Low threshold plasmonic graphene all-optical nonlinear activator and preparation method

Abstract

The application relates to a low-threshold plasmonic graphene all-optical nonlinear activator, which comprises a substrate, a gold film is covered on the substrate, a device pattern is etched on the gold film, and a graphene film is covered on the gold film. The gold film pattern of the plasmonic structure can generate a very strong local field enhancement effect, significantly enhancing the interaction between light and graphene; by utilizing the Pauli blocking effect and the ultrafast carrier dynamics of graphene, combining the local field enhancement effect of the plasmonic structure to strengthen the light-matter interaction, low-threshold nonlinear activation is realized. The device activation threshold is as low as 7.03 nW, and the accuracy in MNIST handwritten digit recognition is 96.7%; by eliminating the photoelectric conversion link, a low-power-consumption, high-energy-efficiency optical domain processing scheme is provided, and the practicability of optical neural network calculation is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of semiconductor device technology, and in particular to a low-threshold plasmonic graphene all-optical nonlinear activator and its preparation method. Background Technology

[0002] Nonlinear activators are core components of optical neural networks (ONNs), used to simulate the nonlinear response characteristics of biological neurons to achieve complex data mapping and feature extraction. Their working principles mainly fall into two categories:

[0003] Electro-optic mechanism: Optical signals are converted into electrical signals using a photodetector, then nonlinear functions such as Sigmoid and ReLU are implemented using electronic circuits (e.g., CMOS), and finally the processed signal is converted back into an optical signal using an electro-optic modulator. This process relies on photoelectric conversion and is prone to introducing delays and energy consumption.

[0004] All-optical mechanism: Directly utilizes the nonlinear optical effects of materials (such as saturated absorption, plasma dispersion, phase transition, etc.) to achieve optical-optical nonlinear mapping.

[0005] In recent years, all-optical nonlinear activators designed based on two-dimensional materials (possessing strong nonlinear optical response and ultrafast response speed characteristics) have been extensively studied. Two-dimensional materials such as MXene and transition metal sulfides provide a physical basis for all-optical nonlinear activators due to their ultrafast carrier dynamics (picosecond relaxation time) and strong nonlinear optical properties (such as saturable absorption). By combining these materials with photonic integrated devices (such as microring resonators and ridge waveguides), high-speed nonlinear modulation has been successfully realized.

[0006] Based on MXene (Ti3C2T) x The nonlinear activator, by combining the saturable absorption properties of microfibers and materials, has achieved efficient nonlinear mapping in optical neural networks and has been successfully applied to super-resolution image reconstruction tasks. Molybdenum disulfide (MoS2), with its saturable absorption properties and high nonlinear optical response, has achieved an accuracy of 89% to 94% in MNIST handwritten digit classification simulation experiments, providing key material support for the realization of all-optical neural networks. Chen et al. realized an all-optical nonlinear activation function by integrating MoTe2 / optical waveguide devices. This device has a low threshold (0.94 μW) and ultrafast response (2.08 THz), achieving an accuracy of 97.6% in MNIST handwritten digit recognition.

[0007] However, the aforementioned nonlinear activators have the following problems: First, the inherent weak nonlinear response of the optical medium makes it difficult to support efficient activation functions; second, traditional electro-optic hybrid schemes severely restrict system performance due to photoelectric conversion delay (nanosecond level) and excessive energy consumption.

[0008] Therefore, there is an urgent need for a low-threshold all-optical nonlinear activator that can solve the above problems. Summary of the Invention

[0009] The present invention aims to provide a low-threshold plasmonic graphene all-optical nonlinear activator and its preparation method to overcome the shortcomings of the prior art. The technical problem to be solved by the present invention is achieved through the following technical solution.

[0010] According to a first aspect of this application, a low-threshold plasmonic graphene all-optical nonlinear activator is provided, comprising a substrate, a gold film covering the substrate, a device pattern etched on the gold film, and a graphene film covering the gold film.

[0011] Preferably, the device pattern is a plasmonic structure pattern, and the size of the plasmonic structure is 10.97 × 5.69 μm².

[0012] According to a second aspect of this application, a method for fabricating the aforementioned low-threshold plasmonic graphene all-optical nonlinear activator is provided, comprising the following steps:

[0013] Step 1: Deposit a gold film on a silicon dioxide substrate;

[0014] Step 2: Use focused ion beam etching to etch a device pattern onto the gold film;

[0015] Step 3: Use dry transfer technology to transfer graphene material onto the patterned and etched substrate to complete the preparation.

[0016] Preferably, in step 1, a gold film is deposited using thermal evaporation technology in a vacuum environment of 1 x 10⁻⁶ Pa.

[0017] Preferably, the deposition rate during the gold film deposition process is 1 Å / s.

[0018] Preferably, in step 2, the specific method for etching to form the device pattern includes:

[0019] Step 21: Fix the silicon dioxide substrate into the FIB chamber and use an electron beam to locate the target area;

[0020] Step 22: Load the preset semi-circular grating metasurface pattern and etch within the selected area;

[0021] Step 23: During the etching process, observe in real time using electron beam imaging and adjust parameters as needed until etching is complete.

[0022] Preferably, in step 23, the FIB etching current is 24 pA.

[0023] Preferably, in step 3, the specific method for transferring graphene material onto the patterned and etched substrate using dry transfer technology is as follows:

[0024] Step 31: Separate the graphene film from the graphene bulk crystal using adhesive tape;

[0025] Step 32: Fix the graphene film onto the triaxial displacement stage; fix the patterned substrate onto the transfer platform.

[0026] Step 33: Use a triaxial displacement stage to bring the graphene film into contact with the substrate;

[0027] Step 34: Heat the substrate to allow the graphene film to adhere to the substrate, thus completing the transfer of the graphene material.

[0028] Preferably, in step 32, after attaching the required graphene film material using PDMS fixed on the glass slide, the glass slide is fixed onto the triaxial displacement stage.

[0029] Preferably, in step 34, the substrate is heated using a heating device to reduce the adhesion between the graphene film and PDMS.

[0030] The embodiments of the present invention have the following advantages:

[0031] The low-threshold plasmonic graphene all-optical nonlinear activator provided in this invention utilizes the Pauli blockade effect and ultrafast carrier dynamics of graphene, combined with the local field enhancement effect of the plasmonic structure, to strengthen the light-matter interaction and achieve low-threshold nonlinear activation. This device has an activation threshold as low as 7.03 nW and achieves 96.7% accuracy in MNIST handwritten digit recognition. By eliminating the photoelectric conversion stage, it provides a low-power, high-efficiency optical domain processing solution, significantly improving the practicality of optical neural network computation. Attached Figure Description

[0032] Figure 1a This is a schematic diagram of a low-threshold plasmonic graphene all-optical nonlinear activator structure according to the present invention.

[0033] Figure 1b Figure 1 shows the SEM image of the low-threshold plasmonic graphene all-optical nonlinear activator.

[0034] Figure 2 a is a schematic diagram of the test device for the low-threshold plasmonic graphene all-optical nonlinear activator in Figure 1;

[0035] Figure 2 b is the Raman spectrum of graphene during the test;

[0036] Figure 2 c is the device's saturation absorption threshold;

[0037] Figure 2 d is a comparison chart of device saturation absorption thresholds;

[0038] Figure 3 a is a diagram showing the relationship between absorption and incidence;

[0039] Figure 3 b is the SA graph for the activation function;

[0040] Figure 4a This is a diagram of the neural network structure.

[0041] Figure 4b Confusion matrix diagram for identification results. Detailed Implementation

[0042] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0043] It should be noted that the above detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0044] like Figure 1a and Figure 1b As shown, a low-threshold plasmonic graphene all-optical nonlinear activator in one embodiment of this application includes a substrate, a gold film covered on the substrate, a device pattern etched on the gold film, and a graphene film covered on the gold film.

[0045] The device pattern is a plasmonic structure pattern, with dimensions of 10.97 × 5.69 μm². The plasmonic structure (gold film pattern) can generate a very strong local field enhancement effect, significantly enhancing the interaction between light and graphene; graphene possesses the Pauli blockade effect and ultrafast carrier dynamics, which, combined with the plasmonic structure, achieves an extremely low-power nonlinear response.

[0046] Through this embodiment, graphene, by leveraging the Pauli blockade effect and ultrafast carrier dynamics, combined with the local field enhancement effect of plasmon structures, significantly enhances the intensity of light-matter interaction, achieving low-threshold nonlinear activation.

[0047] like Figure 1b Scanning electron microscope images of the device clearly demonstrate the morphology and precision of the plasmonic structure (semi-circular grating metasurface) etched on the gold film. The high precision of the structure fabrication and good pattern consistency are beneficial to the stability and repeatability of optical performance. The mature integration process of the plasmonic structure with graphene makes large-scale integration possible.

[0048] The fabrication method of the aforementioned low-threshold plasmonic graphene all-optical nonlinear activator includes the following steps:

[0049] Step 1: Deposit a gold film on a silicon dioxide substrate;

[0050] In this step, a commercial SiO2 substrate and Au with a purity of 99.999% are used. The gold film is prepared using a thermal evaporation technique, depositing gold onto an ultra-flat SiO2 substrate. This process is carried out at 1x10⁻⁶ ppm. -6 The deposition was carried out in a vacuum environment of Pa, with real-time monitoring using a film thickness gauge, and at a stable rate of 1 Å / s to ensure uniform and high-quality gold films.

[0051] Step 2: Use focused ion beam etching to etch a device pattern onto the gold film;

[0052] In this step, focused ion beam (FIB) etching technology is used to etch the pattern of the designed device.

[0053] First, the substrate is fixed to the stage using conductive adhesive, and the FIB chamber is maintained in a high vacuum state. Before etching, the target area is precisely positioned using an electron beam;

[0054] After positioning is complete, adjust the sample stage to 52 degrees to ensure ion beam alignment.

[0055] Subsequently, a pre-defined semi-circular grating metasurface pattern is loaded, and etching is performed in the selected area. During the etching process, the effect is observed in real time using electron beam imaging, and parameters are adjusted as needed to optimize the results.

[0056] Step 3: Use dry transfer technology to transfer graphene material onto the patterned and etched substrate to complete the preparation.

[0057] In this step, graphene material is transferred onto a patterned and etched substrate using a dry transfer technique. The specific method is as follows:

[0058] Step 31: Place the graphene block crystals on the cut 3M tape and fold the tape repeatedly to separate the desired graphene film.

[0059] Step 32: Use PDMS fixed on the glass slide to attach an appropriate amount of graphene film material and fix it on the triaxial displacement stage;

[0060] Step 33: Place the patterned substrate in the vacuum adsorption position of the transfer platform for fixation to prevent deviation during the transfer process;

[0061] Step 34: Using an optical microscope system and a triaxial precision displacement stage, the graphene is precisely positioned at the target location and its height is adjusted to make it contact the target substrate;

[0062] Step 35: Use a heating device to heat the target substrate to weaken the adhesion between graphene and PDMS. After lifting the glass slide, the graphene can be left on the target substrate to complete the entire transfer process.

[0063] The low-threshold plasmonic graphene all-optical nonlinear activator prepared by the above method was tested for performance in the following manner:

[0064] like Figure 2 As shown in Figure a, the experimental optical path configuration for optical testing of the device includes: a pulsed laser (800 nm, 1 kHz), a spectrometer, a sample stage, a power meter, and other testing equipment; the use of pulsed laser testing verifies the nonlinear response of the device at ultra-low power; all-optical testing does not require electrical conversion, highlighting its "all-optical" processing advantage.

[0065] During the experiment: A pulsed laser (model Mai Tai) was used as the signal source. 800nm ​​1kHz pulsed light was directed onto the device through a designed optical path. A power meter was used to detect changes in the light power to determine if the device had reached saturation absorption. Specific experimental data are as follows:

[0066] like Figure 2 The Raman characteristic peaks (G peak, 2D peak, etc.) of the graphene shown in b confirm that the graphene is of good quality and has no obvious defects or uneven doping. High-quality graphene is the foundation for achieving a low-threshold nonlinear response; Raman spectroscopy proves that the transfer process was successful and the material structure is intact.

[0067] like Figure 2 As shown in Figure c, the curve of absorptivity versus incident light intensity yields a saturation absorption threshold of only 7.03 nW. This threshold is extremely low, three orders of magnitude lower than existing devices based on materials such as MoS2 and MoTe2; making it suitable for ultra-low power optical neural network systems.

[0068] like Figure 2 As shown in Figure d, comparing the saturation absorption thresholds of this device with those of a graphene device without plasmon structures clearly demonstrates the performance improvement brought about by the plasmon structure. The plasmon structure reduces the threshold from the microwatt level to the nanowatt level, resulting in a significant performance improvement.

[0069] This experiment validates the effectiveness of plasmon-graphene co-design.

[0070] Based on experimental measurement data, the relationship between the absorptivity α and the incident light intensity I is obtained, and the following formula is used for calculation:

[0071] α(I)=α ns +(α s I s / (I+I s)),

[0072] Where α ns This is nonsaturated absorption, meaning loss independent of intensity, α s For saturated absorption, i.e., the modulation depth of the material, I s Saturation strength is the strength required to reduce absorption to half of its unbleached value under steady-state conditions.

[0073] like Figure 3 a and Figure 3 b. The relationship between transmittance T and incident light intensity I is obtained using the following formula:

[0074] T(I)=1-α(I)=1-α ns -(α s I s / (I+I s ));

[0075] Then, the nonlinear transfer curve is used as the nonlinear activation function SA of the nonlinear activator, specifically:

[0076] f(I) = I*T(I) = I*(1-α) ns -(α s I s / (I+I s ))).

[0077] pass Figure 3 As can be seen from b, the device exhibits typical saturated absorption behavior, making it suitable for simulating neuron activation functions; it can directly realize Sigmoid-type nonlinear mapping in the optical domain without photoelectric conversion.

[0078] Then, it was verified through simulation:

[0079] like Figure 4a As shown, in order to verify the feasibility of using plasmonic graphene all-optical nonlinear activator for optical neural network, the SA nonlinear activation function was used in artificial neural network to realize the recognition of MNIST handwritten characters.

[0080] A four-layer fully connected neural network was constructed, and the SA function obtained by measurement was selected as the nonlinear activation function. The four-layer network structure is as follows: one input layer (784 neurons), two hidden layers (200 and 32 neurons respectively), and one output layer (10 neurons).

[0081] 50,000 images were used for training, 10,000 images were used for testing, and finally a confusion matrix was calculated on the test set. Figure 4bThe MNIST confusion matrix is ​​presented, demonstrating the results of applying this device as a nonlinear activation function to a four-layer fully connected neural network for MNIST handwritten digit recognition. It can be seen that the recognition accuracy for the ten digits 0-9 reaches over 90%, with an average accuracy of 96.7%, verifying the feasibility of the nonlinear activator. The accuracy is comparable to existing electrical solutions, but the power consumption is extremely low and the latency is extremely short, showing significant energy efficiency advantages.

[0082] The aforementioned all-optical nonlinear activator, designed in conjunction with the saturable absorption properties of graphene and plasmonic metasurfaces, utilizes the Pauli blockade effect of graphene and ultrafast carrier dynamics, combined with the local field enhancement effect of plasmonic structures, to strengthen light-matter interactions and achieve low-threshold nonlinear activation.

[0083] Through experimental verification, the device has an activation threshold as low as 7.03 nW and achieves an accuracy of 96.7% in MNIST handwritten digit recognition. By eliminating the photoelectric conversion link, it provides a low-power, high-efficiency optical domain processing solution, which significantly improves the practicality of optical neural network computing.

[0084] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0085] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that the embodiments of this application described herein can be implemented in sequences other than those illustrated or described herein.

[0086] Furthermore, the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or apparatus.

[0087] For ease of description, spatial relative terms such as "above," "on top of," "on the upper surface of," "above," etc., are used herein to describe the spatial positional relationship of a device or feature as shown in the figures to other devices or features. It should be understood that spatial relative terms are intended to encompass different orientations in use or operation beyond the orientation of the device as described in the figures. For example, if the device in the figures were inverted, a device described as "above" or "on top of" other devices or structures would subsequently be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below." The device may also be positioned in other different ways, such as rotated 90 degrees or in other orientations, and the spatial relative descriptions used herein will be interpreted accordingly.

[0088] In the detailed description above, reference has been made to the accompanying drawings, which form part of this document. In the drawings, similar symbols typically identify similar parts unless the context otherwise indicates otherwise. The illustrated embodiments described in the detailed specification, drawings, and claims are not intended to be limiting. Other embodiments may be used and other changes may be made without departing from the spirit or scope of the subject matter presented herein.

[0089] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., 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 low-threshold plasmonic graphene all-optical nonlinear activator, characterized in that, The device includes a substrate covered with a gold film, on which a device pattern is etched. A graphene film is also covered on the gold film. The device pattern is a plasmonic structure pattern of a semi-circular grating metasurface, and the size of the plasmonic structure is 10.97 × 5.69 μm. 2 .

2. A method for fabricating the low-threshold plasmonic graphene all-optical nonlinear activator according to claim 1, characterized in that, Includes the following steps: Step 1: Deposit a gold film on a silicon dioxide substrate; Step 2: Use focused ion beam etching to etch a device pattern onto the gold film; Step 3: Use dry transfer technology to transfer graphene material onto the patterned and etched substrate to complete the preparation.

3. The preparation method according to claim 2, characterized in that, In step 1, in 1x10 -6 Gold films were deposited in a vacuum environment of Pa using thermal evaporation technology.

4. The preparation method according to claim 3, characterized in that, During the gold film deposition process, the deposition rate was 1 Å / s.

5. The preparation method according to claim 2, characterized in that, In step 2, the specific method for etching to form the device pattern includes: Step 21: Fix the silicon dioxide substrate into the chamber of the focused ion beam equipment and use the electron beam to locate the target area; Step 22: Load the preset semi-circular grating metasurface pattern and etch within the selected area; Step 23: During the etching process, observe in real time using electron beam imaging and adjust parameters as needed until etching is complete.

6. The preparation method according to claim 5, characterized in that, In step 23, the etching current of the focused ion beam device is 24 pA.

7. The preparation method according to claim 3, characterized in that, In step 3, the specific method for transferring graphene material onto the patterned and etched substrate using dry transfer technology is as follows: Step 31: Separate the graphene film from the graphene bulk crystal using adhesive tape; Step 32: Fix the graphene film onto the triaxial displacement stage; fix the patterned substrate onto the transfer platform. Step 33: Use a triaxial displacement stage to bring the graphene film into contact with the substrate; Step 34: Heat the substrate to allow the graphene film to adhere to the substrate, thus completing the transfer of the graphene material.

8. The preparation method according to claim 7, characterized in that, In step 32, after attaching the required graphene film material using the polydimethylsiloxane layer fixed on the glass slide, the glass slide is fixed onto the triaxial displacement stage.

9. The preparation method according to claim 8, characterized in that, In step 34, the substrate is heated using a heating device to reduce the adhesion between the graphene film and the polydimethylsiloxane layer.

Images