach the order of 10ns, the controllable mode optical filter can change the control results in real time according to changes in demand. The controllable mode optical filter can realize four functions: "TE0 passes and TE1 blocks", "TE1 passes and TE0 blocks", "TE0 and TE1 are fully blocked", and "TE0 and TE1 are fully passed". By separately regulating the phase state of GSST in the "TE0 pass / resistance control component" and the "TE1 pass / resistance control component", it can effectively realize the separate pass / resistance control of TE0 and TE1 mode light at the same time, thereby greatly reducing the number of devices that need to be designed in coordination, which is beneficial to the large-scale on-chip integration of photonic devices.

[0049] TE0 and TE1 mode optical Figure 1 The light is incident on the A-surface of the silicon waveguide, and the dimension of the fin along the light incidence direction is defined as the length. In this embodiment, the TE0 pass / resistance control component is located near the light incidence end of the silicon waveguide 100. It is understood that the TE1 pass / resistance control component can also be located near the light incidence end of the silicon waveguide 100.

[0050] More specifically, a second fin 2 is provided on the silicon waveguide 100, which can also achieve filtering of TE1 mode light. In this embodiment, two second fins 2 are symmetrically arranged about the central axis of the upper surface of the silicon waveguide 100, and the long side of the second fin 2 is flush with the long side of the silicon waveguide 100. Two second fins 2 achieve the same effect as one second fin 2, and a single second fin 2 can be shortened to less than 1 / 2 of the original length, which can greatly reduce the size of the TE1 pass / resistance control component, and is conducive to large-scale on-chip integration of photonic devices.

[0051] More specifically, the width W0 of the silicon waveguide 100 is 1 μm, the height H0 is 220 nm, the first fins 1 are six rectangular parallelepipeds of equal size, the distance D between the centers of two adjacent first fins 1 is 208 nm, the length L1 of the first fin 1 along the light incident direction is 130 nm, the width W1 is 385 nm, and the height H1 is 340 nm. The axis of the first fin 1 in the light incident direction is set parallel to the axis of the silicon waveguide 100.

[0052] More specifically, the length L2 of the second fin 2 along the light incident direction is 8 μm, the width W2 is 75 nm, and the height H2 is 340 nm. In this embodiment, setting the height of the second fin 2 to be the same as that of the first fin 1 can greatly reduce the processing difficulty and save calculation time.

[0053] Example 2

[0054] Please also refer to Figure 9-12The technical solution provided in this embodiment is basically the same as that in the first embodiment, except that the length of the second fin 2 along the light incident direction is 3.52 μm, the width is 112 nm, and the height is 170 nm.

[0055] When the height of the second fin 2 can be freely selected, the actual occupied area of ​​the designed selective control mode filter is only 1μm×4.69μm, which is very compact. Figure 12 As shown, within the wide bandwidth of 1500-1650nm, the transmittance of the desired mode light under different control conditions is higher than -1dB (approximately 80%). Crosstalk under different control conditions is lower than -8.4dB (approximately 14.4%) within the bandwidth of 150-1600nm, and lower than -10.56dB (approximately 8.8%) at a wavelength of 1550nm. All four indicators are superior to the design results with the two fins of the same height.

[0056] For these two mode filters, if you want to further reduce crosstalk in actual use, you can directly achieve it by increasing the number of fins in the first part and increasing the length of the long fins in the second part.

[0057] According to the comparison between the above results and existing devices, it can be seen that compared with mode filters designed by other methods, the present invention has the advantages of being controllable, small in area, and having a large bandwidth.

[0058] The present invention provides a method for optimizing a controllable mode optical filter, comprising the following steps:

[0059] Determine the initial optimized structure: The initial optimized structure includes a silicon waveguide with a width of 1 μm, five first fins arrayed along the light incident direction, the first fins are evenly spaced on the central axis of the upper surface of the silicon waveguide, and the second fins are symmetrically arranged on both sides of the central axis of the upper surface of the silicon waveguide. Both the first fins and the second fins are rectangular.

[0060] Determine the parameters to be optimized: the parameters to be optimized for the first fin are length L1, width W1, height H1 and the center distance D between two adjacent first fins; the parameters to be optimized for the second fin are length L2, width W2 and height H2.

[0061] TE0 pass / resistance control component optimization: determine a parameter subspace spanned by the parameter ranges of the four parameters of the first fin, length L1, width W1, height H1 and center spacing D (for example, length L1∈[10,100], width W1∈[200,300], height H1∈[10,100], center spacing D∈[100,200], in nm, then these four parameter ranges can be spanned into a parameter space in four-dimensional space. Since the parameter value ranges contained in this space are not the full expected value ranges of L1, W1, H1, and D, for example, the full expected value range of L1 is 10nm~500nm, the full expected value range of W1 is 10nm~1000nm, etc., therefore, this parameter space is only a subspace of the full value ranges of all parameters to be adjusted); then randomly generate a certain number of Parameter combination, that is, randomly generate a number of "length-width-height-center spacing" combinations, and then input each parameter combination into the simulation model of the component to obtain a component configuration, and calculate the transmittance of the TE0 / TE1 mode light corresponding to each parameter combination in the A state and C state respectively by the finite time-domain difference method (because the first fin needs to regulate the TE0 and TE1 mode light differently in different phases, the finite time-domain difference method is used to calculate the transmittance of the two mode lights when the fin is in the A state and the C state, that is, each "parameter combination" needs to be calculated four times in total), and the first output target FOM1 of the device optimization process is calculated from the four transmittances. The first output target FOM1 is the difference between the sum of the transmittances of the phase change material in the A state, the TE1 mode light in the A state, and the TE1 mode light in the C state, and the transmittance of the phase change material in the C state.

[0062]

[0063] in, is the transmittance of TE0 mode light in the component when GSST is in state A; is the transmittance of TE1 mode light in the component when GSST is in state A; is the transmittance of TE0 mode light in the component when GSST is in C state; is the transmittance of TE1 mode light in the component when GSST is in C state.

[0064] Optimization of TE1 pass / resistance control component: The optimization process of TE1 pass / resistance control component is consistent with that of TE0 pass / resistance control component, and the optimized parameters are the length L2, width W2 and height H2 of the second fin; each parameter combination of the second fin is input into the simulation model to obtain a TE1 pass / resistance control component, and the transmittance of TE0 / TE1 mode light corresponding to each parameter combination when the fin is in state A and state C is calculated by finite time-domain difference method. The second output target FOM2 of the device optimization process is calculated from the four transmittances. The second output target FOM2 is the difference between the sum of the transmittances of the phase change material in state A, TE1 in state A and TE0 in state C, and the transmittance of TE1 when the phase change material is in state C.

[0065]

[0066] in, is the transmittance of TE0 mode light in the component when GSST is in state A; is the transmittance of TE1 mode light in the component when GSST is in state A; is the transmittance of TE0 mode light in the component when GSST is in C state; is the transmittance of TE1 mode light in the component when GSST is in C state.

[0067] Obtain the final structure: Determine the structure of the controllable mode optical filter based on the optimal parameter combination corresponding to the first output target convergence value and the second output target convergence value; the FOM expression shows that in the device optimization process, it is equally important to increase the transmittance of the mode light you want to retain and reduce the transmittance of the mode light you want to block. In the algorithm optimization process, the FOM function is an important indicator to guide the optimization direction. The entire optimization process is carried out in the direction of increasing the FOM function. In addition, regarding the conditions for the termination of the optimization, on the one hand, the results of the algorithm must reach a certain degree of convergence, and on the other hand, the value of the FOM function must meet certain expectations, for example: FOM>2.7, and

[0068] More specifically, before obtaining the final structure, a hybrid optimization step is also included, and the hybrid optimization step includes:

[0069] The obtained one-to-one correspondence between parameter combinations and transmittance is treated as a data set. All data sets are then fed into the neural network as training sets. The neural network is trained to obtain a general correspondence between "parameter combination" and "transmittance" in the parameter subspace, thus obtaining a neural network proxy model in the parameter space.

[0070] The neural network proxy model is substituted into the optimization algorithm, and the optimization algorithm is used to calculate the optimal parameter combination that maximizes the FOM function value in this parameter space.

[0071] Since the hybrid optimization process for the TE1 pass / resistance control component is identical to that for the TE0 pass / resistance control component, the hybrid optimization steps for the TE0 pass / resistance control component are used as an example. The resulting one-to-one correspondence between parameter combinations (length L1, width W1, height H1, and center-to-center spacing D) and transmittance is treated as a data set. All these data sets are then fed into a neural network as training sets. The neural network training process then determines the general correspondence between parameter combinations and transmittance within this parameter space, thereby generating a "neural network proxy model" within this parameter space. The significance of the neural network proxy model lies in that, given a new parameter combination, four transmittance values ​​(transmittance for two modes of light in the GSST A and C states) can be quickly obtained without requiring four calculations using the finite time-domain difference method. This significantly accelerates optimization and enables simultaneous optimization of multiple parameters. Once the neural network proxy model is derived, it is substituted into a particle swarm optimization (PSO) algorithm, which then calculates the optimal parameter combination that maximizes the FOM function within this parameter space.

[0072] In this embodiment, the optimization algorithm uses the particle swarm optimization algorithm, which has a fast calculation speed and is easy to implement in code. It is understandable that the genetic algorithm, simulated annealing algorithm or ant colony algorithm can also be used in place of the particle swarm optimization algorithm in the hybrid optimization step.

[0073] More specifically, if one or more parameters in the "optimal parameter combination" are at the boundary, the parameter space is changed to the other side of the boundary (for example, the original parameter space is: L∈[10,100], W∈[200,300], H∈[10,100], D∈[100,200], in nm. In the "optimal parameter combination" obtained after the optimization is completed, the length is 100, and the values ​​of the other parameters are not on the boundary. Then the parameter space is changed to L∈[100,200], W∈[200,300], H∈[10,100], D∈[100,200]), and the TE0 pass / resistance control component optimization, TE1 pass / resistance control component optimization and mixed optimization steps are repeated again until the values ​​of the four parameters in the "optimal parameter combination" are not on the boundary. At this time, the "optimal parameter combination" can be considered as the final optimization result.

[0074] More specifically, before obtaining the final structure, a fin number optimization step is also included, and the fin number optimization step includes:

[0075] After determining the optimal parameter combination, by adjusting the number of the first fin or the second fin and simulating, and considering the transmittance of the TE0 / TE1 mode light in state A and state C respectively according to the results, the optimal number of fins under the optimal parameter combination can be obtained.

[0076] Considering the convenience of processing, the first fin and the second fin are both located on the top surface of the silicon waveguide 100 and are kept at the same height. It is understood that the first fin 1 is located on the top surface of the silicon waveguide 100, and the second fin 2 is located on the side of the silicon waveguide 100. It can also be understood that the design of the TE1 pass / resistance control component can be optimized.

[0077] 1. The present invention uses phase-change material GSST to control the pass / blocking of two modes of light. Since the phase change response time of GSST can reach the order of 10ns, the device can change the control results in real time according to changes in demand.

[0078] 2. The present invention realizes four functions of "TE0 pass and TE1 block", "TE1 pass and TE0 block", "TE0 and TE1 fully block", and "TE0 and TE1 fully pass" on a single device. It can effectively realize the separate pass / block control of TE0 and TE1 mode light at the same time by separately regulating the phase state of GSST in the "TE0 pass / block control component" and the "TE1 pass / block control component", thereby greatly reducing the number of devices that need to be designed in coordination, and facilitating large-scale on-chip integration of photonic devices.

[0079] 3. The present invention optimizes the design of device structures through neural networks and optimization algorithms, which largely solves the drawbacks of devices designed using traditional design methods, such as being too large in size and having complex processing techniques.

[0080] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A controllable mode optical filter, characterized in that: The invention comprises a TE0 pass / resistance control component and a TE1 pass / resistance control component based on a silicon waveguide, wherein the TE0 pass / resistance control component comprises a plurality of first fins arrayed along the light incident direction, and the TE1 pass / resistance control component comprises a second fin extending along the light incident direction, wherein the outer side surface of the second fin is coplanar with the side surface of the silicon waveguide, and the first fin and the second fin are both made of phase change material, and by adjusting the size, number and position of the first fin, a short fin array in the TE0 pass / resistance control component that meets a predetermined first output target is formed. The first output target is the difference between the sum of the transmittances of the phase change material for TE0 mode light in the A state, TE1 mode light in the A state, and TE1 mode light in the C state, and the transmittance of the phase change material for TE0 mode light in the C state. By adjusting the size and number of the second fins, a long fin structure is formed in the TE1 pass / resistance control component that meets the predetermined second output target. The second output target is the difference between the sum of the transmittances of the phase change material for TE0 mode light in the A state, TE1 mode light in the A state, and TE0 mode light in the C state, and the transmittance of the phase change material for TE1 mode light in the C state. The two second fins are symmetrically arranged about the central axis of the upper surface of the silicon waveguide; The phase change material is germanium, antimony, selenium and tellurium.

2. The controllable mode optical filter according to claim 1, wherein: The width of the silicon waveguide is 1 μm and the height is 220 nm. The first fins are six rectangular blocks of equal size. The distance between the centers of two adjacent first fins is 208 nm. The length of each first fin along the direction of light incidence is 130 nm, the width is 385 nm, and the height is 340 nm. The axis of the first fin along the direction of light incidence is arranged parallel to the axis of the silicon waveguide.

3. The controllable mode optical filter according to claim 2, wherein: The second fin has a length of 8 μm, a width of 75 nm, and a height of 340 nm along the light incident direction.

4. The controllable mode optical filter according to claim 2, wherein: The second fin has a length of 3.52 μm, a width of 112 nm, and a height of 170 nm along the light incident direction.

5. A method for optimizing a controllable mode optical filter, characterized in that: The following steps are involved: Determine the initial optimized structure: The initial optimized structure includes a silicon waveguide with a width of 1 μm, five first fins arrayed along the light incident direction, the first fins are evenly spaced on the central axis of the upper surface of the silicon waveguide, and second fins are symmetrically arranged on both sides of the central axis of the upper surface of the silicon waveguide. Both the first fin and the second fin are rectangular parallelepipeds; Determine the parameters to be optimized: the parameters to be optimized for the first fin are length, width, height, and the center distance between two adjacent first fins; the parameters to be optimized for the second fin are length, width, and height; TE0 pass / resistance control component optimization: determine the ranges of the four parameters that need to be optimized for the first fin, and the first parameter space can be formed by the four parameter ranges. A certain number of points are selected in a subspace of the first parameter space, that is, the parameter combinations of this number are selected. Each parameter combination of the first fin is input into the simulation model to obtain the structure of a TE0 pass / resistance control component. The transmittance of the TE0 / TE1 mode light corresponding to each parameter combination in the A state and the C state is calculated by the finite time-domain difference method. The first output target of the device optimization process is calculated from the four transmittances. The first output target is the difference between the sum of the transmittances of the phase change material in the A state, the TE1 mode light in the A state, and the TE1 mode light in the C state, and the transmittance of the phase change material in the C state. TE1 pass / resistance control component optimization: Determine the three parameter ranges that need to be optimized for the second fin, and the three parameter ranges can constitute a second parameter space. Select a certain number of points in a subspace of the second parameter space, that is, select this number of parameter combinations, and input each parameter combination of the second fin into the simulation model to obtain a TE1 pass / resistance control component structure. The transmittance of the TE0 / TE1 mode light corresponding to each parameter combination in the A state and the C state is calculated by the finite time-domain difference method. The second output target of the device optimization process is calculated from the four transmittances. The second output target is the difference between the sum of the transmittances of the phase change material in the A state, the TE1 mode light in the A state, and the TE0 mode light in the C state, and the transmittance of the TE1 mode light when the phase change material is in the C state; Obtaining a final structure: determining a structure of the controllable mode optical filter based on an optimal parameter combination corresponding to the first output target convergence value and the second output target convergence value.

6. The optimization method of the controllable mode optical filter according to claim 5, wherein: Before obtaining the final structure, a hybrid optimization step is also included, and the hybrid optimization step includes: The one-to-one correspondence between parameter combinations in a subspace of the first parameter space / second parameter space and transmittance is taken as a data set, and all data sets are then fed into the neural network as training sets. The neural network is trained to obtain a general correspondence between parameter combinations in the parameter subspace and transmittance, thereby obtaining a neural network proxy model in the parameter space. The neural network proxy model is substituted into the optimization algorithm, and the optimization algorithm is used to calculate the optimal parameter combination that maximizes the first output target / second output target in this parameter subspace.

7. The method for optimizing a controllable mode optical filter according to claim 6, wherein: If one or more parameter values ​​in the optimal parameter combination are at the boundary of the subspace, the parameter subspace is moved to the other side of the boundary, and the TE0 pass / resistance control component optimization, TE1 pass / resistance control component optimization, and mixed optimization steps are repeated again until the values ​​of the four parameters in the "optimal parameter combination" are no longer on the boundary. At this time, the "optimal parameter combination" can be used as the final optimization result.

8. The method for optimizing a controllable mode optical filter according to claim 5, wherein: Before obtaining the final structure, a fin number optimization step is also included, and the fin number optimization step includes: After determining the optimal parameter combination, the optimal number of fins under the optimal parameter combination is obtained by adjusting the number of the first fins or the second fins and performing simulation.

Images