Intelligent method for preparing graphene

Abstract

The invention relates to an intelligent method for preparing graphene. The method comprises the following steps: selecting a carbon-rich material; performing ultrasonic stripping on the carbon-rich material to generate graphene sheets; adjusting stripping parameters to form a synthetic parameter data set; characterizing the generated graphene sheets by using an X-ray diffraction (XRD) analysis method, a Fourier transform infrared spectroscopy (FTIR) analysis method and a carbon-oxygen ratio measurement method respectively to form a graphene characterization data set; the method comprises the following steps: randomly generating a large batch of graphene molecular models with random structures, and carrying out molecular dynamics simulation on each graphene molecular model to obtain corresponding simulation data; identifying a graphene molecular model matched with each piece of experimental characterization data from a large batch of graphene molecular models, and creating a data library; required graphene structure parameters are input, and the deep learning system retrieves corresponding stripping parameters from the database; according to the method, the preparation efficiency is improved, and the high-quality graphene sheet can be prepared more efficiently.

Description

Technical Field

[0001] The invention relates to the technical field of graphene preparation, and in particular to an intelligent method for preparing graphene. Background Art

[0002] As a new type of two-dimensional material, graphene has attracted widespread attention in the fields of electronics, energy, materials science, etc. due to its excellent electrical, thermal and mechanical properties. However, there are still many challenges in the large-scale preparation of graphene with specific properties according to specific needs. Traditional graphene preparation methods, such as chemical vapor deposition and mechanical exfoliation, usually require complex process steps and strict control conditions, with low preparation efficiency and difficulty in large-scale production. Although the ultrasonic exfoliation method is relatively simple and has the potential for large-scale preparation, it is necessary to continuously adjust parameters such as ultrasonic power, time, solvent and temperature to obtain graphene with specific properties (such as specific functional groups, number of layers and size), and repetitive experiments make the operation process complicated and time-consuming.

[0003] The differences in graphene's functional groups, number of layers, and size lead to significant differences in its properties, which in turn affect its specific applications. Since each graphene with a specific property has specific X-ray diffraction (XRD) and Fourier transform infrared spectroscopy (FTIR) characterization results, the existing technology usually studies the properties of graphene through XRD and FTIR characterization. However, since XRD and FTIR in the experiment are a collective description of the functional group content and functional group distribution. For example, the carbon-oxygen single bonds of different functional groups such as carboxyl and epoxy groups are all at 1000cm -1 There are absorption peaks at the wavenumber positions. Therefore, it is difficult to quantitatively and qualitatively describe the functional group content and functional group distribution only through experimental XRD and FTIR characterization.

[0004] Existing molecular dynamics simulation technology can simulate and calculate XRD and FTIR spectra based on a specific graphene molecular structure model and fit them with experimental data, thereby obtaining a molecular structure model that can describe the experimentally prepared graphene, and then accurately obtain functional groups, number of layers and size information. However, the existing traditional empirical potential function has limitations in accuracy, making it difficult to accurately predict the complex properties of graphene, and the calculation results are far from the experimental results. Although the existing density functional theory (DFT) performs well in simulation accuracy, the self-consistent field convergence calculation based on charge density is large, making it difficult to use for simulation calculations of graphene models with larger molecular weights. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide an intelligent method for preparing graphene, which can accurately characterize the physical and chemical properties of graphene and improve the preparation efficiency and quality of graphene.

[0006] The technical solution adopted by the present invention is an intelligent method for preparing graphene, which comprises the following steps:

[0007] S1. Select carbon-rich materials;

[0008] S2, performing ultrasonic exfoliation on the carbon-rich material selected in step S1 to generate graphene sheets;

[0009] S3, adjusting the peeling parameters to generate graphene sheets corresponding to different peeling parameters; recording the peeling parameters adjusted each time to form a synthetic parameter data set;

[0010] S4, characterizing the graphene sheet generated in step S3 using an XRD analysis method, an FTIR analysis method, and a carbon-oxygen ratio measurement method, respectively, to obtain experimental characterization data for each graphene sample, wherein the experimental characterization data includes characterizing an XRD spectrum, characterizing an FTIR spectrum, and characterizing a carbon-oxygen ratio data;

[0011] S5, recording the experimental characterization data corresponding to each graphene sample to form a graphene characterization data set;

[0012] S6, associating the synthesis parameter dataset and the graphene characterization dataset, and inputting them into the deep learning system;

[0013] S7, using the potential function of the deep learning system to randomly generate a large number of graphene molecular models with random structures;

[0014] S8, performing molecular dynamics simulation on each of the graphene molecular models to obtain corresponding simulation data, wherein the simulation data includes a simulated XRD spectrum, a simulated FTIR spectrum, and a simulated carbon-oxygen ratio;

[0015] S9, based on the simulation data obtained in step S8, identifying a graphene molecular model that matches each experimental characterization data from the large batch of graphene molecular models obtained in step S7;

[0016] S10, associating the identified graphene molecular model with the synthesis parameter data set, and creating a database including exfoliation parameters and structural parameters of the graphene molecular model according to the identified graphene molecular model; the structural parameters of the graphene molecular model include graphene functional groups, number of layers and size;

[0017] S11. The user inputs the required graphene structure parameters, and the deep learning system retrieves the corresponding exfoliation parameters from the database, and prepares the required graphene sheet according to the retrieved exfoliation parameters.

[0018] The beneficial effects of the present invention are as follows: through intelligent design and optimization of stripping parameters, parameters such as ultrasonic power, time, solvent and temperature are automatically adjusted, the number of experiments and time are reduced, the preparation efficiency is improved, and high-quality graphene sheets can be prepared more efficiently; deep learning potential functions are used for molecular dynamics simulation, and its accuracy is close to density functional theory (DFT), which greatly improves the reliability of simulation results; compared with traditional potential function methods, deep learning potential functions can more accurately capture the physical and chemical properties of graphene and reduce simulation errors; through high-precision simulation, the structure and properties of graphene can be better predicted and understood, providing an important reference for experimental design; a preparation parameter-graphene property relationship database is established, and users can input the desired graphene properties, and the system will automatically provide optimized preparation parameters, which greatly improves the controllability and repeatability of the graphene preparation process and saves researchers' time and resources.

[0019] Preferably, in step S9, the specific process of identifying the graphene molecular model matching each experimental characterization data from the large batch of graphene molecular models obtained in step S7 comprises the following steps:

[0020] S7.1. For each characterization XRD spectrum and each characterization FTIR spectrum, a characterization XRD spectrum and a characterization FTIR spectrum obtained by experimental characterization constitute an experimental XRD text file and an experimental FTIR text file, respectively; all simulated XRD spectra and all simulated FTIR spectra obtained by simulation constitute a simulated XRD text file and a simulated FTIR text file, respectively; the experimental XRD text file contains a 2×N type XRD matrix, the simulated XRD text file contains respective 2×N type and M×2×N type XRD matrices, the experimental FTIR text file contains a 2×N type FTIR matrix, and the simulated FTIR text file contains an M×2×N type FTIR matrix;

[0021] S7.2, by the four-point difference method, the first-order derivatives of all the simulated M×2×N matrices and all the 2×N matrices obtained in step S7.1 are calculated, and then the second-order derivatives are calculated by the five-point difference method to obtain the second-order derivative results corresponding to each matrix; according to the second-order derivative results, the maximum point and the minimum point are judged respectively to obtain a peak position representing the XRD spectrum, a peak position representing the FTIR spectrum, the peak position of each simulated XRD spectrum, and the peak position of each simulated FTIR spectrum;

[0022] S7.3, compare the peak positions of the XRD spectrum obtained in step S7.2 with the peak positions of each simulated XRD spectrum, and if the difference is less than ±0.1°, compare the peak positions of the FTIR spectrum obtained in step S7.2 with the peak positions of each simulated FTIR spectrum, and if the difference is less than ±5 cm -1 , then it is considered that the simulation data matches the experimental characterization data, and the graphene molecular model corresponding to the simulation data matches the experimental characterization data. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flow chart of an intelligent method for preparing graphene according to the present invention;

[0024] Figure 2 It is a schematic diagram of the accuracy characteristics of the deep learning potential function and the training convergence effect in the present invention;

[0025] Figure 3 Schematic diagram of the experimental process for carrying out specific experiments in the present invention. DETAILED DESCRIPTION

[0026] The invention will be further described below with reference to the accompanying drawings and in combination with specific implementations, so that those skilled in the art can implement the invention with reference to the description. The protection scope of the invention is not limited to the specific implementations.

[0027] The present invention relates to an intelligent method for preparing graphene, which comprises the following steps:

[0028] 1. Prepare synthetic parameter dataset:

[0029] Step 1.1: Select carbon-rich material

[0030] 1) The user selects carbon-rich materials: such as graphite, graphite oxide, biochar, carbon black, etc. One or more raw materials can be selected according to the user's research needs;

[0031] 2) Record the type information of the selected carbon-rich material;

[0032] Step 1.2: Perform ultrasonic exfoliation and adjust exfoliation parameters

[0033] 1) Ultrasonic exfoliation is performed on the selected carbon-rich material to generate graphene sheets: In a specific embodiment, an ultrasonic treatment device is used, and the initial parameters are set as follows: ultrasonic power is 100-500W, and the treatment time is 30-120 minutes.

[0034] 2) Adjust the stripping parameters: adjust the ultrasonic power in the range of 100-500W in increments of 50W; adjust the ultrasonic time in the range of 30-12 minutes in increments of 10 minutes; adjust the type of solvent used, such as water, ethanol, acetone, etc.; adjust the stripping temperature or product drying temperature.

[0035] 3) Record the peeling parameters after each adjustment to form a synthetic parameter data set.

[0036] 2. Preparation of graphene characterization dataset

[0037] Step 2.1: Characterize graphene

[0038] 1) Characterize each graphene sheet produced using the following methods: X-ray diffraction (XRD) analysis, Fourier transform infrared spectroscopy (FTIR) analysis, measurement of the carbon to oxygen ratio (e.g., by X-ray photoelectron spectroscopy (XPS)).

[0039] Step 2.2: Document key features

[0040] 1) Record the XRD spectrum, FTIR spectrum and carbon-oxygen ratio data of each graphene sample to form a graphene characterization data set.

[0041] 2) Associating the graphene characterization data set with the synthesis parameter data set, that is, a set of graphene characterization data corresponds to a set of stripping parameters.

[0042] 3. Deep learning and molecular dynamics simulation

[0043] Step 3.1: Input characterization data

[0044] 1) Input the synthesis parameter dataset and graphene characterization dataset into the deep learning system.

[0045] Step 3.2: Simulate using deep learning potential field

[0046] 1) Using the deep learning potential function trained by the system, a large number of graphene molecular models with random structures are generated and molecular dynamics (MD) simulations are performed to obtain the simulated XRD spectrum and simulated FTIR spectrum corresponding to each graphene molecular model, and the carbon-oxygen ratio of the molecular model is recorded.

[0047] Step 3.3: Identify the molecular structure

[0048] 1) Based on the simulation results (XRD, FTIR, carbon-oxygen ratio), identify the graphene molecular structure that matches the experimental characterization data (XRD, FTIR, carbon-oxygen ratio). The following describes the process for matching a characterization XRD spectrum and a characterization FTIR spectrum: a characterization XRD spectrum and a characterization FTIR spectrum obtained by experimental characterization constitute an experimental XRD text file and an experimental FTIR text file, respectively, and all simulated XRD spectra (M) and all simulated FTIR spectra (M) obtained by simulation constitute a simulated XRD text file and a simulated FTIR text file, respectively, the experimental XRD text file contains a 2×N type XRD matrix, the simulated XRD text file contains respective 2×N type and M×2×N type XRD matrices, the experimental FTIR text file contains a 2×N type FTIR matrix, and the simulated FTIR text file contains an M×2×N type FTIR matrix; wherein N represents the number of sampling scatter points for sampling the characterization XRD spectrum or the characterization FTIR spectrum; M represents the number of simulated XRD spectra or simulated FTIR spectra;

[0049] By using the four-point difference method, first-order derivatives are calculated for all simulated M×2×N matrices and all 2×N matrices obtained in step S7.1, and then second-order derivatives are calculated by using the five-point difference method to obtain the second-order derivative results corresponding to each matrix; according to the second-order derivative results, maximum and minimum points are judged to obtain a peak position representing an XRD spectrum, a peak position representing an FTIR spectrum, the peak position of each simulated XRD spectrum, and the peak position of each simulated FTIR spectrum;

[0050] S7.3, compare the peak positions of the XRD spectrum obtained in step S7.2 with the peak positions of each simulated XRD spectrum, and if the difference is less than ±0.1°, compare the peak positions of the FTIR spectrum obtained in step S7.2 with the peak positions of each simulated FTIR spectrum, and if the difference is less than ±5 cm -1 , then it is considered that the simulation data matches the experimental characterization data, and the graphene molecular model corresponding to the simulation data matches the experimental characterization data.

[0051] 2) Correlate the matching graphene molecular structure with its exfoliation parameters and experimental characterization data.

[0052] 4. Generate mapping relationships and optimize production parameters

[0053] Step 4.1: Create a database of preparation parameters and graphene properties

[0054] 1) Based on the identified graphene molecular structure, a database is created containing the relationship between the exfoliation parameters and the structural parameters of the graphene molecular model. The structural parameters of the graphene molecular model include graphene functional groups (such as hydroxyl, carboxyl, epoxy, etc.), number of layers, and size.

[0055] Step 4.2: Input desired graphene properties

[0056] 1) The user inputs the desired graphene properties, such as specific hydroxyl or carboxyl content, number of layers, and size.

[0057] 2) The system retrieves relevant data from the prepared parameter-graphene property relationship database.

[0058] Step 4.3: Provide optimization parameters

[0059] 1) The system provides optimized ultrasonic power, time, solvent selection and reaction temperature according to the mapping relationship in the database to meet the user's requirements for graphene properties, thereby preparing the required graphene.

[0060] The present invention uses intelligent design and optimization of synthesis parameters to automatically adjust parameters such as ultrasonic power, time, solvent and temperature, thereby reducing the number and time of experiments, improving preparation efficiency, and being able to more efficiently prepare high-quality graphene sheets.

[0061] The deep learning system of the present invention provides a self-developed deep learning potential function with accuracy close to DFT, such as Figure 2 As shown, during the training process, as the training step increases, the root mean square error of the deep learning model predicting the atomic forces gradually decreases and tends to be stable. Ultimately, the accuracy of the deep learning potential function is improved by more than 30% compared to the empirical potential function commonly used at this stage, approaching 90% of the DFT calculation accuracy. The deep learning potential function is used for molecular dynamics simulation, and its accuracy is close to that of density functional theory (DFT), which greatly improves the reliability of the simulation results. Compared with the traditional potential function method, the deep learning potential function can more accurately capture the physical and chemical properties of graphene and reduce simulation errors. Through high-precision simulation, the structure and properties of graphene can be better predicted and understood, providing an important reference for experimental design. The deep learning potential function package of the present invention is frozen into a binary program that can be used by open source molecular dynamics software.

[0062] In the present invention, a preparation parameter-graphene property relationship database is established, and the user can input the desired graphene properties, and the system will automatically provide optimized preparation parameters.

[0063] The present invention optimizes the graphene preparation method and characterization technology, so that high-quality graphene can be applied to multiple fields such as electronics, energy, and environment more quickly. It provides a comprehensive graphene preparation and characterization solution to help researchers and engineers achieve more significant results in graphene application development.

[0064] The effectiveness of the intelligent method for preparing graphene of the present invention is illustrated by specific experiments below.

[0065] First, based on a 100W, 40kHz ultrasonic generator, 5mg of graphite oxide was added to a 500ml deionized water environment. Then, the graphene oxide was collected through a 200nm pore size filter membrane using a suction filtration device, and the obtained graphene oxide will be in a paste state. The obtained graphene oxide was divided into six portions, placed in culture dishes, and allowed to stand in an oven at different ambient temperatures (20°C, 40°C, 60°C, 80°C, 100°C, 120°C) for 24 hours. The dried graphene oxide was taken out and characterized by XRD and FTIR, and then the graphene oxide molecular model was generated in batches and molecular dynamics simulation based on deep learning potential function was performed. The XRD spectrum and FTIR spectrum obtained by simulation calculation were compared with those obtained by experiment to obtain a graphene oxide molecular model that can represent the experimental results. According to the number and distribution of functional groups in the visualized molecular model, a mapping relationship between the drying temperature and the number of functional groups of the product (such as hydroxyl groups) was established. Finally, the directional regulation of the functional groups of graphene oxide was achieved through heating control at different temperatures. Figure 3 As shown, different graphene products are obtained by heating at different temperatures. Through experimental characterization and using the system provided by the present invention to generate a search molecular model, the relationship between the hydroxyl content and the heating temperature is obtained, thereby achieving directional regulation of the graphene hydroxyl functional group. Figure 3 It can be seen that the method provided by the present invention can greatly improve the controllability and repeatability of the graphene preparation process, saving researchers' time and resources.

Claims

1. An intelligent method for preparing graphene, characterized in that: The method comprises the following steps: S1. Select carbon-rich materials; S2, performing ultrasonic exfoliation on the carbon-rich material selected in step S1 to generate graphene sheets; S3, adjusting the peeling parameters to generate corresponding graphene sheets under different peeling parameters; Record the peeling parameters adjusted each time to form a synthetic parameter data set; S4, characterizing the graphene sheet generated in step S3 using an XRD analysis method, an FTIR analysis method, and a carbon-oxygen ratio measurement method, respectively, to obtain experimental characterization data for each graphene sample, wherein the experimental characterization data includes characterizing an XRD spectrum, characterizing an FTIR spectrum, and characterizing a carbon-oxygen ratio data; S5, recording the experimental characterization data corresponding to each graphene sample to form a graphene characterization data set; S6, associating the synthesis parameter dataset and the graphene characterization dataset, and inputting them into the deep learning system; S7, using the potential function of the deep learning system to randomly generate a large number of graphene molecular models with random structures; S8, performing molecular dynamics simulation on each of the graphene molecular models to obtain corresponding simulation data, wherein the simulation data includes a simulated XRD spectrum, a simulated FTIR spectrum, and a simulated carbon-oxygen ratio; S9, based on the simulation data obtained in step S8, identifying a graphene molecular model that matches each experimental characterization data from the large batch of graphene molecular models obtained in step S7; S10, associating the identified graphene molecular model with the synthesis parameter data set, and creating a database including exfoliation parameters and structural parameters of the graphene molecular model according to the identified graphene molecular model; the structural parameters of the graphene molecular model include graphene functional groups, number of layers and size; S11. The user inputs the required graphene structure parameters, and the deep learning system retrieves the corresponding exfoliation parameters from the database, and prepares the required graphene sheet according to the retrieved exfoliation parameters.

2. An intelligent method for preparing graphene according to claim 1, characterized in that: In step S9, the specific process of identifying the graphene molecular model matching each experimental characterization data from the large batch of graphene molecular models obtained in step S7 comprises the following steps: S7.1, for each characterization XRD spectrum and each characterization FTIR spectrum: a characterization XRD spectrum and a characterization FTIR spectrum obtained by experimental characterization constitute an experimental XRD text file and an experimental FTIR text file respectively, all simulated XRD spectra and all simulated FTIR spectra obtained by simulation constitute a simulated XRD text file and a simulated FTIR text file respectively, the experimental XRD text file contains a 2×N type XRD matrix, the simulated XRD text file contains respective 2×N type and M×2×N type XRD matrices, the experimental FTIR text file contains a 2×N type FTIR matrix, and the simulated FTIR text file contains an M×2×N type FTIR matrix; wherein N represents the number of sampling scatter points for sampling the characterization XRD spectrum or the characterization FTIR spectrum; M represents the number of simulated XRD spectra or simulated FTIR spectra; S7.2, by the four-point difference method, the first-order derivatives of all the simulated M×2×N matrices and all the 2×N matrices obtained in step S7.1 are calculated, and then the second-order derivatives are calculated by the five-point difference method to obtain the second-order derivative results corresponding to each matrix; according to the second-order derivative results, the maximum point and the minimum point are judged respectively to obtain a peak position representing the XRD spectrum, a peak position representing the FTIR spectrum, the peak position of each simulated XRD spectrum, and the peak position of each simulated FTIR spectrum; S7.3, compare the peak positions of the XRD spectrum obtained in step S7.2 with the peak positions of each simulated XRD spectrum, and if the difference is less than ±0.1°, compare the peak positions of the FTIR spectrum obtained in step S7.2 with the peak positions of each simulated FTIR spectrum, and if the difference is less than ±5 cm -1 , then it is considered that the simulation data matches the experimental characterization data, and the graphene molecular model corresponding to the simulation data matches the experimental characterization data.

Images