A method for predicting and optimizing loss of amorphous silicon photonic devices
By combining thermal-optical coupling theory with a three-dimensional rough surface model, the loss of amorphous silicon photonic devices can be accurately predicted. This solves the problems of low loss quantification accuracy and trial-and-error dependence on process optimization in existing technologies, and realizes efficient and low-cost loss prediction and optimization, which is suitable for the design and industrialization of integrated photonic devices.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUAZHONG UNIV OF SCI & TECH
- Filing Date
- 2026-02-14
- Publication Date
- 2026-06-02
AI Technical Summary
In existing technologies, amorphous silicon photonic devices suffer from low loss quantification accuracy, low efficiency due to trial-and-error iteration for process optimization, and a disconnect between scattering loss models and actual processes. The lack of a theoretical-experimental closed-loop system results in high costs and long development cycles, making it difficult to meet the needs of mass production and development.
An absorption loss model was constructed using the thermal-optical coupling theory and the thermal conduction equation. A three-dimensional rough surface model was constructed by combining atomic force microscopy scanning and Monte Carlo random simulation. The scattering loss was calculated by the volume current method, forming a closed loop of theory-experiment-process optimization to guide the loss prediction and optimization of amorphous silicon photonic devices.
It enables accurate prediction and optimization of losses in amorphous silicon photonic devices, reduces experimental costs, improves process optimization efficiency, guides photonic chip design, is applicable to existing equipment, requires no high-cost hardware modification, and is suitable for industrial-scale R&D and production.
Smart Images
Figure CN122133328A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of integrated photonic devices, and more specifically, relates to a method for predicting and optimizing the loss of amorphous silicon photonic devices. Background Technology
[0002] Integrated photonic devices are the core foundation of the optoelectronic information industry and are widely used in optical communication, optical computing, biosensing and other fields. Amorphous silicon (a-Si) has become the core material for integrated photonic devices due to its low deposition cost and compatibility with advanced CMOS back-end integration processes. The accurate quantification and verification of its optical loss (absorption + scattering) is a key technical requirement for the optimization of photonic chip design and industrialization, and directly determines the device transmission performance and process development efficiency.
[0003] Currently, the mainstream approach for loss analysis of amorphous silicon photonic devices in the industry is a combination of simulation modeling and offline experimental testing. Theoretically, a simplified waveguide model is constructed using optical simulation software, and absorption and scattering losses are estimated by combining material optical constants and roughness empirical values. Experimentally, absorption losses are inversely calculated by testing optical constants using an ellipsometer, and the total loss is obtained by characterizing roughness using AFM and testing the transmission spectrum. The scattering loss is then further decomposed. To address the insufficient accuracy of weak absorption detection, some solutions use the constant photocurrent method to build a customized platform for supplementary testing. Finally, based on the simulation and test results, the deposition and etching processes are iteratively adjusted through trial and error to optimize the loss.
[0004] Existing loss quantification techniques for amorphous silicon photonic devices have the following key drawbacks: Process iteration relies on blindness: Since loss is strongly correlated with multiple process parameters such as deposition temperature, gas ratio, and etching power, existing technologies lack precise parameter sensitivity analysis methods, which leads to the need for multiple trial and error to complete process optimization. This results in long cycles and high costs, with typical optimization cycles requiring 2-3 months and material loss rates exceeding the target value by more than 30%.
[0005] Insufficient detection capability for weak absorption signals: In the communication band (1550nm), the extinction coefficient k of amorphous silicon is greater than 10. -5 The lower limit of k-value detection for commercial ellipsometers is only 5 × 10⁻⁶. -4 The method cannot capture weak absorption signals, resulting in an absorption loss quantification error of more than 50%. Although the constant photocurrent method can detect weak absorption, it requires the construction of a customized testing platform, with a single platform costing more than 500,000 yuan and a testing cycle of 1-2 weeks, which is difficult to meet the needs of mass R&D.
[0006] The scattering loss modeling is disconnected from the actual process: Existing scattering loss models mostly adopt simplified two-dimensional roughness assumptions, without considering the actual distribution characteristics of the three-dimensional surface profile, and without establishing the correlation between roughness parameters and etching process parameters. As a result, the deviation between the model prediction value and the measured value often exceeds 30%, which cannot effectively guide process optimization.
[0007] Lack of a closed-loop theory-experiment system: In existing technologies, theoretical modeling and experimental testing are independent of each other. There is no standardized experimental procedure or consistency verification criterion, which makes it difficult to effectively verify the model prediction results and lacks clear directional guidance for process optimization. This seriously restricts the performance improvement and industrialization process of amorphous silicon photonic devices. Summary of the Invention
[0008] To address the shortcomings of related technologies, the present invention aims to provide a method for predicting and optimizing the loss of amorphous silicon photonic devices, thereby solving the problems of low accuracy in the existing prediction models for the loss of amorphous silicon photonic devices and low efficiency in process optimization due to its reliance on trial-and-error iteration.
[0009] To achieve the above objectives, in a first aspect, the present invention provides a method for predicting the loss of amorphous silicon photonic devices, comprising: S100. Set the target sample parameters of the amorphous silicon photonic device. Based on the thermal-optical coupling theory, correlate the temperature rise caused by the absorbed light power of amorphous silicon with the resonant frequency shift. Combine the heat conduction equation to solve for the target absorption loss coefficient. S200. Obtain the surface parameters of the target sample by scanning with an atomic force microscope, calculate the root mean square roughness and correlation length from the surface parameters, construct a three-dimensional rough surface model of the target sample including sidewall roughness and top surface roughness using the Monte Carlo random simulation method, and calculate the target scattering loss coefficient using the volume current method. S300. Based on the target absorption loss coefficient and the first error coefficient, the target absorption loss range is obtained; based on the target scattering loss coefficient and the second error coefficient, the target scattering loss range is obtained. S400. The total loss range of the target is obtained based on the target absorption loss range and the target scattering loss range.
[0010] Optionally, step S100 specifically includes: S110. Set the target sample parameters for the amorphous silicon photonic device, establish a three-dimensional waveguide model using COMSOL, and set the simulation parameters. S120. Apply a preset optical power density input to the target sample, simulate and solve the steady-state temperature field, and extract the average temperature rise ΔT; calculate the refractive index change Δn=(dn / dT)·ΔT based on the average temperature rise ΔT, thereby obtaining the resonant frequency shift Δfres = (ng / c)Δn; where ng is the group refractive index, c is the speed of light in vacuum, and dn / dT is the thermo-optic coefficient of amorphous silicon; S130. By fitting the linear slope of Δfres-Pabs, the target absorption loss coefficient αabs is calculated using αabs = γabs·ng / c, where γabs is the thermal-optical coupling coefficient.
[0011] Optionally, step S200 specifically includes: S210. The sidewalls and top surface of the target sample are scanned using an atomic force microscope to obtain the surface parameters of the target sample. The surface parameters are a height matrix h(x,y) containing information on the surface micro-undulations; where x and y are the horizontal and vertical coordinates in the scanning plane, respectively, and h is the height value of the corresponding coordinate point. S220. Calculate the root mean square roughness σ = √[∫∫h²(x,y)dxdy / (S)] based on the height matrix; calculate the autocorrelation function by the height fluctuation data in the horizontal direction, and then fit it using the exponential function R(u)=σ²e^(-u / Lc). The correlation length Lc is obtained by back-calculating the decay rate of the fitted curve; where S is the total area of the AFM scan, u is the horizontal displacement, and R(u) is the autocorrelation value at displacement u. S230, by perturbing Δ with dielectric constant (r)= The coupling rate of the far-field radiation mode is calculated using the volume current method, denoted as 0(n_a-Si² - n_SiO2²)·δ(r), and then expressed as γscat=ω∫∣Sm∣²sinθdθd. The scattering loss coefficient αscat = (ng·γscat) / c is obtained by conversion; where δ(r) is the surface profile function.
[0012] Optional, also includes: Based on the Monte Carlo method, the sidewall roughness fside(z) and top surface roughness ftop(x,y) of the target sample are generated according to the exponential autocorrelation function. Based on the target roughness fside(z) of the sidewall, the target roughness ftop(x,y) of the top surface, and the third error coefficient, the target roughness range is obtained.
[0013] Optionally, the first error coefficient is 10%, the second error coefficient is 15%, and the third error coefficient is ±1nm.
[0014] Secondly, the present invention also provides a method for optimizing the loss of amorphous silicon photonic devices, comprising: T100. Set initial parameters, prepare a sample, and obtain the sample to be tested; substitute the initial parameters into the loss prediction method for amorphous silicon photonic devices as described in any one of the first aspects to obtain the corresponding target total loss range; T200, perform photothermal response testing on the sample to be tested to obtain the total measurement loss of the sample; T300: Determine whether the measured total loss exceeds the target total loss range; if not, the sample to be tested meets the standard; if so, perform atomic force microscopy testing and transmission spectrum testing using an atomic force microscope to obtain the test scattering loss coefficient, sidewall test roughness, and top surface test roughness; obtain the test absorption loss coefficient based on the measured total loss and the test scattering loss coefficient, and execute T400. T400: Optimize the RIE etching process during sample preparation based on the measured absorption loss coefficient, measured scattering loss coefficient, sidewall roughness, and top surface roughness to obtain the optimized sample to be tested. Return to T200 for testing, and repeat until the sample meets the standards.
[0015] Optionally, optimizing the RIE etching process during sample preparation includes: Reduce the power of RIE etching, increase the proportion of O2 used in RIE etching, and extend the RIE etching time.
[0016] Compared with the prior art, the above-described technical solutions conceived in this invention can achieve the following beneficial effects: 1. This invention provides a method for predicting the loss of amorphous silicon photonic devices. It constructs absorption and scattering loss models through multiphysics simulation. The absorption loss model is based on the thermo-optical coupling theory and the heat conduction equation, relating optical power absorption to resonant frequency shift. The scattering loss model extracts roughness parameters using AFM and calculates scattering loss using 3D rough surface modeling and the volume current method. This allows for accurate prediction of the absorption and scattering losses of amorphous silicon photonic devices. This method solves the problems of low loss quantification accuracy and high experimental costs in existing technologies; it achieves the effect of reducing experimental costs, accurately predicting device losses, and guiding photonic chip design.
[0017] 2. This invention provides a method for predicting the loss of amorphous silicon photonic devices. By constructing absorption and scattering loss models in different dimensions, the specific values of the two types of loss and their contribution to the process can be quantified separately. This allows for precise identification of the main sources of device loss, facilitating R&D personnel to develop targeted process optimization solutions, avoiding indiscriminate trial and error adjustments, further reducing the manpower and material costs of process optimization, and achieving precise control and efficient reduction of loss.
[0018] 3. This invention provides a method for optimizing the loss of amorphous silicon photonic devices. During the fabrication process of amorphous silicon photonic devices, loss tests are performed on the fabricated sample devices. The actual loss is compared with the predicted loss to determine whether the device is qualified. If it is not qualified, the fabrication process is optimized. At the same time, the parameters that need to be optimized can be determined through the scattering loss model, guiding the optimization process of the fabrication process. Strict consistency criteria are set to form a closed loop of theory-experiment-process optimization. It has the advantages of strong process guidance and wide applicability, and can be extended to materials such as silicon nitride and indium phosphide, thereby guiding the design and process optimization of photonic chips.
[0019] 4. The present invention provides an optimization method for the loss of amorphous silicon photonic devices. The loss prediction and verification system does not require complex hardware modifications and is compatible with conventional equipment for existing photonic device research and development (such as PECVD, RIE, AFM, spectrometers, etc.). It does not require additional investment in high-cost customized equipment and is easy to implement and promote in the existing research and development and production system, adapting to the needs of large-scale research and development and production in the industry. Attached Figure Description
[0020] Figure 1 This is a schematic cross-sectional view of the amorphous silicon ridge waveguide provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating a method for optimizing the loss of amorphous silicon photonic devices, provided in an embodiment of the present invention. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention 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 merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.
[0022] The following description, in conjunction with a preferred embodiment, illustrates the content involved in the above embodiments.
[0023] Example 1 This invention provides a method for predicting the loss of amorphous silicon photonic devices, comprising: S100. Set the target sample parameters of the amorphous silicon photonic device. Based on the thermal-optical coupling theory, correlate the temperature rise caused by the absorbed light power of amorphous silicon with the resonant frequency shift. Combine the heat conduction equation to solve for the target absorption loss coefficient αabs. S200. Obtain the surface parameters of the target sample by scanning with an atomic force microscope, calculate the root mean square roughness and correlation length from the surface parameters, construct a three-dimensional rough surface model of the target sample including sidewall roughness and top surface roughness using the Monte Carlo random simulation method, and calculate the target scattering loss coefficient αscat using the volume current method. S300. Based on the target absorption loss coefficient αabs and the first error coefficient, the target absorption loss range is obtained; based on the target scattering loss coefficient αscat and the second error coefficient, the target scattering loss range is obtained. S400. The total loss range of the target is obtained based on the target absorption loss range and the target scattering loss range.
[0024] In this embodiment of the invention, the process includes a theoretical prediction stage and an experimental verification stage. In the theoretical prediction stage, a theoretical model of absorption and scattering loss is constructed through multiphysics simulation, and a standardized experimental procedure is used to achieve accurate loss prediction and process optimization. The absorption loss model is based on the thermo-optical coupling theory and the heat conduction equation, relating optical power absorption to resonant frequency shift; the scattering loss model extracts roughness parameters through AFM, and calculates scattering loss using three-dimensional rough surface modeling and the volume current method. In the experimental verification stage, standardized samples are prepared using PECVD deposition and RIE etching, and the model accuracy is verified through photothermal response testing, AFM characterization, and ring resonator testing. Strict consistency criteria are set to form a closed loop of theory-experiment-process optimization.
[0025] For loss prediction of the same type of device, after establishing the absorption loss model and the scattering loss model, only the input parameters need to be changed to obtain the loss prediction results of different devices. For different types of devices, the parameter types are different, and different models need to be established. For example, the parameters of a straight waveguide include length, width and height, while the parameters of a micro-ring also include radius.
[0026] Optionally, step S100 specifically includes: S110. Set the target sample parameters for the amorphous silicon photonic device, establish a three-dimensional waveguide model using COMSOL, and set the simulation parameters. S120. Apply a preset optical power density input to the target sample, simulate and solve the steady-state temperature field, and extract the average temperature rise ΔT; calculate the refractive index change Δn=(dn / dT)·ΔT based on the average temperature rise ΔT, thereby obtaining the resonant frequency shift Δfres = (ng / c)Δn; where ng is the group refractive index, c is the speed of light in vacuum, and dn / dT is the thermo-optic coefficient of amorphous silicon; S130. By fitting the linear slope of Δfres-Pabs, the target absorption loss coefficient αabs is calculated using αabs = γabs·ng / c, where γabs is the thermal-optical coupling coefficient.
[0027] The simulation of the heat conduction equation involves establishing a three-dimensional waveguide model (width W, thickness H, length L) using COMSOL, defining the thermal conductivity of amorphous silicon and SiO2 substrate, setting the air convection heat transfer boundary at the top of the waveguide and the 300K isothermal boundary at the bottom of the substrate, and loading a preset optical power density. In this embodiment, the preset optical power density is Pin = Pabs / [π(W / 2)²], where the absorption efficiency ξ = 10%. The steady-state temperature field is then solved, and the average temperature rise ΔTavg is extracted.
[0028] In this embodiment, an amorphous silicon ridge waveguide model is constructed, referring to... Figure 1 The target sample parameters for the amorphous silicon photonic device were set as follows: amorphous silicon thermal conductivity κ_a-Si = 1.3 W / (m·K), SiO2 substrate thermal conductivity κ_SiO2 = 1.4 W / (m·K), and thermo-optical coefficient dn / dT = 1.8 × 10⁻⁶. -4 K -1 The group refractive index ng = 3.8. The COMSOL simulation settings included: waveguide top air convection heat transfer coefficient h = 10 W / (m²·K), substrate bottom temperature fixed at T0 = 300 K, input optical power Pin set to 100 μW, 300 μW, 500 μW, 700 μW, and 1000 μW, and absorption efficiency ξ = 10%, i.e., Pabs = ξ·Pin. The simulation results showed: average temperature rise ΔTavg at different Pin values were 0.21 K, 0.63 K, 1.05 K, 1.47 K, and 2.10 K, respectively; the slope of the fitted Δfres-Pabs curve was 2.3 GHz / mW; and αabs = 0.28 dB / cm was calculated.
[0029] Optionally, step S200 specifically includes: S210. The sidewalls and top surface of the target sample are scanned using an atomic force microscope to obtain the surface parameters of the target sample. The surface parameters are a height matrix h(x,y) containing information on the surface micro-undulations; where x and y are the horizontal and vertical coordinates in the scanning plane, respectively, and h is the height value of the corresponding coordinate point. S220. Calculate the root mean square roughness σ = √[∫∫h²(x,y)dxdy / (S)] based on the height matrix; calculate the autocorrelation function by the height fluctuation data in the horizontal direction, and then fit it using the exponential function R(u)=σ²e^(-u / Lc). The correlation length Lc is obtained by back-calculating the decay rate of the fitted curve; where S is the total area of the AFM scan, u is the horizontal displacement, and R(u) is the autocorrelation value at displacement u. S230, by perturbing Δ with dielectric constant (r)= The coupling rate of the far-field radiation mode is calculated using the volume current method, denoted as 0(n_a-Si² - n_SiO2²)·δ(r), and then expressed as γscat=ω∫∣Sm∣²sinθdθd. The scattering loss coefficient αscat = (ng·γscat) / c is obtained by conversion; where δ(r) is the surface profile function.
[0030] Optional, also includes: Based on the Monte Carlo method, the sidewall roughness fside(z) and top surface roughness ftop(x,y) of the target sample are generated according to the exponential autocorrelation function. Based on the target roughness fside(z) of the sidewall, the target roughness ftop(x,y) of the top surface, and the third error coefficient, the target roughness range is obtained.
[0031] The scattering loss model is constructed to obtain the target scattering loss coefficient αscat, specifically including: (1) Roughness parameter extraction The sidewalls and top surface of the amorphous silicon waveguide were scanned with high precision using atomic force microscopy (AFM). The scanning area was set to ≥5×5 μm² and the resolution to ≥256×256 pixels to obtain the height matrix h(x,y) containing information on the surface micro-undulations. Based on this height matrix, key parameters were calculated. Root mean square roughness (σ): Reflects the overall severity of surface undulations. It is calculated using the formula σ = √[∫∫h²(x,y)dxdy / (S)], where S = horizontal scanning length × vertical scanning length. The integration process covers the entire scanning area. The roughness index is obtained by averaging the squared height values of all coordinate points and then taking the square root. Correlation length (Lc): describes the spatial correlation of surface undulations (i.e. the influence range of adjacent undulation features). It is calculated by using the autocorrelation function of the horizontal height undulation data and then fitting it with the exponential function R(u)=σ²e^(-u / Lc). The correlation length Lc is obtained by back-calculating the decay rate of the fitted curve. This parameter directly determines the spatial distribution characteristics of the three-dimensional rough surface.
[0032] (2) Generation of three-dimensional rough surfaces (physical model construction) Based on the root mean square roughness σ and correlation length Lc extracted in the above steps, three-dimensional rough surface models of the waveguide sidewalls and top surface are generated using the Monte Carlo stochastic simulation method: Sidewall roughness model fside(z): z is the thickness direction coordinate of the waveguide. According to the exponential autocorrelation function R(u)=σ²e^(-u / Lc), the transverse undulation amount that conforms to the correlation is randomly generated at different positions in the z direction to ensure that the spatial distribution of the sidewall micro profile is consistent with the actual AFM test results.
[0033] The top surface roughness model ftop (x,y) is based on the waveguide length direction coordinate and the waveguide width direction coordinate. It generates a randomly undulating top surface profile in the xy plane, which simulates the micro-uniformity of the top surface caused by etching and deposition processes in actual manufacturing.
[0034] (3) Calculation of scattering loss (3-1) Dielectric constant perturbation analysis (scattering source modeling) The dielectric constant of an ideal smooth waveguide is uniformly distributed. However, the rough surfaces in reality cause irregular fluctuations at the interface between the waveguide and the surrounding medium (such as the substrate SiO2 or air), leading to local perturbations in the dielectric constant. The dielectric constant perturbation function Δ is defined as follows. (r)= 0(n_a-Si² - n_SiO2²) δ(r), where r is the position vector of any point in three-dimensional space. 0 represents the vacuum dielectric constant; n_a-Si represents the refractive index of amorphous silicon, and n_SiO2 represents the refractive index of the substrate (or surrounding medium). The difference between the two is the fundamental source of the dielectric constant; δ(r) is the surface profile function, used to identify the spatial location of a rough surface: when r is located at the rough interface, δ(r) = 1, at which point the dielectric constant changes abruptly; when r is far from the interface, δ(r) = 0, and the dielectric constant remains uniform. This function accurately describes the local variation of the dielectric constant caused by the rough surface, i.e., the spatial distribution of the scattering source.
[0035] (3-2) Volume Current Method (3D-VCM) The volume current method is a key numerical method for solving electromagnetic scattering on rough surfaces. Its core function is to transform the scattering problem caused by dielectric constant perturbation into a radiation problem of equivalent current. By calculating the far-field radiated power of the equivalent current, the magnitude of scattering loss is quantified. The specific steps are as follows: 1) Construction of equivalent current According to Maxwell's equations, the dielectric constant perturbation Δ The disturbance J(r) causes a distortion in the electric field distribution within the waveguide, which can be equivalent to the existence of a "volume current source" J(r) in the disturbance region. The expression for this equivalent current is derived from Maxwell's equations: J(r) = jωΔ E0(r) represents the incident electric field distribution in an ideal smooth waveguide without scattering, where j is the imaginary unit, ω is the angular frequency of the incident light (ω = 2πc / λ), c is the speed of light in vacuum, λ is the wavelength of the incident light, and E0(r) is the incident electric field distribution without scattering. Furthermore, the incident electric field distribution can be obtained through mode solving, such as the fundamental mode TE / TM mode electric field of an amorphous silicon waveguide. The volume current method, through this equivalent transformation, converts the complex interface scattering problem into an easily computable equivalent current radiation problem, avoiding the mathematical difficulties of directly solving for electromagnetic scattering at irregular interfaces.
[0036] 2) Calculation of far-field radiation mode coupling rate Based on the equivalent current J(r), the radiation field distribution generated by this current in the far-field region is calculated using electromagnetic radiation theory. The far-field radiation mode coupling rate Sm is defined, which physically represents the efficiency with which the equivalent current couples incident light energy to scattering modes (i.e., radiation modes detached from the waveguide propagation direction). This is expressed by the integral formula Sm = ∫J(r). The volume current method is calculated as E_m^(r) dV, where E_m^(r) is the conjugate of the electric field of the m-th scattering mode in the far field, and the integration region dV covers the entire dielectric perturbation region. The core advantage of the volume current method is that it can quantify the energy distribution of scattering in different directions and modes by accurately integrating the interaction between the equivalent current and the electric field of the scattering mode.
[0037] 3) Derivation of scattering loss rate γscat The scattering loss rate γscat represents the proportion of energy loss per unit length of waveguide due to scattering, and its magnitude is equal to the ratio of the sum of the radiated power of all far-field scattering modes to the incident power. Based on the far-field radiation mode coupling rate Sm obtained from the above steps, the possible radiation directions in the far field (θ being the azimuth angle, ...) are obtained by integrating. The elevation angle covers 0 ≤ θ ≤ π and 0 ≤ π. For the entire spatial range ≤2π, the scattering loss rate is obtained as: γscat = ω∫∣Sm∣²sinθdθd This integration process transforms the local equivalent current radiation effect into the total scattered energy across the entire space using the volume current method, thus achieving a quantitative calculation of scattering loss.
[0038] 4) Scat conversion of scattering loss coefficient αscat The scattering loss rate γscat is converted to the loss coefficient αscat, which is commonly used in engineering, with units of dB / cm. Based on the relationship between the group refractive index ng of the waveguide (reflecting the speed of light in the waveguide) and the speed of light c in vacuum, the conversion formula is derived: αscat = (ng The physical meaning of this formula is to convert the energy loss ratio per unit length (γscat) into an intuitive attenuation coefficient, taking into account the actual transmission characteristics of light in a waveguide, so as to facilitate comparison and verification with experimental test results.
[0039] The volume current method plays a "bridging role" in the calculation of scattering loss—by equating the dielectric perturbation of a rough surface to a volume current, the complex electromagnetic scattering problem is transformed into a quantifiable equivalent current radiation calculation, and finally, through integral solution and unit conversion, the accurate scattering loss coefficient αscat is obtained.
[0040] In one specific embodiment, the AFM test data includes: scanning the waveguide sidewalls and top surface to obtain σside = 4.2 nm and σtop = 2.8 nm, and fitting an autocorrelation curve to obtain a correlation length Lc = 300 nm. Based on the Monte Carlo method, a random surface conforming to R(u) = σ²e^(-u / Lc) is generated with a meshing accuracy of 50 nm. Dielectric constant. _a-Si=11.7 0, SiO2=3.9 0, the scattering loss coefficient αscat is calculated to be 0.62dB / cm.
[0041] Based on the above embodiments, the loss prediction method is verified experimentally. The specific steps of the experimental verification include: (1) Samples were prepared according to the standardized process. Amorphous silicon thin films were deposited using PECVD. Waveguide structures were formed by reactive ion etching (RIE) to control the sidewall roughness σside < 5 nm. The deposition process parameters for PECVD deposition of amorphous silicon were: substrate temperature 300℃, SiH4 / H2 flow ratio 1:5, SiH4 flow rate 20 sccm, H2 flow rate 100 sccm, RF power 200 W, and amorphous silicon thin film thickness 500 nm. The RIE etching equipment used CF4 / O2 (flow ratio 4:1) as the etching gas, with an etching power of 80 W and an etching time of 60 s. The final measured sidewall roughness σside = 4.0 nm.
[0042] The target sample prepared in this embodiment is an amorphous silicon strip waveguide device in the 1550nm communication band. The waveguide dimensions are set as follows: width W=2μm, thickness H=500nm, length L=1mm. The substrate is SiO2 material. The target total loss of the waveguide is ≤1dB / cm (absorption loss ≤0.3dB / cm, scattering loss ≤0.7dB / cm).
[0043] (2) Photothermal response test: Using a 1550nm tunable laser, a high-precision thermocouple and a spectrometer, the input power was scanned from 100 to 1000μW, and the redshift of the resonant wavelength Δλ and the temperature rise ΔT were recorded to verify the linear relationship between Δλ and ΔT.
[0044] Specifically, the output of a 1550nm tunable laser is coupled to the input of a waveguide via an optical fiber, a spectrometer is connected to the output of the waveguide, and a thermocouple is attached to the center of the waveguide surface. The laser's output is gradually increased from Pin = 100μW to 1000μW, with each step stabilizing for 30 seconds. The resonant wavelength λ and temperature T are then recorded. The slope of the Δλ-Pin curve is 2.2GHz / mW. The experimental αabs,exp = 0.27dB / cm is calculated, with a relative deviation of 3.6% (≤10%) from the simulated value.
[0045] (3) Roughness and scattering characteristics: The waveguide surface was scanned by AFM in a 5×5μm² area at a resolution of 256×256 pixels. The root mean square roughness (σ) and correlation length (Lc) were extracted, and the scattering loss γscat was tested in combination with the transmission spectrum of the ring resonator. The accuracy of the testing equipment used is as follows: thermocouple temperature measurement accuracy ±0.1K, spectrometer resolution 1pm, and AFM scanning resolution ≥256×256 pixels.
[0046] Specifically, the scanning area for AFM testing was 5×5μm², with a resolution of 256×256 pixels. The scan was repeated 3 times and the average value was taken, resulting in σside=4.0nm and σtop=2.7nm, with an error of ≤0.2nm compared to the model input parameters.
[0047] (4) Consistency verification: Set the criteria of relative deviation of absorption loss ≤10%, relative deviation of scattering loss ≤15%, and roughness parameter error ±1nm to verify the consistency between theoretical model and experimental results. That is, the first error coefficient is 10%, the second error coefficient is 15%, and the third error coefficient is ±1nm.
[0048] Furthermore, universality verification was conducted. Specifically, the material parameters were adjusted as follows: silicon nitride refractive index n=2.0, thermal conductivity κ=3.0W / (m·K), and thermo-optic coefficient dn / dT=2.5×10^-5K^-1. Using the method of this invention, the predicted silicon nitride waveguide αabs=0.12dB / cm and αscat=0.35dB / cm were obtained; experimentally measured αabs,exp=0.13dB / cm and αscat,exp=0.38dB / cm, with relative deviations of 8.3% and 8.6% respectively, satisfying the consistency criterion and verifying the universality of the method.
[0049] The above experiments verified that the error between the predicted loss and the actual loss of the amorphous silicon photonic device obtained by the loss prediction method proposed in this embodiment of the invention is within a preset range, meeting the consistency verification requirements. This solves the problems of low loss quantification accuracy and high experimental costs in existing technologies; it achieves the effect of reducing experimental costs, accurately predicting device loss, and guiding photonic chip design.
[0050] Furthermore, the method provided in this embodiment of the invention can be extended to other photonic device materials such as silicon nitride and indium phosphide. By adjusting the optical constants (n,k), thermal conductivity, and roughness characteristics of the material, loss prediction and verification of the corresponding material device can be achieved.
[0051] Example 2 like Figure 2 As shown, the present invention also provides a method for optimizing the loss of amorphous silicon photonic devices, comprising: T100. Set initial parameters, prepare the sample, and obtain the sample to be tested; substitute the initial parameters into the loss prediction method for amorphous silicon photonic devices described in any one of Examples 1 to obtain the corresponding target total loss range; T200, perform photothermal response testing on the sample to be tested to obtain the total measurement loss of the sample; T300: Determine whether the measured total loss exceeds the target total loss range; if not, the sample to be tested meets the standard; if so, perform atomic force microscopy testing and transmission spectrum testing using an atomic force microscope to obtain the test scattering loss coefficient, sidewall test roughness, and top surface test roughness; obtain the test absorption loss coefficient based on the measured total loss and the test scattering loss coefficient, and execute T400. T400: Optimize the RIE etching process during sample preparation based on the measured absorption loss coefficient, measured scattering loss coefficient, sidewall roughness, and top surface roughness to obtain the optimized sample to be tested. Return to T200 for testing, and repeat until the sample meets the standards.
[0052] Since scattering loss is related to the surface roughness of the sample, optimizing the etching process includes: controlling the sidewall roughness by adjusting the etching power and O2 ratio during reactive ion etching (RIE); for example, reducing the RIE etching power, increasing the O2 ratio used in RIE etching, and extending the RIE etching time.
[0053] When the difference between the total measurement loss of the sample under test and the target total loss obtained by the prediction model is within the preset range, the prepared sample can be considered qualified; if the difference is greater than the preset range, the prepared sample is considered unqualified, the RIE etching process is optimized, a new sample is prepared again, and the test is performed again until the sample under test meets the standard; the preparation parameters when the sample under test is qualified are recorded, and the preparation is carried out according to these parameters.
[0054] Based on the above embodiments, exemplarily, the initial sample preparation and testing showed an absorption loss αabs,exp = 0.42 dB / cm, exceeding the predicted value of 0.14 dB / cm. AFM detection revealed σside = 6.0 nm. This exceeded the target total loss range, necessitating adjustments to the RIE etching process. The etching power was reduced from 100 W to 80 W, the O2 ratio was increased from 1:4 to 1:1, and the etching time was extended to 70 s. After optimization, the sample was prepared again. The second prepared sample had a sidewall roughness σside = 3.8 nm, αabs,exp = 0.26 dB / cm, αscat,exp = 0.59 dB / cm, and a total loss of 0.85 dB / cm, meeting the target requirements.
[0055] This invention implements a loss test on the fabricated sample device during the fabrication process of amorphous silicon photonic devices, compares the actual loss with the predicted loss to determine whether the device is qualified, and optimizes the fabrication process if it is not qualified. At the same time, the scattering loss model can be used to determine the parameters that need to be optimized, guide the optimization process of the fabrication process, set strict consistency criteria, and form a closed loop of theory-experiment-process optimization. It has the advantages of strong process guidance and wide applicability, and can be extended to materials such as silicon nitride and indium phosphide, thereby guiding the design and process optimization of photonic chips.
[0056] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for predicting the loss of amorphous silicon photonic devices, characterized in that, include: S100. Set the target sample parameters of the amorphous silicon photonic device. Based on the thermal-optical coupling theory, correlate the temperature rise caused by the absorbed light power of amorphous silicon with the resonant frequency shift. Combine the heat conduction equation to solve for the target absorption loss coefficient. S200. Obtain the surface parameters of the target sample by scanning with an atomic force microscope, calculate the root mean square roughness and correlation length from the surface parameters, construct a three-dimensional rough surface model of the target sample including sidewall roughness and top surface roughness using the Monte Carlo random simulation method, and calculate the target scattering loss coefficient using the volume current method. S300. Based on the target absorption loss coefficient and the first error coefficient, the target absorption loss range is obtained; Based on the target scattering loss coefficient and the second error coefficient, the target scattering loss range is obtained; S400. The total loss range of the target is obtained based on the target absorption loss range and the target scattering loss range.
2. The method as described in claim 1, characterized in that, Step S100 specifically includes: S110. Set the target sample parameters for the amorphous silicon photonic device, establish a three-dimensional waveguide model using COMSOL, and set the simulation parameters. S120. Apply a preset optical power density input to the target sample, simulate and solve the steady-state temperature field, and extract the average temperature rise ΔT; calculate the refractive index change Δn=(dn / dT)·ΔT based on the average temperature rise ΔT, thereby obtaining the resonant frequency shift Δfres = (ng / c)Δn; where ng is the group refractive index, c is the speed of light in vacuum, and dn / dT is the thermo-optic coefficient of amorphous silicon; S130. By fitting the linear slope of Δfres-Pabs, the target absorption loss coefficient αabs is calculated using αabs = γabs·ng / c, where γabs is the thermal-optical coupling coefficient.
3. The method as described in claim 2, characterized in that, Step S200 specifically includes: S210. The sidewalls and top surface of the target sample are scanned using an atomic force microscope to obtain the surface parameters of the target sample. The surface parameters are a height matrix h(x,y) containing information on the surface micro-undulations; where x and y are the horizontal and vertical coordinates in the scanning plane, respectively, and h is the height value of the corresponding coordinate point. S220. Calculate the root mean square roughness σ = √[∫∫h²(x,y)dxdy / (S)] based on the height matrix; calculate the autocorrelation function by the height fluctuation data in the horizontal direction, and then fit it using the exponential function R(u)=σ²e^(-u / Lc). The correlation length Lc is obtained by back-calculating the decay rate of the fitted curve; where S is the total area of the AFM scan, u is the horizontal displacement, and R(u) is the autocorrelation value at displacement u. S230, by perturbing Δ with dielectric constant (r)= The coupling rate of the far-field radiation mode is calculated using the volume current method, denoted as 0(n_a-Si² - n_SiO2²)·δ(r), and then expressed as γscat=ω∫∣Sm∣²sinθdθd. The scattering loss coefficient αscat = (ng·γscat) / c is obtained by conversion; where δ(r) is the surface profile function.
4. The method as described in claim 3, characterized in that, Also includes: Based on the Monte Carlo method, the sidewall roughness fside(z) and top surface roughness ftop(x,y) of the target sample are generated according to the exponential autocorrelation function. Based on the target roughness fside(z) of the sidewall, the target roughness ftop(x,y) of the top surface, and the third error coefficient, the target roughness range is obtained.
5. The method as described in claim 4, characterized in that, The first error coefficient is 10%, the second error coefficient is 15%, and the third error coefficient is ±1nm.
6. A method for optimizing the loss of an amorphous silicon photonic device, characterized in that, include: T100: Set initial parameters, prepare the sample, and obtain the sample to be tested; Substituting the initial parameters into the loss prediction method for amorphous silicon photonic devices as described in any one of claims 1-5, the corresponding target total loss range is obtained; T200, perform photothermal response testing on the sample to be tested to obtain the total measurement loss of the sample; T300: Determine whether the measured total loss exceeds the target total loss range; if not, the sample to be tested meets the standard; if so, perform atomic force microscopy testing and transmission spectrum testing using an atomic force microscope to obtain the test scattering loss coefficient, sidewall test roughness, and top surface test roughness; obtain the test absorption loss coefficient based on the measured total loss and the test scattering loss coefficient, and execute T400. T400: Optimize the RIE etching process during sample preparation based on the measured absorption loss coefficient, measured scattering loss coefficient, sidewall roughness, and top surface roughness to obtain the optimized sample to be tested. Return to T200 for testing, and repeat until the sample meets the standards.
7. The method as described in claim 6, characterized in that, The optimization of the RIE etching process during sample preparation includes: Reduce the power of RIE etching, increase the proportion of O2 used in RIE etching, and extend the RIE etching time.