Electro-optic modulator, modulation method and manufacturing method

Abstract

The application discloses an electro-optical modulator, a modulation method and a preparation method. The electro-optical modulator comprises a silicon substrate, a silicon dioxide insulating layer, an input waveguide, an input beam splitter, an output coupler, an output waveguide, waveguide arms and a silicon dioxide-graphene composite structure arranged on both sides of the waveguide arms. The waveguide arms are in two groups, and the two groups of waveguide arms are symmetrically arranged. Each group of waveguide arms is composed of a curved segment and a straight segment. The silicon dioxide-graphene composite structure comprises a silicon dioxide modulation arm and a graphene layer. The silicon dioxide modulation arm is symmetrically arranged with the straight segment. The graphene layer is arranged on the top plane of the silicon dioxide modulation arm and the side surface facing the waveguide arm, and forms a three-dimensional coupling structure which is coupled with the evanescent field around the waveguide arm. The electro-optical modulator can realize high-speed and low-loss conversion of electrical signals to optical signals, thereby improving the transmission quality and transmission efficiency of the optical communication system.

Description

Technical Field

[0001] This invention relates to the field of integrated photonics, specifically to an electro-optic modulator, modulation method, and fabrication method. Background Technology

[0002] As optical communication technology iterates towards ultra-high speeds of 400G, 800G, and even 1.6T and 3.2T, the performance requirements for core components in optical communication systems are becoming increasingly stringent. Electro-optic modulators, as key components for converting electrical signals to optical signals, directly determine the speed, power consumption, integration, and transmission quality of the optical transmission system. In ultra-high-speed transmission scenarios, electro-optic modulators must possess core performance characteristics such as high modulation sensitivity, high optical field utilization, wide modulation bandwidth, and low power consumption to meet the demands of high-capacity, high-speed signal transmission and adapt to the application requirements of emerging fields such as AI data centers and 6G communications.

[0003] However, existing electro-optic modulators still suffer from numerous performance bottlenecks in practical applications, making it difficult to fully meet the stringent requirements of ultra-high-speed optical communication systems. Existing solutions often employ a single-sided structure with graphene on one side of the waveguide, which can only couple with a very small area of ​​the evanescent field on the waveguide surface, resulting in a coupling overlap generally less than 40% and an optical field energy utilization rate of less than 50%. Higher driving voltages are required for effective modulation, increasing device power consumption and limiting the modulation rate. Furthermore, existing algorithm compensation schemes are mostly pre-distortion compensation for general silicon-based modulators, failing to address the unique errors arising from the asynchronous carrier responses of the top and side graphene layers and the uneven three-dimensional distribution of the evanescent field in three-dimensional coupled structures, thus hindering high-precision and stable modulation in ultra-high-speed scenarios. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide an electro-optic modulator that can realize high-speed, low-loss conversion of electrical signals to optical signals, thereby improving the transmission quality and transmission efficiency of optical communication systems.

[0005] A second objective of the present invention is to provide a modulation method for the electro-optic modulator.

[0006] The third objective of this invention is to provide a method for preparing an electro-optic modulator.

[0007] The technical solution of the present invention to solve the above-mentioned technical problems is:

[0008] An electro-optic modulator includes a silicon substrate, a silicon dioxide insulating layer disposed on the silicon substrate, an input waveguide disposed on the silicon dioxide insulating layer, an input beamsplitter, an output coupler, an output waveguide, a waveguide arm disposed between the input beamsplitter and the output coupler, and silicon dioxide-graphene composite structures respectively disposed on both sides of the waveguide arm. The waveguide arm is in two sets, symmetrically arranged, each set consisting of a curved segment and a straight segment. The silicon dioxide-graphene composite structure includes a silicon dioxide modulation arm and a graphene layer disposed on the silicon dioxide modulation arm. The silicon dioxide modulation arm is symmetrically arranged with the straight segment. The graphene layer is disposed on the top plane of the silicon dioxide modulation arm and on the inner side facing the waveguide arm, forming a three-dimensional coupling structure that is three-dimensionally coupled to the evanescent field surrounding the waveguide arm.

[0009] Preferably, a hexagonal boron nitride insulating layer is disposed on the outer side of the silicon dioxide modulation arm away from the waveguide arm, and the thickness of the hexagonal boron nitride insulating layer is 20-50 nm.

[0010] Preferably, the vertical cross-section of the silicon dioxide modulation arm is rectangular, the top plane of the silicon dioxide modulation arm is symmetrically arranged with the silicon dioxide insulating layer, and the side is perpendicular to the silicon dioxide insulating layer.

[0011] Preferably, the thickness of the silicon dioxide modulation arm is 200-300 nm.

[0012] A modulation method for an electro-optic modulator includes the following steps:

[0013] S1: Input a first single-mode optical signal in the 1550nm band into the input waveguide of the electro-optic modulator. After the first single-mode optical signal enters the input waveguide, it is split into two second single-mode optical signals. The two second single-mode optical signals are transmitted along two sets of symmetrically arranged waveguide arms respectively. During the transmission process, each second single-mode optical signal will form an evanescent wave of the optical field around the corresponding waveguide arm.

[0014] S2: Based on the silica-graphene composite structure on both sides of the waveguide arm, the evanescent wave of the optical field penetrates to the surface of the silica-graphene composite structure and undergoes near-field coupling with the graphene layer in the silica-graphene composite structure facing the waveguide arm, thereby realizing the efficient interaction between the optical field and the graphene layer.

[0015] S3: Build an LSTM timing error compensation model adapted to the three-dimensional coupling structure. The model is trained using the system's measured three-dimensional coupled optical field timing sample dataset. The model is then iteratively optimized using a weighted mean square loss function to obtain an LSTM timing error compensation model that adapts to the electro-optic modulator error correction and meets the modulation accuracy requirements.

[0016] S4: Real-time acquisition of driving electrical signal parameters, environmental parameters and optical feedback parameters, input of multi-source timing parameters into the trained LSTM timing error compensation model, correction of nonlinear error and timing drift through model calculation, output of two differential driving electrical signals with voltage correction and frequency fine-tuning;

[0017] S5: The differential driving electrical signal is applied to the graphene layer in the silica-graphene composite structure on the outside of the two sets of waveguide arms by means of preset metal electrodes. The graphene carrier concentration and Fermi level are changed by electrical modulation. The absorption and refractive index characteristics of the optical field are changed by means of the electro-optic effect of graphene, so that the optical signals transmitted in the two sets of waveguide arms generate differential phase / intensity modulation, thereby achieving high-precision and stable electro-optic modulation.

[0018] Preferably, in step S3, the training process of the LSTM time-series error compensation model is as follows:

[0019] S301: Construct an LSTM temporal error compensation model, which includes an input layer, a bidirectional LSTM hidden layer, a three-dimensional light field feature attention layer, a fully connected layer, and an output layer; wherein, the number of forward hidden units and backward hidden units in the bidirectional LSTM hidden layer are 32-64 respectively; the three-dimensional light field feature attention layer adopts an additive attention mechanism to assign adaptive weights to the three-dimensional coupled optical feedback parameters; the number of neurons in the fully connected layer is 8-16, and the ReLU activation function is used;

[0020] S302: Select the actual measured time-series sample dataset of the system as training data, use the driving electrical signal parameters, environmental parameters, and optical feedback parameters as model input parameters, and use the electrical signal voltage correction and frequency fine-tuning corresponding to the model input parameters as training labels; during the training process, the weighted mean square error loss function is used to measure the deviation between the predicted value and the label value, and the Adam optimizer is selected and combined with the cosine annealing learning rate scheduling strategy and early stopping mechanism to iteratively train the LSTM time-series error compensation model until the prediction accuracy of the LSTM time-series error compensation model meets the preset modulation requirements, and the trained LSTM time-series error compensation model is obtained.

[0021] Preferably, in step S302, the driving electrical signal includes voltage, frequency, and phase; the environmental parameters include temperature and humidity; and the optical feedback parameters include absorption coefficient and refractive index feedback values.

[0022] A method for fabricating the electro-optic modulator includes the following steps:

[0023] Step 1: Select a single-crystal silicon wafer of a specified thickness as the silicon substrate. Use the RCA standard cleaning method to remove organic matter, metal impurities and natural oxide layer from the surface of the silicon substrate. After drying, it is ready for use. Use plasma-enhanced chemical vapor deposition technology to deposit a silicon dioxide insulating layer on the surface of the silicon substrate. By controlling the deposition temperature and radio frequency power, the thickness of the silicon dioxide insulating layer is made to reach the preset value.

[0024] Step 2: Photoresist is coated on the surface of the silicon dioxide insulating layer using electron beam lithography. After exposure and development, waveguide patterns are formed, including an input waveguide, an input beam splitter, two sets of symmetrically arranged waveguide arms, an output coupler, and an output waveguide. Using the photoresist as a mask, waveguide structures corresponding to the waveguide patterns are etched on the silicon dioxide insulating layer by dry etching, ensuring that the width of the waveguide arms meets the preset width range and the spacing between the two sets of symmetrically arranged waveguide arms meets the first preset spacing range. After etching, the residual photoresist is removed.

[0025] Step 3: Deposit silicon dioxide material on the surface of the silicon dioxide insulating layer on both sides of the waveguide arm using plasma-enhanced chemical vapor deposition (PECVD) to construct a silicon dioxide modulation arm; use dry etching to trim the edges of the silicon dioxide modulation arm so that the side of the silicon dioxide modulation arm is perpendicular to the silicon dioxide insulating layer, and the top plane is parallel to the silicon dioxide insulating layer, and ensure that the distance between the inner side of the silicon dioxide modulation arm facing the waveguide arm and the side of the waveguide arm facing the silicon dioxide modulation arm conforms to a second preset distance range, and the length of the silicon dioxide modulation arm is consistent with the length of the straight segment of the waveguide arm;

[0026] Step 4: Prepare monolayer graphene by mechanical exfoliation. Using PMMA-assisted transfer process, first perform oxygen plasma activation treatment on the inner side of the silicon dioxide modulation arm, then attach and fix the defect-free monolayer graphene to the inner side of the silicon dioxide modulation arm. After constant temperature baking, adhesive removal, cleaning, and nitrogen drying, the preparation and integration of the graphene layer on the inner side of the silicon dioxide modulation arm is completed.

[0027] Step 5: Using the same graphene layer preparation process as in Step 4, fabricate a graphene-PMMA composite film. Transfer the graphene-PMMA composite film to the top plane of the silica modulation arm. After desizing, cleaning, and morphology inspection, the preparation of the graphene layer on the top plane of the silica modulation arm is completed.

[0028] Step 6: Obtain hBN sheets using mechanical exfoliation. Apply the hBN sheets to the outer side of the silicon dioxide modulation arm away from the waveguide arm using dry transfer technology, serving as an insulating protection and carrier modulation layer for the graphene layer. Deposit Cr / Au metal electrodes on the surface of the hBN sheets using electron beam evaporation technology. One end of the Cr / Au metal electrode is in ohmic contact with the graphene layer, while the other end extends to the edge of the electro-optic modulator for connecting an external driving electrical signal source.

[0029] Preferably, step 4 consists of the following steps:

[0030] Step 401: Repeatedly peel off the highly oriented pyrolytic graphite block with transparent adhesive to obtain a single-layer graphene sheet; screen the graphene sheets that are wrinkle-free, pore-free, and large enough to completely cover the inner side of the silica modulation arm using an optical microscope.

[0031] Step 402: Place the screened single-layer graphene sheet in a PMMA solution and bake it on a heating table to make PMMA evenly cover the surface of the single-layer graphene sheet to form a graphene-PMMA composite film; then peel the graphene-PMMA composite film off the transparent adhesive residue layer and soak it in deionized water to completely remove the bottom adhesive tape residue.

[0032] Step 403: Using a micro-alignment system, precisely align the graphene side of the graphene-PMMA composite film with the inner side of the silicon dioxide modulation arm near the waveguide arm. Use a micromanipulator to press and adhere the film, ensuring that the graphene completely covers the entire area of ​​the inner side of the silicon dioxide modulation arm. After transfer, place the entire sample on a heating stage for baking, so that the graphene and the inner side of the silicon dioxide modulation arm are tightly bonded together by van der Waals forces.

[0033] Step 404: Immerse the sample in acetone solution for ultrasonic treatment to dissolve and completely remove the PMMA support layer; then rinse the sample surface multiple times with isopropanol and dry it with nitrogen gas to complete the preparation of the graphene layer on the inner side of the silicon dioxide modulation arm.

[0034] Preferably, step 5 consists of the following steps:

[0035] Step 501: Repeat steps 401 and 402 to prepare another graphene-PMMA composite film whose size is completely matched with the top plane size of the silicon dioxide modulation arm, and ensure that the graphene-PMMA composite film is free from structural defects such as cracks and wrinkles.

[0036] Step 502: Using a micro-alignment system, the graphene side of the graphene-PMMA composite film is placed downwards and precisely covers the top plane of the silicon dioxide modulation arm. This ensures that the edges of the graphene in the graphene-PMMA composite film are aligned with the edges of the top plane of the silicon dioxide modulation arm without any offset. A heating stage is then used for heat drying and leveling to eliminate air bubbles at the contact interface and strengthen the bonding force between the graphene and the top plane of the silicon dioxide modulation arm.

[0037] Step 503: Immerse the sample in acetone solution and let it stand for a while until the PMMA is completely dissolved and removed. Rinse it several times with deionized water and dry it with nitrogen. Perform microscopic characterization and detection using atomic force microscopy to ensure that the roughness of the graphene surface meets the preset requirements and that there is no residual solvent or impurities. This completes the preparation of the graphene layer on the top plane of the silica modulation arm.

[0038] Compared with the prior art, the present invention has the following advantages:

[0039] Graphene, as a novel two-dimensional material, possesses ultra-high-speed carrier response characteristics. Combined with the low-loss and excellent insulation properties of silicon dioxide, this invention effectively expands the operational range of the electro-optic modulator and optical field by constructing a silicon dioxide-graphene composite structure on both sides of the waveguide arm, thereby improving the utilization rate of the optical field. Specifically, by applying a driving electrical signal to the silicon dioxide-graphene composite structure on both sides of the waveguide arm, the Fermi level of graphene can be precisely controlled, thereby causing controllable changes in the absorption coefficient and refractive index of graphene. After the optical signal is split by the input beam splitter, it enters two symmetrically arranged waveguide arms, generating efficient optical coupling with the graphene layers in the silicon dioxide-graphene composite structure on both sides of the waveguide arm. This results in a stable phase difference between the two optical signals. These two optical signals with phase difference are then interfered by the output coupler to achieve amplitude modulation of the optical signal, ultimately achieving a high-speed, low-loss conversion from electrical signal to optical signal. This significantly improves the transmission quality and efficiency of the optical communication system, meeting the application requirements of ultra-high-speed optical communication. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the electro-optic modulator of the present invention.

[0041] Figure 2 and 3 These are schematic diagrams from two different perspectives of the waveguide arm and the silica-graphene composite structure on both sides.

[0042] Figure 4 This is a cross-sectional view of the waveguide arm and the silicon dioxide-graphene composite structure on both sides.

[0043] In the figure: 1-1, silicon substrate; 1-2, silicon dioxide insulating layer; 1-3, input waveguide; 1-4, waveguide arm; 1-5, silicon dioxide-graphene composite structure; 1-6, output waveguide; 1-7, graphene layer; 1-8, silicon dioxide modulation arm; 1-9, hBN insulating layer. Detailed Implementation

[0044] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0045] See Figures 1-4 The electro-optic modulator of the present invention includes a silicon substrate, a silicon dioxide insulating layer disposed on the silicon substrate, an input waveguide disposed on the silicon dioxide insulating layer, an input beam splitter, an output coupler, an output waveguide, a waveguide arm disposed between the input beam splitter and the output coupler, and silicon dioxide-graphene composite structures respectively disposed on both sides of the waveguide arm. The waveguide arm is in two sets, symmetrically arranged, and each set consists of curved segments and straight segments. The silicon dioxide-graphene composite structure includes a silicon dioxide modulation arm and a graphene layer disposed on the silicon dioxide modulation arm. The thickness of the silicon dioxide modulation arm is 200-300 nm, which provides support for the graphene layer and ensures effective coupling of the evanescent wave of the optical field, with an optical transmission loss of <0.5 dB / cm. The silicon dioxide modulation arm is symmetrically arranged with the straight segment. The graphene layer is disposed on the top plane of the silicon dioxide modulation arm and on the inner side facing the waveguide arm, forming a three-dimensional coupling structure that is coupled three-dimensionally with the evanescent field around the waveguide arm. The graphene layer consists of 1-3 layers, preferably a single layer, which ensures that the coupling overlap with the evanescent wave of the optical signal transmitted in the waveguide arm is >80%. Its absorption coefficient varies with electro-optic modulation by 0.1-0.3, and its refractive index varies with electro-optic modulation by 0.05-0.15. A hexagonal boron nitride (hBN) insulating layer is disposed on the outer side of the silicon dioxide modulation arm away from the waveguide arm. The thickness of the hexagonal boron nitride (hBN) insulating layer is 20-50 nm, which is used to ensure insulation performance and carrier modulation efficiency.

[0046] See Figures 1-4 The working principle of the electro-optic modulator of the present invention is as follows:

[0047] The electro-optic modulator of this invention uses a Mach-Zehnder interferometer (MZI) as its core structure. A single-mode optical signal in the 1550nm band is introduced through an input waveguide and then split into two beams by an input beamsplitter, which enter two symmetrically arranged waveguide arms. As the single-mode optical signal propagates within the waveguide arms, an evanescent wave naturally forms around it. Because both sides of the waveguide arms are equipped with silica-graphene composite structures, and the graphene layers in these structures face the waveguide arms, the evanescent wave can fully contact the graphene layers on both sides, achieving highly efficient coupling with an overlap of over 80%. This high coupling efficiency ensures that changes in the graphene optical feedback parameters effectively affect the single-mode optical signal, providing a guarantee for precise modulation. An external driving electrical signal is applied to the graphene layer through electrodes, triggering the electro-optic effect of the graphene. Specifically, the driving electrical signal changes the carrier distribution inside the graphene, thereby causing a change in its Fermi level. Furthermore, changes in the Fermi level will further increase the absorption coefficient of graphene. and refractive index The optical feedback parameters undergo periodic changes, and these changes are synchronized in real time with the changes in the electrical signal. Therefore, by applying a differential driving electrical signal to the silica-graphene composite structure on both sides of the two symmetrically arranged waveguide arms, the two optical signals can be modulated to different degrees during transmission. Changes in the absorption coefficient affect the intensity of the optical signal, while changes in the refractive index alter the propagation speed, ultimately creating a phase difference between the two optical signals. This phase difference changes linearly with the amplitude of the modulation signal, and the modulation sensitivity is... This effectively ensures the linearity and accuracy of the modulation. After the two differentially modulated optical signals enter the output coupler, coherent interference occurs. When the phase difference between the two beams is 0°, constructive interference occurs, and the output optical signal intensity reaches its maximum; when the phase difference is 180°, destructive interference occurs, and the output optical signal intensity drops to its minimum. By controlling the phase difference between the two beams to continuously vary between 0° and 180° using a driving electrical signal, amplitude modulation of the output optical signal can be achieved. This allows the information carried by the driving electrical signal to be loaded onto the optical signal, and the final modulated optical signal is output through the output waveguide.

[0048] See Figures 1-4 The fabrication process of the electro-optic modulator of the present invention is as follows:

[0049] Step 1: Preparation of Silicon Substrate and Insulating Layer: A 500μm thick single-crystal silicon wafer was selected as the silicon substrate. The organic matter, metallic impurities, and natural oxide layer on the surface of the silicon substrate were removed using the RCA standard cleaning method (ammonia-hydrogen peroxide-deionized water mixed solution, temperature 75℃). After drying, it was ready for use. A silicon dioxide insulating layer was deposited on the surface of the silicon substrate using plasma-enhanced chemical vapor deposition (PECVD) technology. The deposition temperature was controlled at 300℃ and the RF power at 150W, so that the thickness of the silicon dioxide insulating layer reached 250nm. This thickness can simultaneously meet the insulation and isolation requirements of the silicon substrate and the effective transmission of evanescent waves in the subsequent optical field. Preliminary tests showed that the optical transmission loss of this insulating layer system was <0.5dB / cm.

[0050] Step 2, Waveguide Structure Fabrication: Using electron beam lithography, photoresist (model AZ5214) is coated onto the surface of the silicon dioxide insulating layer. After exposure and development, the target pattern of "input waveguide - input beam splitter - parallel waveguide arm - output coupler - output waveguide" is formed. Using the photoresist as a mask, a dry etching process (reactive gas) is performed. and A complete waveguide structure is formed by etching on a silicon dioxide insulating layer with a flow ratio of 3:1. The width of the waveguide arm is controlled at 450nm (to adapt to 1550nm single-mode optical transmission), and the spacing between two symmetrically arranged waveguide arms is 2μm. After etching, the residual photoresist on the surface is removed.

[0051] Step 3: Fabrication of the silicon dioxide modulation arm: Using PECVD technology, silicon dioxide material is deposited on the surface of the silicon dioxide insulating layer on both sides of the waveguide arm to construct the silicon dioxide modulation arm; the deposition parameters are consistent with the aforementioned deposition parameters of the silicon dioxide insulating layer, and the deposition thickness is controlled at 280nm; the edges of the silicon dioxide modulation arm are trimmed using a dry etching process to give it a regular structure with vertical sides and a horizontal top, wherein the distance between the inner side of the silicon dioxide modulation arm facing the waveguide arm and the opposite side of the waveguide arm is 0.8μm, and the length of the top plane is completely consistent with the length of the straight segment of the waveguide arm.

[0052] Step 4, the preparation of the monolayer graphene on the side of the silica modulation arm, specifically:

[0053] Step 401: Repeatedly peel off the highly oriented pyrolytic graphite (HOPG) bulk (99.99% purity) with transparent tape (300 cP) to obtain a single-layer graphene sheet with a thickness of about 0.34 nm; screen the graphene sheets that are wrinkle-free, pore-free, and large enough to completely cover the sides of the silica modulation arm using an optical microscope.

[0054] Step 402: Place the screened single-layer graphene sheet in a 5% PMMA solution and bake it on a 60°C heating table for 10 minutes to make PMMA uniformly cover the surface of the graphene and form a graphene-PMMA composite film; then peel the graphene-PMMA composite film off the transparent adhesive residue layer and soak it in deionized water for 5 minutes to remove the bottom adhesive tape residue.

[0055] Step 403: Using a micro-alignment system, precisely align the graphene side of the graphene-PMMA composite film with the inner side of the silicon dioxide modulation arm near the waveguide arm. Slowly adhere the film using a micromanipulator to ensure that the graphene completely covers the target side area. After the transfer is complete, place the sample on a 120°C heating stage and bake for 30 minutes to allow the graphene and the inner side of the silicon dioxide modulation arm to bond tightly through van der Waals forces, thereby enhancing the interfacial adhesion.

[0056] Step 404: Immerse the sample in an acetone solution at 25°C and sonicate for 5 minutes (power 50W, frequency 40kHz to avoid damaging the graphene) to dissolve and remove the PMMA support layer; then rinse the sample surface twice with isopropanol and dry it with high-purity nitrogen (purity 99.99%) to complete the attachment of a single layer of graphene on the side of the silica modulation arm.

[0057] Step 5: Preparation of the monolayer graphene at the top of the silica modulation arm. The specific steps are as follows:

[0058] Step 501: Repeat steps 401 and 402 to prepare another graphene-PMMA composite film with dimensions matching the top plane of the silicon dioxide modulation arm, and ensure that the graphene in the graphene-PMMA composite film is free of any defects.

[0059] Step 502: Using a micro-alignment system, precisely cover the top plane of the silicon dioxide modulation arm with the graphene side of the graphene-PMMA composite film facing down, ensuring that the edges of the graphene are completely aligned with the edges of the top plane of the silicon dioxide modulation arm; use a vacuum press (vacuum degree...) Press at 0.1 MPa (pressure 0.1 MPa, temperature 80℃) for 10 minutes to remove interfacial bubbles (bubble rate <1%) and strengthen the bonding force between graphene and the top plane of the silicon dioxide modulation arm.

[0060] Step 503: Immerse the sample in acetone solution for 15 minutes until PMMA is completely dissolved, rinse three times with deionized water, and dry with high-purity nitrogen. Observe and detect with an atomic force microscope to ensure that the graphene surface is flat (roughness < 1 nm) and free of residual solvent or impurities, thus completing the preparation and attachment of the graphene layer on the top plane of the silicon dioxide modulation arm.

[0061] Step 6: Fabrication of hexagonal boron nitride insulating layer and metal electrode

[0062] A 35 nm thick hexagonal boron nitride (hBN) sheet was obtained by mechanical exfoliation. The hBN sheet was then coated onto the outer side of the silicon dioxide-graphene composite structure using a dry transfer technique, serving as an insulating and protective layer for the graphene layer and a carrier modulation layer. Cr / Au electrodes (5 nm / 100 nm thick) were deposited on the hBN layer surface using electron beam evaporation. One end of the Cr / Au electrode forms an ohmic contact with the graphene layer, while the other end extends to the edge of the electro-optic modulator of this invention for connecting an external driving electrical signal source.

[0063] After the electro-optic modulator of the present invention is fabricated, its performance needs to be verified, specifically as follows:

[0064] (1) Optical signal input and beam splitting: A single-mode continuous laser (power 10mW) in the 1550nm band is coupled to the input waveguide through an optical fiber. The single-mode continuous laser is evenly split into two beams after passing through the input beam splitter and enters two symmetrical waveguide arms respectively. When the optical signal is transmitted in the waveguide arm, an evanescent wave of optical field will be formed around the waveguide arm. The evanescent wave of optical field is fully coupled with the graphene layer in the silicon dioxide-graphene composite structure on both sides of the waveguide arm. The measured coupling overlap reaches 85%.

[0065] (2) Electrical signal modulation and optical feedback parameter changes: A differential driving electrical signal (voltage range 0-5V) is applied to the electrodes on both sides of the two waveguide arms by an external driving electrical signal source. This differential driving electrical signal causes a change in the carrier concentration in the graphene layer, resulting in the generation of the Fermi level in the graphene. offset (at) (Within the design range); the change in the absorption coefficient of graphene with the dynamic change of the Fermi level. (Within the design range of 0.1-0.3), Refractive index change (Within the design range of 0.05-0.15), and the changes in the above optical feedback parameters are kept in real-time synchronization with the frequency of the electrical signal (1-100GHz).

[0066] (3) Interference Modulation and Output: Two differentially modulated optical signals converge into the output coupler. Due to the difference in propagation speed caused by the change in refractive index, the two optical signals form a stable phase difference. When the phase difference between the two optical signals is 0°, constructive interference occurs, and the output optical power reaches 9.8mW. When the phase difference is 180°, destructive interference occurs, and the output optical power is <0.2mW. By adjusting the external driving electrical signal in real time, the amplitude modulation of the output optical signal can be achieved. The modulation sensitivity is tested to reach With a modulation rate of up to 50GHz, it can fully meet the application requirements of ultra-high-speed optical communication.

[0067] Example 2

[0068] The difference between this embodiment and Embodiment 1 is that, in the field of integrated photonics, efficiently loading driving electrical signals into optical signals to achieve ultra-high-speed modulation is a key issue facing optical communication systems. Existing electro-optic modulators not only suffer from structural problems such as insufficient modulation sensitivity and optical field utilization, but also have limitations in modulation accuracy and stability due to nonlinearity of electro-optic effects, carrier response delay, and environmental drift. To solve these problems, this invention also introduces the LSTM (Long Short-Term Memory) algorithm, which captures the nonlinear correlation between driving electrical signal parameters, environmental parameters, and optical feedback parameters through time-series modeling, and performs pre-distortion compensation on the input driving electrical signal to achieve high-speed, low-loss, and high-linearity conversion of electrical signals to optical signals, meeting the stringent requirements of ultra-high-speed optical communication.

[0069] The modulation method of the electro-optic modulator of the present invention includes the following steps:

[0070] S1: Input a first single-mode optical signal in the 1550nm band into the input waveguide of the electro-optic modulator. After the first single-mode optical signal enters the input waveguide, it is split into two second single-mode optical signals. The two second single-mode optical signals are transmitted along two sets of symmetrically arranged waveguide arms respectively. During the transmission process, each second single-mode optical signal will form an evanescent wave of the optical field around the corresponding waveguide arm.

[0071] S2: Based on the silica-graphene composite structure on both sides of the waveguide arm, the evanescent wave of the optical field penetrates to the surface of the silica-graphene composite structure and undergoes near-field coupling with the graphene layer in the silica-graphene composite structure facing the waveguide arm, thereby realizing the efficient interaction between the optical field and the graphene layer.

[0072] S3: Build an LSTM timing error compensation model adapted to the three-dimensional coupling structure. The model is trained using the system's measured three-dimensional coupled optical field timing sample dataset. The model is then iteratively optimized using a weighted mean square loss function to obtain an LSTM timing error compensation model that adapts to the electro-optic modulator error correction and meets the modulation accuracy requirements.

[0073] In this embodiment, the LSTM timing error compensation model adopts a lightweight bidirectional LSTM attention hybrid model, and the training process of this LSTM timing error compensation model is as follows:

[0074] S301: Construct an LSTM timing error compensation model, which sequentially includes an input layer with a feature dimension of 7 and a time step of 20, a bidirectional LSTM hidden layer, a three-dimensional light field feature attention layer, a fully connected layer, and an output layer; wherein, the number of forward hidden units and backward hidden units in the bidirectional LSTM hidden layer are 32-64, preferably 32, and the activation function is tanh; the three-dimensional light field feature attention layer adopts an additive attention mechanism to assign adaptive weights to the three-dimensional coupled optical feedback parameters; the number of neurons in the fully connected layer is 8-16, preferably 12, and the ReLU activation function is used; the output layer is used to output two differential driving electrical signals with voltage correction and frequency fine-tuning;

[0075] S302: Model training employs a transfer learning strategy combining COMSOL simulation data pre-training and experimental data fine-tuning to improve model generalization ability and training efficiency. Formal training uses a time-series sample dataset from actual system measurements as training data, with driving electrical signal parameters, environmental parameters, and optical feedback parameters as model input parameters. The corresponding electrical signal voltage correction and frequency fine-tuning values ​​are used as training labels. During training, a weighted mean square error (MSE) loss function is used to measure the deviation between predicted and labeled values, and the Adam optimizer (initial learning rate 0.0) is selected. 01) Combining the cosine annealing learning rate scheduling strategy and the early stopping mechanism, the LSTM timing error compensation model is iteratively trained until the prediction accuracy of the LSTM timing error compensation model meets the preset modulation requirements, thus obtaining the trained LSTM timing error compensation model; finally, the trained LSTM timing error compensation model is quantized into a 16-bit fixed-point number and deployed on a field-programmable gate array (FPGA), which can realize parallel inference of the model with an inference latency of <1ps, and can perform real-time pre-distortion compensation for the driving electrical signal of the electro-optic modulator, ensuring high precision and high stability of the modulation process.

[0076] During training, the driving electrical signal is acquired via a high-speed ADC sampler, and includes voltage (0-5V), frequency (1-100GHz), and phase (0-360°); the environmental parameters are acquired via a temperature and humidity sensor, and include temperature (-20-85℃) and humidity (10%-90%); the optical feedback parameters are acquired via a photoelectric sampler, and include the absorption coefficient. and refractive index feedback value In addition, in this embodiment, time-series training data with a frequency range of 1-100GHz and a temperature range of -20-85℃ are collected; the input electrical signal parameters, environmental parameters, and optical feedback parameters are used as the model input, and the electrical signal voltage correction and frequency fine-tuning are used as the model output to construct a positive compensation model; the model parameters are iteratively optimized using a weighted mean square error loss function. For extreme operating conditions with frequencies >80GHz and temperatures <-10℃ or >70℃, the weighting coefficients of the corresponding samples are set to 2-3 times to enhance the model fitting ability under extreme operating conditions and improve the modulation compensation accuracy under all operating conditions.

[0077] S4: Real-time acquisition of driving electrical signal parameters, environmental parameters and optical feedback parameters, input of multi-source timing parameters into the trained LSTM timing error compensation model, correction of nonlinear error and timing drift through model calculation, output of two differential driving electrical signals with voltage correction and frequency fine-tuning;

[0078] S5: The differential driving electrical signal is applied to the graphene layer in the silica-graphene composite structure on the outer side of the two sets of waveguide arms through electrodes. The graphene carrier concentration and Fermi level are changed by electro-modulation. The absorption and refractive index characteristics of the optical field are changed by the electro-optic effect of graphene, so that the optical signals transmitted in the two sets of waveguide arms generate differential phase / intensity modulation, thus completing the electro-optic modulation.

[0079] Finally, the electro-optic modulator in this embodiment employs a collaborative working mechanism of Mach-Zehnder interferometer (MZI) and LSTM algorithm intelligent compensation. The entire workflow is as follows:

[0080] A single-mode optical signal in the 1550nm band is introduced into the electro-optic modulator of this invention via an input waveguide, and then split into two paths by an input beam splitter, which enter two symmetrically arranged waveguide arms respectively. During the transmission of the optical signal within the waveguide arms, an evanescent wave of the optical field is formed around the waveguide arms, which forms an efficient coupling with the graphene layer in the silicon dioxide-graphene composite structure on both sides of the waveguide arms.

[0081] Simultaneously, real-time timing data of the driving electrical signal, environmental parameters, and optical feedback parameters are acquired. A pre-trained LSTM timing error compensation model is used to pre-distort the input driving electrical signal parameters. Addressing modulation deviations caused by the nonlinearity of the graphene electro-optic effect, carrier response delay, and ambient temperature drift, two optimized differentiated driving electrical signals are output. These optimized differentiated driving electrical signals are applied to the graphene layer through electrodes, triggering the electro-optic effect of graphene and causing a controllable change in the carrier distribution within the graphene. This allows for precise modulation of its Fermi level, ultimately enabling the absorption coefficient and refractive index of graphene to change synchronously and with high precision with the driving electrical signal. During the transmission of the two optical signals within the waveguide arm, the propagation speed changes differentially due to the dynamic changes in the graphene refractive index, resulting in a linearly controllable phase difference. This phase difference changes linearly with the amplitude of the optimized driving electrical signal, fully ensuring the accuracy of the modulation process.

[0082] The two differentially modulated optical signals are ultimately merged into the output coupler and undergo coherent interference: constructive interference occurs when the phase difference between the two optical signals is 0°, and the output optical power reaches its maximum value; destructive interference occurs when the phase difference is 180°, and the output optical power drops to its minimum value. By dynamically adjusting the driving electrical signal using the LSTM algorithm, continuous and precise control of the phase difference between the two optical signals within the range of 0°-180° can be achieved, ultimately outputting a high-linearity amplitude-modulated optical signal loaded with electrical signal information, completing the efficient electro-optic signal conversion.

[0083] The above are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above content. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. An electro-optic modulator, characterized in that, The device includes a silicon substrate, a silicon dioxide insulating layer disposed on the silicon substrate, an input waveguide disposed on the silicon dioxide insulating layer, an input beam splitter, an output coupler, an output waveguide, a waveguide arm disposed between the input beam splitter and the output coupler, and silicon dioxide-graphene composite structures respectively disposed on both sides of the waveguide arm. The waveguide arm is in two sets, symmetrically arranged, with each set consisting of a curved segment and a straight segment. The silicon dioxide-graphene composite structure includes a silicon dioxide modulation arm and a graphene layer disposed on the silicon dioxide modulation arm. The silicon dioxide modulation arm is symmetrically arranged with the straight segment. The graphene layer is disposed on the top plane of the silicon dioxide modulation arm and on the inner side facing the waveguide arm, forming a three-dimensional coupling structure that is three-dimensionally coupled to the evanescent field surrounding the waveguide arm.

2. The electro-optic modulator according to claim 1, characterized in that, A hexagonal boron nitride insulating layer is disposed on the outer side of the silicon dioxide modulation arm away from the waveguide arm, and the thickness of the hexagonal boron nitride insulating layer is 20-50 nm.

3. The electro-optic modulator according to claim 1, characterized in that, The vertical cross-section of the silicon dioxide modulation arm is rectangular. The top plane of the silicon dioxide modulation arm is symmetrically arranged with the silicon dioxide insulating layer, and the side plane is perpendicular to the silicon dioxide insulating layer.

4. The electro-optic modulator according to claim 1, characterized in that, The thickness of the silicon dioxide modulation arm is 200-300 nm.

5. A modulation method for the electro-optic modulator according to any one of claims 1-4, characterized in that, Includes the following steps: S1: Input a first single-mode optical signal in the 1550nm band into the input waveguide of the electro-optic modulator. After the first single-mode optical signal enters the input waveguide, it is split into two second single-mode optical signals. The two second single-mode optical signals are transmitted along two sets of symmetrically arranged waveguide arms respectively. During the transmission process, each second single-mode optical signal will form an evanescent wave of the optical field around the corresponding waveguide arm. S2: Based on the silica-graphene composite structure on both sides of the waveguide arm, the evanescent wave of the optical field penetrates to the surface of the silica-graphene composite structure and undergoes near-field coupling with the graphene layer in the silica-graphene composite structure facing the waveguide arm, thereby realizing the efficient interaction between the optical field and the graphene layer. S3: Build an LSTM timing error compensation model adapted to the three-dimensional coupling structure. The model is trained using the system's measured three-dimensional coupled optical field timing sample dataset. The model is then iteratively optimized using a weighted mean square loss function to obtain an LSTM timing error compensation model that adapts to the electro-optic modulator error correction and meets the modulation accuracy requirements. S4: Real-time acquisition of driving electrical signal parameters, environmental parameters and optical feedback parameters, input of multi-source timing parameters into the trained LSTM timing error compensation model, correction of nonlinear error and timing drift through model calculation, output of two differential driving electrical signals with voltage correction and frequency fine-tuning; S5: The differential driving electrical signal is applied to the graphene layer in the silica-graphene composite structure on the outside of the two sets of waveguide arms by means of preset metal electrodes. The graphene carrier concentration and Fermi level are changed by electrical modulation. The absorption and refractive index characteristics of the optical field are changed by means of the electro-optic effect of graphene, so that the optical signals transmitted in the two sets of waveguide arms generate differential phase / intensity modulation, thereby achieving high-precision and stable electro-optic modulation.

6. The modulation method according to claim 5, characterized in that, In step S3, the training process of the LSTM time-series error compensation model is as follows: S301: Construct an LSTM temporal error compensation model, which includes an input layer, a bidirectional LSTM hidden layer, a three-dimensional light field feature attention layer, a fully connected layer, and an output layer; wherein, the number of forward hidden units and backward hidden units in the bidirectional LSTM hidden layer are 32-64 respectively; the three-dimensional light field feature attention layer adopts an additive attention mechanism to assign adaptive weights to the three-dimensional coupled optical feedback parameters; the number of neurons in the fully connected layer is 8-16, and the ReLU activation function is used; S302: Select the actual measured time-series sample dataset of the system as training data, use the driving electrical signal parameters, environmental parameters, and optical feedback parameters as model input parameters, and use the electrical signal voltage correction and frequency fine-tuning corresponding to the model input parameters as training labels; during the training process, the weighted mean square error loss function is used to measure the deviation between the predicted value and the label value, and the Adam optimizer is selected and combined with the cosine annealing learning rate scheduling strategy and early stopping mechanism to iteratively train the LSTM time-series error compensation model until the prediction accuracy of the LSTM time-series error compensation model meets the preset modulation requirements, and the trained LSTM time-series error compensation model is obtained.

7. The modulation method according to claim 6, characterized in that, In step S302, the driving electrical signal includes voltage, frequency, and phase; the environmental parameters include temperature and humidity; and the optical feedback parameters include absorption coefficient and refractive index feedback values.

8. A preparation method, characterized in that, The method for preparing the electro-optic modulator according to any one of claims 1-4 includes the following steps: Step 1: Select a single-crystal silicon wafer of a specified thickness as the silicon substrate. Use the RCA standard cleaning method to remove organic matter, metal impurities and natural oxide layer from the surface of the silicon substrate. After drying, it is ready for use. Use plasma-enhanced chemical vapor deposition technology to deposit a silicon dioxide insulating layer on the surface of the silicon substrate. By controlling the deposition temperature and radio frequency power, the thickness of the silicon dioxide insulating layer is made to reach the preset value. Step 2: Photoresist is coated on the surface of the silicon dioxide insulating layer using electron beam lithography. After exposure and development, waveguide patterns are formed, including an input waveguide, an input beam splitter, two sets of symmetrically arranged waveguide arms, an output coupler, and an output waveguide. Using the photoresist as a mask, waveguide structures corresponding to the waveguide patterns are etched on the silicon dioxide insulating layer by dry etching, ensuring that the width of the waveguide arms meets the preset width range and the spacing between the two sets of symmetrically arranged waveguide arms meets the first preset spacing range. After etching, the residual photoresist is removed. Step 3: Deposit silicon dioxide material on the surface of the silicon dioxide insulating layer on both sides of the waveguide arm using plasma-enhanced chemical vapor deposition (PECVD) to construct a silicon dioxide modulation arm; use dry etching to trim the edges of the silicon dioxide modulation arm so that the side of the silicon dioxide modulation arm is perpendicular to the silicon dioxide insulating layer, and the top plane is parallel to the silicon dioxide insulating layer, and ensure that the distance between the inner side of the silicon dioxide modulation arm facing the waveguide arm and the side of the waveguide arm facing the silicon dioxide modulation arm conforms to a second preset distance range, and the length of the silicon dioxide modulation arm is consistent with the length of the straight segment of the waveguide arm; Step 4: Prepare monolayer graphene by mechanical exfoliation. Using PMMA-assisted transfer process, first perform oxygen plasma activation treatment on the inner side of the silicon dioxide modulation arm, then attach and fix the defect-free monolayer graphene to the inner side of the silicon dioxide modulation arm. After constant temperature baking, adhesive removal, cleaning, and nitrogen drying, the preparation and integration of the graphene layer on the inner side of the silicon dioxide modulation arm is completed. Step 5: Using the same graphene layer preparation process as in Step 4, fabricate a graphene-PMMA composite film. Transfer the graphene-PMMA composite film to the top plane of the silica modulation arm. After desizing, cleaning, and morphology inspection, the preparation of the graphene layer on the top plane of the silica modulation arm is completed. Step 6: Obtain hBN sheets using mechanical exfoliation. Apply the hBN sheets to the outer side of the silicon dioxide modulation arm away from the waveguide arm using dry transfer technology, serving as an insulating protection and carrier modulation layer for the graphene layer. Deposit Cr / Au metal electrodes on the surface of the hBN sheets using electron beam evaporation technology. One end of the Cr / Au metal electrode is in ohmic contact with the graphene layer, while the other end extends to the edge of the electro-optic modulator for connecting an external driving electrical signal source.

9. The preparation method according to claim 8, characterized in that, The specific steps of step 4 are as follows: Step 401: Repeatedly peel off the highly oriented pyrolytic graphite block with transparent adhesive to obtain a single-layer graphene sheet; screen the graphene sheets that are wrinkle-free, pore-free, and large enough to completely cover the inner side of the silica modulation arm using an optical microscope. Step 402: Place the screened single-layer graphene sheet in a PMMA solution and bake it on a heating table to make PMMA evenly cover the surface of the single-layer graphene sheet to form a graphene-PMMA composite film; then peel the graphene-PMMA composite film off the transparent adhesive residue layer and soak it in deionized water to completely remove the bottom adhesive tape residue. Step 403: Using a micro-alignment system, precisely align the graphene side of the graphene-PMMA composite film with the inner side of the silicon dioxide modulation arm near the waveguide arm. Use a micromanipulator to press and adhere the film, ensuring that the graphene completely covers the entire area of ​​the inner side of the silicon dioxide modulation arm. After transfer, place the entire sample on a heating stage for baking, so that the graphene and the inner side of the silicon dioxide modulation arm are tightly bonded together by van der Waals forces. Step 404: Immerse the sample in acetone solution for ultrasonic treatment to dissolve and completely remove the PMMA support layer; then rinse the sample surface multiple times with isopropanol and dry it with nitrogen gas to complete the preparation of the graphene layer on the inner side of the silicon dioxide modulation arm.

10. The preparation method according to claim 9, characterized in that, The specific steps of step 5 are as follows: Step 501: Repeat steps 401 and 402 to prepare another graphene-PMMA composite film whose size is completely matched with the top plane size of the silicon dioxide modulation arm, and ensure that the graphene-PMMA composite film is free from structural defects such as cracks and wrinkles. Step 502: Using a micro-alignment system, the graphene side of the graphene-PMMA composite film is placed downwards and precisely covers the top plane of the silicon dioxide modulation arm. This ensures that the edges of the graphene in the graphene-PMMA composite film are aligned with the edges of the top plane of the silicon dioxide modulation arm without any offset. A heating stage is then used for heat drying and leveling to eliminate air bubbles at the contact interface and strengthen the bonding force between the graphene and the top plane of the silicon dioxide modulation arm. Step 503: Immerse the sample in acetone solution and let it stand for a while until the PMMA is completely dissolved and removed. Then rinse it several times with deionized water and dry it with nitrogen. Microscopic characterization and detection using atomic force microscopy ensured that the surface roughness of the graphene met the preset requirements and that there were no residual solvents or impurities, thus completing the preparation of the graphene layer on the top plane of the silica modulation arm.

Images