A hybrid optimization method for multilayer film metamaterial particles and enhanced optical power

Abstract

The present application relates to a kind of mixed optimization method of multilayer film metamaterial microparticle and enhanced optical force, multilayer film metamaterial microparticle is made of multiple periodic structures superposition, each periodic structure is made of different refractive index material layer superposition, the universal multilayer film metamaterial microparticle of the present application, both can simulate the characteristics of existing crystal, also can be realized by design optical constant New crystal, will further expand the applicability of optical force and moment photonic probe in optical tweezers system.In addition, using the mixed optimization method of enhanced optical force of the present application, the time cost of photonic probe design is greatly reduced, which is helpful to the comprehensive study of optical force enhancement.

Description

Technical Field

[0001] This invention belongs to the field of optical micromanipulation technology, specifically relating to a hybrid optimization method for multilayer film metamaterial particles and enhanced optical power. Background Technology

[0002] Optical tweezers technology possesses the ability to manipulate optical forces and precisely detect linear and angular motions without contact, enabling applications such as the measurement of forces and torques of individual biomolecules, quantum optomechanics, colloidal self-assembly, and light-driven nanomachines. For these applications, silica or polystyrene microspheres have become popular choices as photonic probes. For the combined controllable manipulation of forces and torques, optically birefringent photonic probes, due to their advantages in reducing scattering, minimizing thermal effects, and completely restricting rotational degrees of freedom, have been successfully applied in optical tweezers systems, serving as optical force and torque sensors. These probes primarily include calcite (CaCO3), aragonite (CaCO3), liquid crystal RM257, and rutile TiO2.

[0003] While these media probes each have their advantages, they also possess specific limitations. Although silica and polystyrene microspheres offer high uniformity and chemical stability, their lack of optical birefringence makes them unsuitable as torque sensors. Calcite / spherulite CaCO3 crystals exhibit birefringence (n... r While RM257 contains materials with high birefringence (n=0.1-0.16), it is chemically unstable in water, making it less suitable for biological and microfluidic environments. RM257's shape is not tunable, limited to spherical shapes, which is detrimental to the regulation of transverse moments dependent on shape anisotropy. Rutile TiO2, although possessing the highest birefringence (n=0.1-0.16), is a different story. r It is a natural crystal with a refractive index of 0.1-0.16, and its shape can be optimized through top-down processing. However, its refractive index (n=2.6) is too high, which will cause high light scattering. This will cause the particles to be unable to be stably captured in the axial direction, and relatively expensive high numerical aperture objectives are required to compensate for it, thus limiting its application range.

[0004] Therefore, existing birefringent photon probes are still insufficient to meet the requirements. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a hybrid optimization method for multilayer film metamaterial microparticles and optical enhancement, which solves the problem that multilayer film metamaterial microparticles are difficult to achieve optimal optical and torque performance. The optimization method of this application solves the problems of low optimization efficiency and long optimization time.

[0006] The present invention includes a multilayer film metamaterial microparticle, which is composed of multiple periodic structures stacked together, each periodic structure being composed of layers of materials with different refractive indices stacked together.

[0007] Preferably, each periodic structure is composed of a high-refractive-index material layer and a low-refractive-index material layer stacked together. More preferably, each periodic structure is composed of one high-refractive-index material layer and one low-refractive-index material layer stacked together.

[0008] Preferably, the high refractive index material is Si (refractive index 3.48) or Nb2O5 (refractive index 2.26), with Si being the most preferred; the low refractive index material is Si3N4 (refractive index 1.98) or SiO2 (refractive index 1.45), with Si3N4 being the most preferred.

[0009] Preferably, each periodic structure has the same length and width.

[0010] Preferably, in each periodic structure, the ratio ρ of the thickness d2 of the high refractive index material to the total thickness d is 0.2, the length and width W of each periodic structure are both 450 nm, and the ratio AR of the height H to the width W of each periodic structure, i.e., H / W, is 2.5.

[0011] Preferably, the total number N of the periodic structures, i.e., N = W / d, is 20-25, and more preferably 22 or 23.

[0012] This invention provides a hybrid optimization method for enhancing the optical power of multilayer film metamaterial microparticles, namely, a structural optimization method for multilayer film metamaterial microparticles, comprising:

[0013] For any periodic structure of the multilayer film metamaterial microparticles to be optimized, the relationship between the width and aspect ratio of the periodic structure and the stiffness of the first axial optical trap is calculated using the finite element simulation method, with the ratio of the high refractive index material layer to the thickness of the periodic structure as a premise of large grid search (preferably the width and aspect ratio of the periodic structure are the premise of large grid search), and the target thickness ratio with the best first axial optical trap stiffness is selected.

[0014] Based on the target thickness ratio, the parameter space formed by the width and aspect ratio is divided into multiple grids (preferably, the grid is a fine grid with a higher wire density than the large grid mentioned above), and one grid is selected as the initial first grid;

[0015] A first preset number of points are randomly selected from the first grid as the initial population. The second axial optical trap stiffness of each point is calculated using the trained neural network surrogate model. The points are moved using the particle swarm optimization algorithm based on the second axial optical trap stiffness. The second axial optical trap stiffness of the moved points is calculated using the trained neural network surrogate model until the optimal point with the maximum second axial optical trap stiffness is obtained.

[0016] Based on the location of the optimal point in the first grid, a second grid is searched, and the first grid is updated to the second grid and iterative calculation is performed until the optimal point is located inside the first grid.

[0017] The width and aspect ratio of the optimal point located inside the first grid are used as the optimal parameter combination for the multilayer film metamaterial microparticles.

[0018] Preferably, when calculating the thickness ratio of the high-refractive-index material layer to the periodic structure using the finite element simulation method, the relationship between the width and aspect ratio of the periodic structure and the stiffness of the first axial optical trap is determined, and the target thickness ratio with the largest first axial optical trap stiffness is selected, including:

[0019] Within different preset ranges of the ratio of the high refractive index material layer to the thickness of the periodic structure, the width of the periodic structure, and the aspect ratio, the first axial optical trap stiffness corresponding to any width and aspect ratio is calculated using the finite element simulation method based on the relationship between the width and aspect ratio of the periodic structure and the first axial optical trap stiffness, with different first spacings, and with different first spacings. This forms an array of axial optical trap stiffness corresponding to each of the thickness ratios.

[0020] From the various thickness ratios, select at least one target thickness ratio that is optimal for the axial optical trap stiffness array.

[0021] Preferably, before randomly selecting a first preset number of points from the first grid as the initial population, the process includes:

[0022] Select a second preset number of points from the first grid at a second spacing;

[0023] The third axial optical trap stiffness of each point is calculated based on the width and aspect ratio, forming a data set for each point including the width, the aspect ratio, and the third axial optical trap stiffness;

[0024] The neural network proxy model is trained using the second preset number of points as the training set to obtain the trained neural network proxy model.

[0025] Preferably, the step of searching for the second grid based on the location of the optimal point in the first grid includes:

[0026] Determine the position of the optimal point in the first grid;

[0027] If the optimal point is located at the boundary of the first grid, then search for a second grid located on the other side of the boundary of the first grid.

[0028] This application utilizes the equivalent dielectric film theory (EMT) to design a multilayer metamaterial microparticle composed of alternating high and low refractive index materials. The equivalent dielectric constant of this microparticle, relative to the parallel component of the z-axis (ε... ∥ =ε x =ε z ) and vertical component (ε ⊥ =ε y The following are respectively represented:

[0029]

[0030] Here, ε1 and ε2 are the dielectric constants of the high- and low-refractive-index materials, respectively. By changing the thickness of the two materials, different equivalent refractive indices and birefringences can be achieved.

[0031] The corresponding refractive index n o =n ∥ =√ε ∥ n e =n ⊥ =√ε ⊥ Equivalent refractive index n eff For (n e +n o ) / 2, birefringence n r For |n ∥ -n ⊥ |

[0032] Preliminary realization of enhanced optical force. The optical force F experienced by the multilayer film metamaterial particles is divided into gradient force F0. grad and scattering force F scat F = F grad +F scat The gradient force is primarily used to stably capture multilayer metamaterial particles, while the scattering force mainly pushes the particles away from the stable capture point. Therefore, to achieve stable particle capture, it is necessary to enhance the gradient force and reduce the scattering force.

[0033] According to the Rayleigh model, the magnitude of the optical gradient force is proportional to the refractive index difference Δn between the particle and the surrounding medium, and related to W. 3 Proportional:

[0034]

[0035] Where c is the speed of light in a vacuum, and ε0 is the dielectric constant of a vacuum. It is the square of the electric field gradient.

[0036] Re(α) is the equivalent polarizability of the particle:

[0037]

[0038] Where, n m Let m be the refractive index of the surrounding medium, and n be the refractive index of the medium. eff / n m The terms related to refractive index can be written as:

[0039]

[0040] ∝ indicates that they are directly proportional.

[0041] The magnitude of light scattering force is related to Δn 2 Proportional to W 6 Proportional:

[0042]

[0043] in, <s>σ is the time-averaged Poynting vector, and σ is the scattering cross section.

[0044]

[0045] Where a is related to wavelength and n m Relevant constants.

[0046] Therefore, our approach is to increase the n of the particles. eff This increases Δn and enhances the optical gradient force. At the same time, it improves the shape of the particles. That is, while keeping the total volume constant, the spheres commonly found in optical tweezers systems are replaced with cuboids that are "stretched" along the optical axis, thereby reducing the contact area between the incident light and the particle scattering cross section, reducing light scattering, and thus reducing the scattering force.

[0047] Since the particles also exhibit birefringence, the spin of the particles around the optical axis can be achieved by adjusting the linear polarization angle of the incident light.

[0048] Axial optical trap stiffness k z It can be used to evaluate the stability of axial trapping of particles and is calculated using the following formula:

[0049] F z =-k z (z-z0) (7)

[0050] Where F z This refers to the axial optical force, specifically the component of the optical force F along the axial direction. Here, we take the maximum value, F. z Calculate, z is when F z z is the position of the particle at its maximum value, and z0 is the position of the equilibrium point.

[0051] The beneficial effect of this invention is that the microparticles of this invention can allow for any combination of refractive index and birefringence, thereby providing an optimal balance between the performance of force and torque sensors.

[0052] In this work, we develop and implement a general computational method for the design of micro / nano photonic probes to achieve photodynamic enhancement and torque with minimal computation time and resources. This method is scalable to the design of all linear micro / nano photonic probes.

[0053] This invention proposes using multilayer metamaterial microparticles to achieve optical force enhancement and torque, and presents a hybrid optimization algorithm for rapid global optimization. By employing EMT (Electromechanical Metamaterials), multilayer metamaterial microparticles with alternating high and low refractive indices are designed. Parameters ρ, W, and AR are adjusted to increase Δn while decreasing W, thereby enhancing the optical gradient force while reducing light scattering force. The parameter adjustment process is accomplished using a hybrid optimization algorithm. First, under the premise of a large grid search, the adjustment range of the ρ value is narrowed. Then, at an optimal ρ value, W and AR are changed, and optimization is performed using a particle swarm optimization algorithm. Finally, a neural network surrogate model is used to predict simulation results.

[0054] The goal of the optimization method in this application is to save computation time and quickly find the optimal solution that meets our requirements.

[0055] Under the large grid condition in step 5, the values ​​of ρ have been determined to be 0.2 and 0.3, which means that the global optimal solution cannot appear in other ρ values. Then, for the values ​​of ρ to be 0.2 and 0.3, grids were actually selected in the top, bottom, left and right of the two-dimensional graph shown in Figure 3(a) and (b) for trial and error. However, no clear trend was found in which grid the optimal point exists, i.e., the left, right and bottom as starting points. Finally, the calculation was performed from the top according to the steps shown by the arrows. The optimal solution can be found with the fewest calculation steps, so this process is regarded as the global optimal solution.

[0056] The optical constants of the microparticles are tunable throughout the entire process of achieving photodynamic enhancement in this application, allowing for the artificial design of suitable optical parameters based on optimization goals. Furthermore, the microparticles are fabricated using a top-down process based on a sacrificial layer, enabling flexible control of particle geometry, uniformity, and high yield.

[0057] The universal multilayer metamaterial microparticles of this invention can both simulate the properties of existing crystals and create entirely new crystals by designing optical constants, further expanding the applicability of force and torque photonic probes in optical tweezers systems. Furthermore, the introduction of a hybrid optimization algorithm significantly reduces the time cost of photonic probe design, contributing to comprehensive research on photomechanical enhancement. Attached Figure Description

[0058] Figure 1 This is a schematic diagram illustrating the optimization principle of one of the optimization methods in this embodiment of the invention.

[0059] Figure 2 This is a schematic diagram of the structure of the multilayer film metamaterial microparticles of the present invention, wherein 1 is a light source, 2 is a high refractive index material, and 3 is a low refractive index material.

[0060] Figure 3a In this embodiment of the invention, under strong focused Gaussian beam illumination with a p = 0.2, W and AR are related to k. z The correspondence.

[0061] Figure 3b In this embodiment of the invention, under strong focused Gaussian beam illumination with a p = 0.3, W and AR are related to k. z The correspondence.

[0062] Figure 4a In this embodiment of the invention, under optimal parameter conditions, the particle tilt angle and the normalizing torque τ are... x The size relationship.

[0063] Figure 4b In this embodiment of the invention, under optimal parameter conditions, the particle tilt angle and the normalizing torque τ are... y The size relationship.

[0064] Figure 4c In this embodiment of the invention, under optimal parameter conditions, the spin angle and spin torque τ of the microparticles are determined by a strongly focused Gaussian beam. z The size relationship. Detailed Implementation

[0065] The following detailed description of an embodiment of the present invention, in conjunction with the accompanying drawings, should not be construed as limiting the scope of protection of the present invention.

[0066] Example 1

[0067] A detailed method for preparing Si / Si3N4 multilayer film metamaterial microparticles includes:

[0068] (1) Thoroughly clean the silicon substrate.

[0069] (2) Electron beam evaporation deposition of chromium (Cr) sacrificial layer.

[0070] (3) Plasma-assisted reactive magnetron sputtering deposition or thermal evaporation of Si / Si3N4 multilayer films. The total thickness of a single periodic structure, i.e., one Si layer and one Si3N4 layer, is controlled to be 50 nm; the thickness of the Si layer is 10 nm, and the thickness of the Si3N4 layer is 40 nm; the length and width of the Si layer and the Si3N4 layer are both 450 nm, and the total thickness is 1150 nm.

[0071] (4) Spin-coat a layer of positron beam resist.

[0072] (5) The resist layer is etched using electron beam lithography.

[0073] (6) Argon plasma sputtering deposition of Cr corrosion mask.

[0074] (7) Remove unwanted Cr in the unpatterned area with tape, and then remove the remaining scale in the thermal stripping solution.

[0075] (8) Etching of multilayer structures using reactive ion etching.

[0076] (9) Immerse the sample in Cr wet etching solution (bright yellow) to remove the top Cr etching mask layer and the bottom sacrificial Cr layer.

[0077] (10) First rinse the sample, then immerse the sample in deionized water (deionized water is light yellow) to remove the Cr etchant.

[0078] (11) Wash the sample a second time by immersing it in fresh deionized water (the color of the deionized water remains unchanged) without stirring to completely remove the Cr etchant.

[0079] (12) The sample was immersed in a plastic centrifuge tube filled with deionized water and vortexed for 30 seconds to obtain multilayer film cuboid nanoparticles. The remaining silicon substrate was removed, leaving only the multilayer film cuboid nanoparticle solution in the plastic centrifuge tube.

[0080] Example 2

[0081] A hybrid optimization method for enhancing the optical power of Si / Si3N4 multilayer metamaterial particles includes:

[0082] Step 1: Determine the basic configuration of the multilayer metamaterial film particles to be optimized, i.e., as follows Figure 2 The image shows multilayer membrane metamaterial microparticles;

[0083] Step 2: Determine the parameters to be optimized and the optimization objective. The parameters to be optimized are: ρ, W, and AR. Set the range of ρ to 0.1-0.9, the range of W to 50-500nm, and the range of AR to 1-5.

[0084] Step 3: Reduce the number of optimization parameters. Based on a large grid search, i.e., setting the pixel interval to Δρ = 0.1, ΔW = 10nm, and ΔAR = 0.2, the finite element method is used to calculate W, AR, and axial optical trap stiffness k under different ρ values. z Based on the corresponding relationship, two optimal ρ values ​​were found, which are 0.2 and 0.3 respectively;

[0085] Step 4: When ρ takes values ​​of 0.2 and 0.3 respectively, divide the parameter space spanned by the two parameters W and AR into several grids. Randomly select one grid (e.g., W = 350-450nm, AR = 1-2) as the initial grid, and select 51 × 11 = 561 points at equal intervals from this grid to calculate k. z The parameter values ​​are spaced as follows: ΔW = 1 nm, ΔAR = 0.1.

[0086] Step 5: Create a neural network surrogate model, that is, use a neural network to fit and predict the simulation results. This involves applying the calculated W, AR-k values... z Data set (i.e., W and AR are independent variables, k) z The neural network is fed with W and AR as the input and k as the dependent variable. z This is the output neural network proxy model.

[0087] The function of the neural network surrogate model is to use the computation of a neural network to replace the finite element simulation method to obtain calculation results. After training, the neural network surrogate model can quickly calculate the corresponding k′ based on any given set of W and AR. Z , and k′ Z With k z The error is very small, which greatly shortens the simulation time and makes global algorithm optimization in a grid more feasible.

[0088] Step 6: Substitute the neural network agent model into particle swarm optimization to determine the movement directions of W and AR.

[0089] After obtaining the neural network surrogate model, a certain number of points are randomly selected within the grid as the initial population for particle swarm optimization. Then, the neural network surrogate model is used to quickly calculate the k corresponding to each individual in the initial population. z Then, the particle swarm optimization algorithm is used to determine the movement direction of each individual within the grid and move these points accordingly. The "neural network surrogate model calculation" and "particle swarm optimization" processes are iterated multiple times until the result converges to the optimal point. If the optimal point is at the boundary of the grid, the search grid is changed to the other side of that boundary, and steps 4, 5, and 6 are repeated until the optimal point in a certain grid is inside that grid. At this point, the W and AR corresponding to the optimal point can be considered the optimal parameter combination.

[0090] For ease of description, the optimization process of the optimization method is illustrated as follows: Figure 1 As shown.

[0091] Before optimization, the parameters and optimization objectives to be optimized using this method were determined. The parameters to be optimized are: width W and aspect ratio AR. Then, the parameter space spanned by the two parameters W and AR is first divided into several grids. One grid is randomly selected (e.g., W = 350-400nm, AR = 1-2) as the initial grid. From this grid, 51 × 11 = 561 points are selected at equal intervals to calculate k. z The parameter values ​​are spaced as follows: ΔW = 1nm, ΔAR = 0.1; then a neural network surrogate model is created, which takes the values ​​of "W, AR-k" obtained in the previous step as an interval. z The "data set" is fed as a training set into the neural network to train W and AR as inputs, k z This is the output neural network surrogate model. The neural network surrogate model is then substituted into particle swarm optimization. First, a certain number of points are randomly selected within the grid as the initial population for particle swarm optimization. Then, the neural network surrogate model is used to quickly calculate k for each individual in the initial population. z Then, the particle swarm optimization algorithm is used to move these points to new positions. The "neural network surrogate model calculation" and "particle swarm optimization" processes are iterated multiple times until the result converges to the optimal point. If the optimal point is at the boundary of the grid, the search grid is changed to the other side of the boundary, and the above steps are repeated until the optimal point in a certain grid is inside that grid. At this point, the W and AR corresponding to the optimal point can be considered as the optimal parameter combination.

[0092] In each periodic structure, the optical constant of the multilayer metamaterial particles (the ratio of the thickness of the high-refractive-index material to the total thickness) is ρ = 0.2, the length and width of each periodic structure are W = 450 nm, the height-to-width ratio of each periodic structure is AR = 2.5, and H = 1150 nm.

[0093] Subsequently, the microparticles with optimal parameters are processed and synthesized, and then captured using an optical tweezers system. For example... Figure 2 As shown, in the optical tweezers system, multilayer metamaterial particles are captured near the beam waist of light source 1. The light beam output from light source 1 illuminates the multilayer metamaterial particles. Light source 1 emits a Gaussian beam with a wavelength equal to the transmission wavelength of the multilayer metamaterial particles. After passing through an oil immersion objective with a numerical aperture (NA) of 1.4, the Gaussian beam outputs a strongly focused Gaussian beam (i.e., the spot size of the strongly focused Gaussian beam is: a spot radius of 0.41 μm in the x-direction and a spot radius of 0.28 μm in the y-direction). The multilayer metamaterial particles are placed at the focal point of the strongly focused Gaussian beam, which illuminates them. The multilayer metamaterial particles include a high-refractive-index layer 2 and a low-refractive-index layer 3, possessing a high Δn and a small W, thereby achieving the function of optical enhancement.

[0094] like Figure 2 As shown, in this embodiment, a wavelength-tunable laser light source is used. After passing through an oil immersion objective lens with an NA of 1.4, a strongly focused Gaussian beam is output. The multilayer metamaterial particles are composed of alternating layers of high refractive index layer 2 and low refractive index layer 3. The thickness d1 of the high refractive index layer 2 is 10 nm, and the material is Si. The thickness d2 of the low refractive index layer 2 is 40 nm, and the material is Si3N4. ρ = 0.2, W = 450 nm, H = 1150 nm, and aspect ratio AR = 2.5.

[0095] Because the optical axis (n⊥) of the particle is perpendicular to its long side ( Figure 2 By utilizing the optical birefringence of particles in the xy plane and rotating to capture the linear polarization of the light beam, the spin of the particles around the z-axis can be achieved, with the direction as follows: Figure 2 As shown by the circular arrow in the image, the spin torque τ z The calculation is as follows:

[0096] τ z =-Asin(Hk0n r sin(2γ) (8)

[0097] Where, A=Sε0n eff c(E0) 2 / (2ω), S is the cross-sectional area of ​​the particle, ε o is the vacuum permittivity, c is the speed of light in vacuum, E0 is the amplitude intensity of the incident light beam, ω is the angular frequency of the incident light beam. k0=2π / λ0, λ0 is the wavelength of the incident light, H is the altitude, n r Let be the birefringence, and γ be the spin angle of the particle about the z-axis. Furthermore, due to the high AR of the particle and its anisotropic shape, when the particle is offset by an angle α or β relative to the z-axis in the x or y direction, a restoring torque τ will be generated due to the asymmetrical distribution of radiation pressure on the particle surface. x Or τ y This ensures that the orientation of the particles remains unchanged, meaning their major axis is always parallel to the direction of light source propagation (z). The aligning torque is calculated as follows:

[0098]

[0099] Where S is an arbitrary closed surface enclosing the particle, and dS represents the surface integral around the closed surface S. For the surface normal vector, For Maxwell's stress tensor, It is the lever arm.

[0100] The optimal calculations yielded W, AR, and k values ​​for SSN particles with ρ values ​​of 0.2 and 0.3. z The relationship is shown in Figure 3. For particles with ρ = 0.2, as shown in Figure 3(a), only six grid regions need to be calculated. Following the direction indicated by the white arrows, a total of five steps are required to find the optimal solution under the condition of ρ = 0.2, as shown by the black dots in the figure. The corresponding values ​​are W = 450 nm, H = 1150 nm, AR = 2.5, and k z 1.7 pNμm -1 mW -1 .

[0101] For particles with ρ = 0.3, as shown in Figure 3(b), only four grid regions need to be calculated. In grid region 2, the optimal solution under the condition of ρ = 0.3 can be found directly in region 3 by following the black arrow, requiring only two steps. To verify the rationality of this optimal solution, as shown by the black dashed arrow in region 2, the data in grid region 4 was calculated, ultimately pointing to grid region 3, as shown by the point in the figure. The corresponding values ​​are W = 383 nm, H = 1110.7 nm, AR = 2.9, and k z 1.4 pNμm -1 mW -1 .

[0102] Taking ρ=0.2 as an example, without using a hybrid optimization algorithm to find the optimal value, it takes 8-10 minutes to simulate a set of parameters (such as ρ=0.2, W=50nm, AR=1) using the finite element method. If all the data are manually adjusted and calculated (i.e. ρ=0.2, W ranges from 50-500nm, pixel interval ΔW is 1nm, AR ranges from 1-5, pixel interval ΔAR is 0.1), the total simulation time is about 4 months.

[0103] Using a hybrid optimization algorithm, the data is divided into grids. First, the parameters of a grid (e.g., ρ = 0.2, W = 250-300nm, AR = 4-5) are randomly calculated, and the corresponding k is determined. z By fitting and predicting the maximum k through neural networks z If the location is such that only 6 grid cells need to be computed to find the global optimum, i.e., the maximum k, then... z The simulation time for this simulation, including its location, is now approximately 3 days, a reduction to 1 / 40th of the original time. This is highly beneficial for optimizing the optical parameters of the optical force probe, allowing the optimal k to be found in the shortest possible time with minimal computational effort. z Or other required performance.

[0104] After obtaining the global optimal solution for the entire region, which is the optimal solution under the condition of ρ = 0.2, the normalizing torque and spin torque under this condition were calculated, as shown in Figure 4. Figure 4a , Figure 4b The restoring torque τ caused by shape anisotropy is respectively x and τ y , Figure 4c The spin torque τ z .

[0105] As can be seen from the above, in addition to the advantages of system stability, ease of synthesis, and tunable optical parameters, this invention also has the following advantages:

[0106] The SSN microparticles designed in this invention possess photodynamic enhancement properties, and their k z Up to 1.7 pNμm -1 mW -1 .

[0107] The SSN microparticles designed in this invention have birefringence characteristics, and can achieve spin around the optical axis by adjusting the linear polarization angle of the light source.

[0108] This invention employs a hybrid optimization algorithm for optimal design, which can reduce the computation time to 1 / 40 of the original, greatly improving computational efficiency.

[0109] Due to the current development and maturity of micro-nano fabrication technologies, dielectric materials such as Si, Nb2O5, and Si3N4 are suitable for thin-film deposition techniques (such as evaporation and sputtering) to synthesize stable multilayer structures, thus enabling mass production. Furthermore, these materials exhibit low cytotoxicity and high chemical stability; therefore, these multilayer metamaterial particles formed by alternating high and low refractive indices can become efficient and multifunctional bioimaging and therapeutic tools. In addition, this hybrid optimization algorithm will be able to design photonic probes entirely according to the user's desired performance specifications, which will promote comprehensive research in optical tweezers and levitation photomechanics.

[0110] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of protection of this application is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of this application as described above, which are not provided in detail for the sake of brevity.

[0111] One or more embodiments in this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of this application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments in this application should be included within the protection scope of this application.< / s>

Claims

1. A hybrid optimization method for enhancing the optical power of multilayer film metamaterial microparticles, characterized in that, The multilayer film metamaterial microparticle is composed of a plurality of periodic structures, each periodic structure is composed of a plurality of layers of different refractive index materials, and each periodic structure is composed of a layer of high refractive index material and a layer of low refractive index material; The method comprises: For any periodic structure of the multilayer film metamaterial microparticle to be optimized, the relationship between the width and the aspect ratio of the periodic structure and the first axial optical trap stiffness is calculated under the condition that the thickness ratio of the high refractive index material layer to the periodic structure is adopted by using the finite element simulation method, and a target thickness ratio with the best first axial optical trap stiffness is selected; Based on the target thickness ratio, the parameter space composed of the width and the aspect ratio is divided into a plurality of grids, and an optional grid is an initial first grid; A first preset number of points are randomly selected from the first grid as an initial population, the second axial optical trap stiffness of each point is calculated by applying the trained neural network proxy model, the points are moved according to the second axial optical trap stiffness by using the particle swarm optimization algorithm, the second axial optical trap stiffness of the moved points is calculated by applying the trained neural network proxy model, and the optimal point with the maximum second axial optical trap stiffness is obtained until the optimal point is located inside the first grid; The second grid is searched according to the position of the optimal point in the first grid, the first grid is updated to the second grid, and iterative calculation is performed until the optimal point is located inside the first grid; The width and the aspect ratio of the optimal point located inside the first grid are taken as the optimal parameter combination of the multilayer film metamaterial microparticle.

2. The hybrid optimization method for enhancing optical power as claimed in claim 1, wherein, The relationship between the width and the aspect ratio of the periodic structure and the first axial optical trap stiffness is calculated under the condition that the thickness ratio of the high refractive index material layer to the periodic structure is adopted by using the finite element simulation method, and a target thickness ratio with the best first axial optical trap stiffness is selected, comprising: In different preset ranges of the thickness ratio of the high refractive index material layer to the periodic structure, the width of the periodic structure, and the aspect ratio, different first intervals are respectively adopted, the relationship between the width and the aspect ratio of the periodic structure and the first axial optical trap stiffness is adopted, the first axial optical trap stiffness corresponding to any width and aspect ratio under each thickness ratio is calculated by using the finite element simulation method, and each axial optical trap stiffness array corresponding to each thickness ratio is constructed; From each thickness ratio, at least one target thickness ratio with the best axial optical trap stiffness array is selected.

3. The hybrid optimization method for enhancing optical power of claim 1, wherein, Before the first preset number of points are randomly selected from the first grid as the initial population, comprising: Second preset number of points are selected from the first grid at a second interval; The third axial optical trap stiffness of each point is calculated according to the width and the aspect ratio, and a data group including the width, the aspect ratio, and the third axial optical trap stiffness of each point is constructed; The second preset number of points are taken as a training set to train the neural network proxy model, and the trained neural network proxy model is obtained.

4. The hybrid optimization method for enhancing optical power of claim 1, wherein, The second grid is searched according to the position of the optimal point in the first grid, comprising: The position of the optimal point in the first grid is judged; If the optimum point is located at a border of the first grid, a second grid located on the other side of the border of the first grid is searched.

5. The hybrid optimization method for enhancing optical power of claim 1, wherein, The high refractive index material is Si or Nb2O5, and the low refractive index material is Si3N4 or SiO2.

6. The hybrid optimization method for enhancing optical power of claim 1, wherein, The length and width of each periodic structure are the same.

7. The hybrid optimization method of enhancing optical power as claimed in claim 1, 5 or 6, wherein, The ratio of the thickness of the high refractive index material layer to the total thickness in each periodic structure ρ 0.2, the length and width of each periodic structure W 450 nm, the ratio of the height to the width of each periodic structure AR 2.

5.

8. The hybrid optimization method for enhancing optical power as claimed in claim 1, 5 or 6, wherein, The total number of the periodic structures is 20-25.

Images